From 852109a31f7e1cbe680e95cf8a793c9f4ee63f28 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 18:00:45 +0800 Subject: [PATCH 01/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 1 + 1 file changed, 1 insertion(+) create mode 100644 src/Unmanned_Aircraft_System/README.md diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md new file mode 100644 index 0000000000..8620aae73b --- /dev/null +++ b/src/Unmanned_Aircraft_System/README.md @@ -0,0 +1 @@ +使用强化学习和模拟器进行无人机/无人驾驶飞机路 径规划 \ No newline at end of file From 6fe47f51daaa1ade9e5dcdab13d7928bccb402e9 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 18:08:41 +0800 Subject: [PATCH 02/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index 8620aae73b..62f5b7cc72 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路 径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From af14f998a9070884c7cd8c97c1b2aa54b101c45e Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 18:11:18 +0800 Subject: [PATCH 03/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index 62f5b7cc72..bd2f2f59ca 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From c1ae4acbbef045b93e2c188222df56ebde143228 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 18:14:42 +0800 Subject: [PATCH 04/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index bd2f2f59ca..62f5b7cc72 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From ddaa9d522d82660b22dda5fd46466d0b009f6ce0 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 18:15:54 +0800 Subject: [PATCH 05/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index 62f5b7cc72..bd2f2f59ca 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From d66962dc189cb673507c7c23cc5a30cc268b08e0 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 18:20:49 +0800 Subject: [PATCH 06/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index bd2f2f59ca..62f5b7cc72 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From 9293e17535979018e820d0fb5feb1f4eb775cc8c Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 18:23:15 +0800 Subject: [PATCH 07/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index 62f5b7cc72..bd2f2f59ca 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From e8fef316bc62a650a2f87294dba71e840670be83 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 18:27:09 +0800 Subject: [PATCH 08/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index bd2f2f59ca..62f5b7cc72 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From 7a9fab5526fdd74c47be19721805b231891899d7 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 18:50:42 +0800 Subject: [PATCH 09/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index 62f5b7cc72..bd2f2f59ca 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From 79cc5314e7877fa7f6df8af5500bd0108a121d4e Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 18:51:48 +0800 Subject: [PATCH 10/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index bd2f2f59ca..62f5b7cc72 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From 29f77705706b6019d217b13d23af6c639bc60b5c Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 19:29:47 +0800 Subject: [PATCH 11/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index 62f5b7cc72..bd2f2f59ca 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From 4b1aca1e95f41e2141387c92a6487d8a7027ff28 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 19:41:49 +0800 Subject: [PATCH 12/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index bd2f2f59ca..62f5b7cc72 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From edcd60e09ca3e17b693a474637aae2bbc88eaf5a Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 19:43:49 +0800 Subject: [PATCH 13/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index 62f5b7cc72..bd2f2f59ca 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From e43731a1b15a6760ba58613b26ce8ee5d348d7f1 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 19:46:43 +0800 Subject: [PATCH 14/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index bd2f2f59ca..62f5b7cc72 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From 0f8575a199751c6514077f110c7b4825eb5cede0 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 19:50:09 +0800 Subject: [PATCH 15/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index 62f5b7cc72..bd2f2f59ca 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From 60db4b42c3ffbab3936bc0caff6c3ba3a2f05b7c Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 19:51:38 +0800 Subject: [PATCH 16/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index bd2f2f59ca..62f5b7cc72 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From d9535bae15e08ac997a25c9e277c9128b2591229 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 18 Dec 2025 22:43:17 +0800 Subject: [PATCH 17/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index 62f5b7cc72..bd2f2f59ca 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file From 78869806c6590301db64df7a712ce2336412a730 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Fri, 19 Dec 2025 00:11:07 +0800 Subject: [PATCH 18/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=EF=BC=9A=E6=9B=B4=E6=96=B0Unmanned=5FAircraft=5FSystem?= =?UTF-8?q?=E7=9A=84README.md?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index bd2f2f59ca..645fb4d244 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1 +1,2 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 \ No newline at end of file +使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 + From b4985379cce2f25beda21ede15815c56b9b0d9bd Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Fri, 19 Dec 2025 14:28:06 +0800 Subject: [PATCH 19/45] =?UTF-8?q?=E6=8F=90=E4=BA=A4=E9=80=89=E9=A2=98?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index 645fb4d244..34b39fbed7 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1,2 +1,2 @@ -使用强化学习和模拟器进行无人机/无人驾驶飞机路径规划 +基于 YOLOv8 的无人机地面简易目标识别（如行人 / 小型车辆） From bc149727947fd991bdbb476179644bc62b823cd0 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Fri, 19 Dec 2025 16:56:24 +0800 Subject: [PATCH 20/45] =?UTF-8?q?=E4=BF=AE=E6=94=B9=E9=80=89=E9=A2=98?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md index 34b39fbed7..6f3bf20ddc 100644 --- a/src/Unmanned_Aircraft_System/README.md +++ b/src/Unmanned_Aircraft_System/README.md @@ -1,2 +1,2 @@ -基于 YOLOv8 的无人机地面简易目标识别（如行人 / 小型车辆） +无人机航拍图像的小目标识别算法研究 From c0679cbcd98c710f74249ff454e4e7d9a608577b Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Fri, 19 Dec 2025 17:28:53 +0800 Subject: [PATCH 21/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Unmanned_Aircraft_System/README.md | 2 -- 1 file changed, 2 deletions(-) delete mode 100644 src/Unmanned_Aircraft_System/README.md diff --git a/src/Unmanned_Aircraft_System/README.md b/src/Unmanned_Aircraft_System/README.md deleted file mode 100644 index 6f3bf20ddc..0000000000 --- a/src/Unmanned_Aircraft_System/README.md +++ /dev/null @@ -1,2 +0,0 @@ -无人机航拍图像的小目标识别算法研究 - From 2d1e296235088c7f2036bb31121a8a5b399ed66e Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Fri, 19 Dec 2025 17:37:27 +0800 Subject: [PATCH 22/45] =?UTF-8?q?=E9=87=8D=E5=91=BD=E5=90=8D=E6=96=87?= =?UTF-8?q?=E4=BB=B6=E5=A4=B9=EF=BC=9AUnmanned=5FAircraft=5FSystem=20?= =?UTF-8?q?=E2=86=92=20UAV=5FSmall=5FTarget=5FRecognition?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/UAV_Small_Target_Recognition/README.md | 3 +++ 1 file changed, 3 insertions(+) create mode 100644 src/UAV_Small_Target_Recognition/README.md diff --git a/src/UAV_Small_Target_Recognition/README.md b/src/UAV_Small_Target_Recognition/README.md new file mode 100644 index 0000000000..c9ac2e269b --- /dev/null +++ b/src/UAV_Small_Target_Recognition/README.md @@ -0,0 +1,3 @@ +无人机航拍图像的小目标识别算法 + + From d49e121e292de767c0d6312f0c5b3c9b29be7b44 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Fri, 19 Dec 2025 18:01:45 +0800 Subject: [PATCH 23/45] =?UTF-8?q?=E5=AE=8C=E6=88=90=E9=80=89=E9=A2=98?= =?UTF-8?q?=E5=B9=B6=E6=8F=90=E4=BA=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/UAV_Small_Target_Recognition/README.md | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/src/UAV_Small_Target_Recognition/README.md b/src/UAV_Small_Target_Recognition/README.md index c9ac2e269b..a05dec4c58 100644 --- a/src/UAV_Small_Target_Recognition/README.md +++ b/src/UAV_Small_Target_Recognition/README.md @@ -1,3 +1 @@ -无人机航拍图像的小目标识别算法 - - +无人机航拍图像的小目标识别算法 \ No newline at end of file From 2c572930f54bab14ac8e011a32afbbf50d45c476 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Fri, 19 Dec 2025 19:41:09 +0800 Subject: [PATCH 24/45] =?UTF-8?q?=E9=87=8D=E5=91=BD=E5=90=8D=E6=96=87?= =?UTF-8?q?=E4=BB=B6=E5=A4=B9=E4=B8=BAGesture=5FControl=5Ffor=5FAirsim=5FU?= =?UTF-8?q?AV=EF=BC=8C=E5=B9=B6=E6=9B=B4=E6=96=B0README.md=E8=A1=A5?= =?UTF-8?q?=E5=85=85=E6=97=A0=E4=BA=BA=E6=9C=BA=E6=89=8B=E5=8A=BF=E6=8E=A7?= =?UTF-8?q?=E5=88=B6=E5=86=85=E5=AE=B9?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Gesture_Control_for_Airsim_UAV/README.md | 2 ++ src/Gesture_Control_for_Airsim_UAV/main.py | 1 + src/UAV_Small_Target_Recognition/README.md | 1 - 3 files changed, 3 insertions(+), 1 deletion(-) create mode 100644 src/Gesture_Control_for_Airsim_UAV/README.md create mode 100644 src/Gesture_Control_for_Airsim_UAV/main.py delete mode 100644 src/UAV_Small_Target_Recognition/README.md diff --git a/src/Gesture_Control_for_Airsim_UAV/README.md b/src/Gesture_Control_for_Airsim_UAV/README.md new file mode 100644 index 0000000000..4f05828e92 --- /dev/null +++ b/src/Gesture_Control_for_Airsim_UAV/README.md @@ -0,0 +1,2 @@ +允许用户使用手势控制 Airsim 无人机 + diff --git a/src/Gesture_Control_for_Airsim_UAV/main.py b/src/Gesture_Control_for_Airsim_UAV/main.py new file mode 100644 index 0000000000..c6d640eaa5 --- /dev/null +++ b/src/Gesture_Control_for_Airsim_UAV/main.py @@ -0,0 +1 @@ +pressed) \ No newline at end of file diff --git a/src/UAV_Small_Target_Recognition/README.md b/src/UAV_Small_Target_Recognition/README.md deleted file mode 100644 index a05dec4c58..0000000000 --- a/src/UAV_Small_Target_Recognition/README.md +++ /dev/null @@ -1 +0,0 @@ -无人机航拍图像的小目标识别算法 \ No newline at end of file From d5f2c9d16e3500dce621d7ceb7965e2f7c5f0d6f Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Fri, 19 Dec 2025 21:16:23 +0800 Subject: [PATCH 25/45] =?UTF-8?q?=E9=87=8D=E5=91=BD=E5=90=8D=E6=96=87?= =?UTF-8?q?=E4=BB=B6=E5=A4=B9=EF=BC=9AGesture=5FControl=5Ffor=5FAirsim=5FU?= =?UTF-8?q?AV=20=E2=86=92=20Drone=5Fgesture=5Foperation?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../README.md | 0 src/Gesture_Control_for_Airsim_UAV/main.py | 1 - 2 files changed, 1 deletion(-) rename src/{Gesture_Control_for_Airsim_UAV => Drone_gesture_operation}/README.md (100%) delete mode 100644 src/Gesture_Control_for_Airsim_UAV/main.py diff --git a/src/Gesture_Control_for_Airsim_UAV/README.md b/src/Drone_gesture_operation/README.md similarity index 100% rename from src/Gesture_Control_for_Airsim_UAV/README.md rename to src/Drone_gesture_operation/README.md diff --git a/src/Gesture_Control_for_Airsim_UAV/main.py b/src/Gesture_Control_for_Airsim_UAV/main.py deleted file mode 100644 index c6d640eaa5..0000000000 --- a/src/Gesture_Control_for_Airsim_UAV/main.py +++ /dev/null @@ -1 +0,0 @@ -pressed) \ No newline at end of file From f1d316714ea393683f4e830c370d226ba183c967 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Fri, 19 Dec 2025 22:01:09 +0800 Subject: [PATCH 26/45] =?UTF-8?q?=E6=9B=B4=E6=96=B0Drone=5Fgesture=5Fopera?= =?UTF-8?q?tion/README.md=EF=BC=9A=E5=9F=BA=E4=BA=8E=E8=BD=BB=E9=87=8F?= =?UTF-8?q?=E7=BA=A7=20CNN=20=E7=9A=84=E6=97=A0=E4=BA=BA=E6=9C=BA=E6=89=8B?= =?UTF-8?q?=E5=8A=BF=E6=93=8D=E6=8E=A7=E6=8C=87=E4=BB=A4=E8=AF=86=E5=88=AB?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Drone_gesture_operation/README.md b/src/Drone_gesture_operation/README.md index 4f05828e92..3454d8e7dd 100644 --- a/src/Drone_gesture_operation/README.md +++ b/src/Drone_gesture_operation/README.md @@ -1,2 +1,2 @@ -允许用户使用手势控制 Airsim 无人机 +基于轻量级 CNN 的无人机手势操控指令识别 From 5a5639ca81060ddcba63d4cd8b20d16697f1074e Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Sat, 20 Dec 2025 16:08:12 +0800 Subject: [PATCH 27/45] =?UTF-8?q?=E6=97=A0=E4=BA=BA=E6=9C=BA=E6=89=8B?= =?UTF-8?q?=E5=8A=BF=E8=AF=86=E5=88=AB?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 109 ++++++++++++++++++++++++++++ 1 file changed, 109 insertions(+) create mode 100644 src/Drone_gesture_operation/main.py diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py new file mode 100644 index 0000000000..ca2ffe128d --- /dev/null +++ b/src/Drone_gesture_operation/main.py @@ -0,0 +1,109 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +import numpy as np +import tensorflow as tf +import os +import warnings +import pathlib + +# 关闭oneDNN提示 +os.environ['TF_ENABLE_ONEDNN_OPTS'] = '0' +# 关闭TF Lite警告 +warnings.filterwarnings('ignore', category=UserWarning, module='tensorflow.lite.python.interpreter') + + +class PointHistoryClassifier(object): + def __init__( + self, + model_path='model/point_history_classifier/point_history_classifier.tflite', + score_th=0.5, + invalid_value=0, + num_threads=1, + ): + if not os.path.isfile(model_path): + raise FileNotFoundError(f"模型文件不存在：{model_path}") + if not model_path.endswith('.tflite'): + raise ValueError(f"路径不是.tflite文件：{model_path}") + + with open(model_path, 'rb') as f: + model_data = f.read() + self.interpreter = tf.lite.Interpreter(model_content=model_data, num_threads=num_threads) + self.interpreter.allocate_tensors() + self.input_details = self.interpreter.get_input_details() + self.output_details = self.interpreter.get_output_details() + + self.score_th = score_th + self.invalid_value = invalid_value + print(f"✅ 模型加载成功！") + print(f"输入张量形状：{self.input_details[0]['shape']}") + print(f"输出张量形状：{self.output_details[0]['shape']}") + + def __call__(self, point_history): + if not isinstance(point_history, (list, np.ndarray)): + raise TypeError("输入必须是列表或numpy数组") + + input_data = np.array([point_history], dtype=np.float32) + if input_data.shape != tuple(self.input_details[0]['shape']): + raise ValueError( + f"输入形状不匹配！模型要求：{self.input_details[0]['shape']}，实际：{input_data.shape}" + ) + + self.interpreter.set_tensor(self.input_details[0]['index'], input_data) + self.interpreter.invoke() + + result = self.interpreter.get_tensor(self.output_details[0]['index']) + result_squeezed = np.squeeze(result) + result_index = np.argmax(result_squeezed) + + print(f"\n原始预测得分：{result_squeezed}") + print(f"最高得分索引：{result_index}，得分值：{result_squeezed[result_index]}") + + if result_squeezed[result_index] < self.score_th: + result_index = self.invalid_value + print(f"⚠️ 得分低于阈值({self.score_th})，返回无效值：{self.invalid_value}") + + return result_index + + +def preprocess_point_history(point_history): + """归一化关键点数据到0~1""" + point_history = np.array(point_history, dtype=np.float32) + min_val = np.min(point_history) + max_val = np.max(point_history) + # 避免除零 + point_history = (point_history - min_val) / (max_val - min_val + 1e-6) + return point_history + + +if __name__ == "__main__": + # 配置 + MODEL_PATH = pathlib.Path( + r"E:\无人机\dronehandgesture2023P1\model\point_history_classifier\point_history_classifier.tflite").resolve() + SCORE_THRESHOLD = 0.5 + # 手势映射（根据实际训练标签调整） + gesture_mapping = {0: "无手势/降落", 1: "起飞/前进"} + + print(f"当前模型路径：{MODEL_PATH}") + try: + # 实例化分类器 + classifier = PointHistoryClassifier( + model_path=str(MODEL_PATH), + score_th=SCORE_THRESHOLD, + num_threads=4 + ) + + # 1. 测试数据（随机生成32维，匹配模型输入） + test_point_history = np.random.rand(32).astype(np.float32) + # 2. 预处理（替换为真实数据时注释掉随机数，启用下面两行） + # real_point_history = [0.1,0.2,...,0.3] # 32个真实关键点数值 + # test_point_history = preprocess_point_history(real_point_history) + + # 分类推理 + result = classifier(test_point_history) + print(f"\n🎯 最终分类结果：{result} → {gesture_mapping.get(result, '未知手势')}") + + except Exception as e: + print(f"\n❌ 错误：{e}") + import traceback + + traceback.print_exc() \ No newline at end of file From 0ecd007042d36d15111f85915c6762860426dc10 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Sat, 20 Dec 2025 16:15:47 +0800 Subject: [PATCH 28/45] =?UTF-8?q?=E6=9B=B4=E6=96=B0Drone=5Fgesture=5Fopera?= =?UTF-8?q?tion=EF=BC=9A=E5=90=8C=E6=AD=A5main.py=E4=BB=A3=E7=A0=81?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/README.md | 1 - 1 file changed, 1 deletion(-) diff --git a/src/Drone_gesture_operation/README.md b/src/Drone_gesture_operation/README.md index 3454d8e7dd..93d5444e0d 100644 --- a/src/Drone_gesture_operation/README.md +++ b/src/Drone_gesture_operation/README.md @@ -1,2 +1 @@ 基于轻量级 CNN 的无人机手势操控指令识别 - From 786c5984bc6e56f89a7348408257575e783b1934 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Sat, 20 Dec 2025 16:20:07 +0800 Subject: [PATCH 29/45] =?UTF-8?q?=E6=97=A0=E4=BA=BA=E6=9C=BA=E6=89=8B?= =?UTF-8?q?=E5=8A=BF=E8=AF=86=E5=88=AB?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 355 ++++++++++++++++++++-------- 1 file changed, 256 insertions(+), 99 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index ca2ffe128d..fbbf1930c2 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -1,109 +1,266 @@ #!/usr/bin/env python # -*- coding: utf-8 -*- +import copy +import argparse +import itertools +from collections import Counter +from collections import deque +import time + +import cv2 as cv import numpy as np -import tensorflow as tf -import os -import warnings -import pathlib - -# 关闭oneDNN提示 -os.environ['TF_ENABLE_ONEDNN_OPTS'] = '0' -# 关闭TF Lite警告 -warnings.filterwarnings('ignore', category=UserWarning, module='tensorflow.lite.python.interpreter') - - -class PointHistoryClassifier(object): - def __init__( - self, - model_path='model/point_history_classifier/point_history_classifier.tflite', - score_th=0.5, - invalid_value=0, - num_threads=1, - ): - if not os.path.isfile(model_path): - raise FileNotFoundError(f"模型文件不存在：{model_path}") - if not model_path.endswith('.tflite'): - raise ValueError(f"路径不是.tflite文件：{model_path}") - - with open(model_path, 'rb') as f: - model_data = f.read() - self.interpreter = tf.lite.Interpreter(model_content=model_data, num_threads=num_threads) - self.interpreter.allocate_tensors() - self.input_details = self.interpreter.get_input_details() - self.output_details = self.interpreter.get_output_details() - - self.score_th = score_th - self.invalid_value = invalid_value - print(f"✅ 模型加载成功！") - print(f"输入张量形状：{self.input_details[0]['shape']}") - print(f"输出张量形状：{self.output_details[0]['shape']}") + +# ========== FPS计算 ========== +class CvFpsCalc: + def __init__(self, buffer_len=10): + self.buffer_len = buffer_len + self.times = deque(maxlen=buffer_len) + + def get(self): + self.times.append(time.perf_counter()) + if len(self.times) < 2: + return 0 + return int(len(self.times) / (self.times[-1] - self.times[0])) + + +# ========== 手势分类器（简化版） ========== +class KeyPointClassifier: + def __call__(self, landmark_list): + return 7 # 模拟点手势 + + +class PointHistoryClassifier: def __call__(self, point_history): - if not isinstance(point_history, (list, np.ndarray)): - raise TypeError("输入必须是列表或numpy数组") - - input_data = np.array([point_history], dtype=np.float32) - if input_data.shape != tuple(self.input_details[0]['shape']): - raise ValueError( - f"输入形状不匹配！模型要求：{self.input_details[0]['shape']}，实际：{input_data.shape}" - ) - - self.interpreter.set_tensor(self.input_details[0]['index'], input_data) - self.interpreter.invoke() - - result = self.interpreter.get_tensor(self.output_details[0]['index']) - result_squeezed = np.squeeze(result) - result_index = np.argmax(result_squeezed) - - print(f"\n原始预测得分：{result_squeezed}") - print(f"最高得分索引：{result_index}，得分值：{result_squeezed[result_index]}") - - if result_squeezed[result_index] < self.score_th: - result_index = self.invalid_value - print(f"⚠️ 得分低于阈值({self.score_th})，返回无效值：{self.invalid_value}") - - return result_index - - -def preprocess_point_history(point_history): - """归一化关键点数据到0~1""" - point_history = np.array(point_history, dtype=np.float32) - min_val = np.min(point_history) - max_val = np.max(point_history) - # 避免除零 - point_history = (point_history - min_val) / (max_val - min_val + 1e-6) - return point_history - - -if __name__ == "__main__": - # 配置 - MODEL_PATH = pathlib.Path( - r"E:\无人机\dronehandgesture2023P1\model\point_history_classifier\point_history_classifier.tflite").resolve() - SCORE_THRESHOLD = 0.5 - # 手势映射（根据实际训练标签调整） - gesture_mapping = {0: "无手势/降落", 1: "起飞/前进"} - - print(f"当前模型路径：{MODEL_PATH}") - try: - # 实例化分类器 - classifier = PointHistoryClassifier( - model_path=str(MODEL_PATH), - score_th=SCORE_THRESHOLD, - num_threads=4 + return 0 + + +# ========== 参数解析 ========== +def get_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--device", type=int, default=0) + parser.add_argument("--width", type=int, default=960) + parser.add_argument("--height", type=int, default=540) + return parser.parse_args() + + +# ========== 辅助函数 ========== +def select_mode(key, mode): + number = -1 + if 48 <= key <= 57: + number = key - 48 + if key == ord('n'): + mode = 0 + if key == ord('k'): + mode = 1 + if key == ord('h'): + mode = 2 + return number, mode + + +def calc_bounding_rect(image): + """模拟手部边界框（屏幕中心）""" + h, w = image.shape[:2] + cx, cy = w // 2, h // 2 + bw, bh = 200, 200 # 边界框大小 + return [cx - bw // 2, cy - bh // 2, cx + bw // 2, cy + bh // 2] + + +def calc_landmark_list(image): + """模拟21个手部关键点（适配原代码逻辑）""" + h, w = image.shape[:2] + cx, cy = w // 2, h // 2 + landmark_list = [] + + # 手掌中心（0号点） + landmark_list.append([cx, cy]) + + # 拇指（1-4号点） + landmark_list.append([cx - 50, cy - 30]) + landmark_list.append([cx - 80, cy - 60]) + landmark_list.append([cx - 100, cy - 90]) + landmark_list.append([cx - 110, cy - 110]) + + # 食指（5-8号点） + landmark_list.append([cx + 50, cy - 30]) + landmark_list.append([cx + 80, cy - 60]) + landmark_list.append([cx + 100, cy - 90]) + landmark_list.append([cx + 110, cy - 110]) # 8号点（点手势关键） + + # 中指（9-12号点） + landmark_list.append([cx + 30, cy - 10]) + landmark_list.append([cx + 50, cy - 40]) + landmark_list.append([cx + 70, cy - 70]) + landmark_list.append([cx + 80, cy - 90]) + + # 无名指（13-16号点） + landmark_list.append([cx + 10, cy + 10]) + landmark_list.append([cx + 20, cy - 20]) + landmark_list.append([cx + 30, cy - 50]) + landmark_list.append([cx + 40, cy - 70]) + + # 小指（17-20号点） + landmark_list.append([cx - 10, cy + 10]) + landmark_list.append([cx - 20, cy - 20]) + landmark_list.append([cx - 30, cy - 50]) + landmark_list.append([cx - 40, cy - 70]) + + return landmark_list + + +def pre_process_landmark(landmark_list): + temp = copy.deepcopy(landmark_list) + if not temp: + return [] + base_x, base_y = temp[0][0], temp[0][1] + for i in range(len(temp)): + temp[i][0] -= base_x + temp[i][1] -= base_y + temp = list(itertools.chain.from_iterable(temp)) + max_val = max(map(abs, temp)) if temp else 1 + return [x / max_val for x in temp] + + +def pre_process_point_history(image, point_history): + temp = copy.deepcopy(point_history) + if not temp: + return [] + base_x, base_y = temp[0][0], temp[0][1] + image_w, image_h = image.shape[1], image.shape[0] + for i in range(len(temp)): + temp[i][0] = (temp[i][0] - base_x) / image_w + temp[i][1] = (temp[i][1] - base_y) / image_h + return list(itertools.chain.from_iterable(temp)) + + +def draw_landmarks(image, landmark_list): + """绘制模拟关键点""" + if len(landmark_list) == 0: + return image + # 绘制手指连线 + links = [(2, 3), (3, 4), (5, 6), (6, 7), (7, 8), (9, 10), (10, 11), (11, 12), + (13, 14), (14, 15), (15, 16), (17, 18), (18, 19), (19, 20), + (0, 1), (1, 2), (2, 5), (5, 9), (9, 13), (13, 17), (17, 0)] + for (p1, p2) in links: + if p1 < len(landmark_list) and p2 < len(landmark_list): + cv.line(image, tuple(landmark_list[p1]), tuple(landmark_list[p2]), (0, 0, 0), 6) + cv.line(image, tuple(landmark_list[p1]), tuple(landmark_list[p2]), (255, 255, 255), 2) + # 绘制关键点 + for i, (x, y) in enumerate(landmark_list): + size = 8 if i in [4, 8, 12, 16, 20] else 5 + cv.circle(image, (x, y), size, (255, 255, 255), -1) + cv.circle(image, (x, y), size, (0, 0, 0), 1) + return image + + +def draw_bounding_rect(image, brect): + cv.rectangle(image, (brect[0], brect[1]), (brect[2], brect[3]), (0, 255, 0), 2) + return image + + +def draw_info_text(image, brect, hand_sign_text, finger_gesture_text): + cv.rectangle(image, (brect[0], brect[1] - 30), (brect[2], brect[1]), (0, 255, 0), -1) + info = f"Hand: {hand_sign_text}" + cv.putText(image, info, (brect[0] + 5, brect[1] - 5), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) + if finger_gesture_text: + cv.putText(image, f"Gesture: {finger_gesture_text}", (10, 60), + cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255), 3) + return image + + +def draw_point_history(image, point_history): + for i, (x, y) in enumerate(point_history): + if x != 0 and y != 0: + cv.circle(image, (x, y), 2 + i // 2, (0, 255, 0), -1) + return image + + +def draw_info(image, fps, mode, number): + cv.putText(image, f"FPS: {fps}", (10, 30), cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255), 3) + mode_text = ["Idle", "Log Keypoint", "Log Point History"][mode] if 0 <= mode <= 2 else "Idle" + cv.putText(image, f"Mode: {mode_text}", (10, 90), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) + if 0 <= number <= 9: + cv.putText(image, f"Num: {number}", (10, 120), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) + return image + + +# ========== 主函数 ========== +def main(): + args = get_args() + # 初始化摄像头 + cap = cv.VideoCapture(args.device) + cap.set(cv.CAP_PROP_FRAME_WIDTH, args.width) + cap.set(cv.CAP_PROP_FRAME_HEIGHT, args.height) + + # 初始化工具类 + cvFpsCalc = CvFpsCalc(buffer_len=10) + keypoint_classifier = KeyPointClassifier() + point_history_classifier = PointHistoryClassifier() + + # 标签和历史数据 + keypoint_labels = ["None", "Point", "Fist", "OK", "Peace", "ThumbUp", "ThumbDown", "PointGesture"] + point_history_labels = ["None", "MoveUp", "MoveDown", "MoveLeft", "MoveRight"] + history_length = 16 + point_history = deque(maxlen=history_length) + finger_gesture_history = deque(maxlen=history_length) + mode = 0 + + while True: + # FPS计算 + fps = cvFpsCalc.get() + + # 按键处理 + key = cv.waitKey(1) & 0xFF + if key == 27: # ESC退出 + break + number, mode = select_mode(key, mode) + + # 读取摄像头帧 + ret, frame = cap.read() + if not ret: + break + frame = cv.flip(frame, 1) # 镜像显示 + debug_frame = copy.deepcopy(frame) + + # 【核心修改】跳过真实检测，直接模拟手部数据 + brect = calc_bounding_rect(debug_frame) + landmark_list = calc_landmark_list(debug_frame) + + # 预处理 + pre_landmark = pre_process_landmark(landmark_list) + pre_point_history = pre_process_point_history(debug_frame, point_history) + + # 手势分类 + hand_sign_id = keypoint_classifier(pre_landmark) + point_history.append(landmark_list[8] if hand_sign_id == 7 else [0, 0]) + + # 手指手势分类 + finger_gesture_id = 0 + if len(pre_point_history) == history_length * 2: + finger_gesture_id = point_history_classifier(pre_point_history) + finger_gesture_history.append(finger_gesture_id) + most_common = Counter(finger_gesture_history).most_common(1) + + # 绘制UI + debug_frame = draw_bounding_rect(debug_frame, brect) + debug_frame = draw_landmarks(debug_frame, landmark_list) + debug_frame = draw_info_text( + debug_frame, brect, + keypoint_labels[hand_sign_id] if hand_sign_id < len(keypoint_labels) else "Unknown", + point_history_labels[most_common[0][0]] if most_common else "Unknown" ) - # 1. 测试数据（随机生成32维，匹配模型输入） - test_point_history = np.random.rand(32).astype(np.float32) - # 2. 预处理（替换为真实数据时注释掉随机数，启用下面两行） - # real_point_history = [0.1,0.2,...,0.3] # 32个真实关键点数值 - # test_point_history = preprocess_point_history(real_point_history) + # 绘制辅助信息 + debug_frame = draw_point_history(debug_frame, point_history) + debug_frame = draw_info(debug_frame, fps, mode, number) + + # 显示窗口 + cv.imshow('Hand Gesture Recognition (ESC to exit)', debug_frame) - # 分类推理 - result = classifier(test_point_history) - print(f"\n🎯 最终分类结果：{result} → {gesture_mapping.get(result, '未知手势')}") + # 释放资源 + cap.release() + cv.destroyAllWindows() - except Exception as e: - print(f"\n❌ 错误：{e}") - import traceback - traceback.print_exc() \ No newline at end of file +if __name__ == '__main__': + main() \ No newline at end of file From 2bd59549f4678f23489b060527b58dbfaeed6e73 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Sat, 20 Dec 2025 21:02:48 +0800 Subject: [PATCH 30/45] =?UTF-8?q?=E4=BC=98=E5=8C=96Drone=5Fgesture=5Fopera?= =?UTF-8?q?tion/main.py=EF=BC=9A=E7=BB=99=E4=BB=A3=E7=A0=81=E6=B7=BB?= =?UTF-8?q?=E5=8A=A0=E6=B3=A8=E9=87=8A=EF=BC=8C=E5=A2=9E=E5=8A=A0=E5=8F=AF?= =?UTF-8?q?=E8=AF=BB=E6=80=A7?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 283 ++++++++++++++++++---------- 1 file changed, 183 insertions(+), 100 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index fbbf1930c2..b5da6dc69d 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -1,5 +1,7 @@ #!/usr/bin/env python # -*- coding: utf-8 -*- +# 功能：无人机手势识别模拟程序（适配Python3.13+Windows） +# 说明：移除MediaPipe/TensorFlow依赖，通过模拟手部数据实现核心逻辑展示 import copy import argparse import itertools @@ -11,157 +13,219 @@ import numpy as np -# ========== FPS计算 ========== +# ===================== 工具类：FPS计算 ===================== class CvFpsCalc: + """帧率计算类，基于时间队列滑动平均计算FPS""" + def __init__(self, buffer_len=10): - self.buffer_len = buffer_len - self.times = deque(maxlen=buffer_len) + self.buffer_len = buffer_len # 缓存长度（计算最近N帧的平均FPS） + self.times = deque(maxlen=buffer_len) # 时间戳队列 def get(self): - self.times.append(time.perf_counter()) - if len(self.times) < 2: + """获取当前FPS值""" + self.times.append(time.perf_counter()) # 记录当前时间戳 + if len(self.times) < 2: # 至少需要2个时间戳才能计算 return 0 - return int(len(self.times) / (self.times[-1] - self.times[0])) + # 计算平均FPS：帧数 / 总时间 + fps = len(self.times) / (self.times[-1] - self.times[0]) + return int(fps) -# ========== 手势分类器（简化版） ========== +# ===================== 模拟分类器（适配原逻辑） ===================== class KeyPointClassifier: + """关键点分类器（简化版）""" + def __call__(self, landmark_list): - return 7 # 模拟点手势 + # 固定返回7（对应PointGesture点手势，模拟分类结果） + return 7 class PointHistoryClassifier: + """轨迹点分类器（简化版）""" + def __call__(self, point_history): + # 固定返回0（对应None，模拟分类结果） return 0 + # ===================== 参数解析 ===================== + -# ========== 参数解析 ========== def get_args(): - parser = argparse.ArgumentParser() - parser.add_argument("--device", type=int, default=0) - parser.add_argument("--width", type=int, default=960) - parser.add_argument("--height", type=int, default=540) + """解析命令行参数""" + parser = argparse.ArgumentParser(description="手部手势识别模拟程序") + parser.add_argument("--device", type=int, default=0, help="摄像头设备号（默认0）") + parser.add_argument("--width", type=int, default=960, help="摄像头画面宽度") + parser.add_argument("--height", type=int, default=540, help="摄像头画面高度") return parser.parse_args() -# ========== 辅助函数 ========== +# ===================== 核心辅助函数 ===================== def select_mode(key, mode): + """根据按键切换操作模式 + Args: + key: 按键值 + mode: 当前模式（0:空闲 1:记录关键点 2:记录轨迹点） + Returns: + number: 按键数字（0-9），-1表示非数字键 + mode: 更新后的模式 + """ number = -1 - if 48 <= key <= 57: + if 48 <= key <= 57: # 数字键0-9 number = key - 48 - if key == ord('n'): + if key == ord('n'): # n键：切换到空闲模式 mode = 0 - if key == ord('k'): + if key == ord('k'): # k键：切换到记录关键点模式 mode = 1 - if key == ord('h'): + if key == ord('h'): # h键：切换到记录轨迹点模式 mode = 2 return number, mode def calc_bounding_rect(image): - """模拟手部边界框（屏幕中心）""" + """生成模拟手部边界框（屏幕中心固定位置） + Args: + image: 摄像头帧画面 + Returns: + brect: 边界框坐标 [x1, y1, x2, y2] + """ h, w = image.shape[:2] - cx, cy = w // 2, h // 2 - bw, bh = 200, 200 # 边界框大小 + cx, cy = w // 2, h // 2 # 屏幕中心 + bw, bh = 200, 200 # 边界框尺寸 return [cx - bw // 2, cy - bh // 2, cx + bw // 2, cy + bh // 2] def calc_landmark_list(image): - """模拟21个手部关键点（适配原代码逻辑）""" + """生成模拟21个手部关键点（适配原代码21点逻辑） + Args: + image: 摄像头帧画面 + Returns: + landmark_list: 21个关键点坐标列表 [[x1,y1], [x2,y2], ...] + """ h, w = image.shape[:2] - cx, cy = w // 2, h // 2 + cx, cy = w // 2, h // 2 # 屏幕中心（关键点基准位置） landmark_list = [] - # 手掌中心（0号点） + # 0号点：手掌中心 landmark_list.append([cx, cy]) - - # 拇指（1-4号点） - landmark_list.append([cx - 50, cy - 30]) - landmark_list.append([cx - 80, cy - 60]) - landmark_list.append([cx - 100, cy - 90]) - landmark_list.append([cx - 110, cy - 110]) - - # 食指（5-8号点） - landmark_list.append([cx + 50, cy - 30]) - landmark_list.append([cx + 80, cy - 60]) - landmark_list.append([cx + 100, cy - 90]) - landmark_list.append([cx + 110, cy - 110]) # 8号点（点手势关键） - - # 中指（9-12号点） - landmark_list.append([cx + 30, cy - 10]) - landmark_list.append([cx + 50, cy - 40]) - landmark_list.append([cx + 70, cy - 70]) - landmark_list.append([cx + 80, cy - 90]) - - # 无名指（13-16号点） - landmark_list.append([cx + 10, cy + 10]) - landmark_list.append([cx + 20, cy - 20]) - landmark_list.append([cx + 30, cy - 50]) - landmark_list.append([cx + 40, cy - 70]) - - # 小指（17-20号点） - landmark_list.append([cx - 10, cy + 10]) - landmark_list.append([cx - 20, cy - 20]) - landmark_list.append([cx - 30, cy - 50]) - landmark_list.append([cx - 40, cy - 70]) + # 1-4号点：拇指 + landmark_list.extend([[cx - 50, cy - 30], [cx - 80, cy - 60], [cx - 100, cy - 90], [cx - 110, cy - 110]]) + # 5-8号点：食指（8号点为指尖，点手势关键） + landmark_list.extend([[cx + 50, cy - 30], [cx + 80, cy - 60], [cx + 100, cy - 90], [cx + 110, cy - 110]]) + # 9-12号点：中指 + landmark_list.extend([[cx + 30, cy - 10], [cx + 50, cy - 40], [cx + 70, cy - 70], [cx + 80, cy - 90]]) + # 13-16号点：无名指 + landmark_list.extend([[cx + 10, cy + 10], [cx + 20, cy - 20], [cx + 30, cy - 50], [cx + 40, cy - 70]]) + # 17-20号点：小指 + landmark_list.extend([[cx - 10, cy + 10], [cx - 20, cy - 20], [cx - 30, cy - 50], [cx - 40, cy - 70]]) return landmark_list def pre_process_landmark(landmark_list): + """关键点预处理：相对坐标转换+归一化（适配原逻辑） + Args: + landmark_list: 原始关键点列表 + Returns: + 预处理后的一维归一化列表 + """ temp = copy.deepcopy(landmark_list) - if not temp: + if not temp: # 空值保护 return [] + + # 相对坐标：以0号点（手掌中心）为基准 base_x, base_y = temp[0][0], temp[0][1] for i in range(len(temp)): temp[i][0] -= base_x temp[i][1] -= base_y + + # 一维化 + 归一化（消除尺度影响） temp = list(itertools.chain.from_iterable(temp)) - max_val = max(map(abs, temp)) if temp else 1 + max_val = max(map(abs, temp)) if temp else 1 # 除零保护 return [x / max_val for x in temp] def pre_process_point_history(image, point_history): + """轨迹点预处理：相对坐标转换+归一化（适配原逻辑） + Args: + image: 摄像头帧画面 + point_history: 轨迹点历史列表 + Returns: + 预处理后的一维归一化列表 + """ temp = copy.deepcopy(point_history) - if not temp: + if not temp: # 空值保护 return [] + + # 相对坐标：以第一个点为基准 base_x, base_y = temp[0][0], temp[0][1] image_w, image_h = image.shape[1], image.shape[0] for i in range(len(temp)): - temp[i][0] = (temp[i][0] - base_x) / image_w + temp[i][0] = (temp[i][0] - base_x) / image_w # 归一化到[0,1] temp[i][1] = (temp[i][1] - base_y) / image_h + + # 一维化 return list(itertools.chain.from_iterable(temp)) +# ===================== 绘制函数（UI展示） ===================== def draw_landmarks(image, landmark_list): - """绘制模拟关键点""" + """绘制手部关键点和连线 + Args: + image: 待绘制的画面 + landmark_list: 关键点列表 + Returns: + 绘制后的画面 + """ if len(landmark_list) == 0: return image - # 绘制手指连线 + + # 手指连线定义（关键点索引对） links = [(2, 3), (3, 4), (5, 6), (6, 7), (7, 8), (9, 10), (10, 11), (11, 12), (13, 14), (14, 15), (15, 16), (17, 18), (18, 19), (19, 20), (0, 1), (1, 2), (2, 5), (5, 9), (9, 13), (13, 17), (17, 0)] + + # 绘制连线：黑色粗线+白色细线（立体效果） for (p1, p2) in links: - if p1 < len(landmark_list) and p2 < len(landmark_list): + if p1 < len(landmark_list) and p2 < len(landmark_list): # 索引保护 cv.line(image, tuple(landmark_list[p1]), tuple(landmark_list[p2]), (0, 0, 0), 6) cv.line(image, tuple(landmark_list[p1]), tuple(landmark_list[p2]), (255, 255, 255), 2) - # 绘制关键点 + + # 绘制关键点：指尖8号/12号等用大圆点，其余用小圆点 for i, (x, y) in enumerate(landmark_list): size = 8 if i in [4, 8, 12, 16, 20] else 5 - cv.circle(image, (x, y), size, (255, 255, 255), -1) - cv.circle(image, (x, y), size, (0, 0, 0), 1) + cv.circle(image, (x, y), size, (255, 255, 255), -1) # 白色填充 + cv.circle(image, (x, y), size, (0, 0, 0), 1) # 黑色边框 return image def draw_bounding_rect(image, brect): + """绘制手部边界框 + Args: + image: 待绘制的画面 + brect: 边界框坐标 [x1,y1,x2,y2] + Returns: + 绘制后的画面 + """ cv.rectangle(image, (brect[0], brect[1]), (brect[2], brect[3]), (0, 255, 0), 2) return image def draw_info_text(image, brect, hand_sign_text, finger_gesture_text): + """绘制手势信息文本 + Args: + image: 待绘制的画面 + brect: 边界框坐标 + hand_sign_text: 手部手势文本 + finger_gesture_text: 手指轨迹手势文本 + Returns: + 绘制后的画面 + """ + # 绘制背景框（覆盖边界框上方） cv.rectangle(image, (brect[0], brect[1] - 30), (brect[2], brect[1]), (0, 255, 0), -1) + # 绘制手部手势标签 info = f"Hand: {hand_sign_text}" cv.putText(image, info, (brect[0] + 5, brect[1] - 5), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) + # 绘制轨迹手势标签 if finger_gesture_text: cv.putText(image, f"Gesture: {finger_gesture_text}", (10, 60), cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255), 3) @@ -169,95 +233,114 @@ def draw_info_text(image, brect, hand_sign_text, finger_gesture_text): def draw_point_history(image, point_history): + """绘制轨迹点历史（指尖移动轨迹） + Args: + image: 待绘制的画面 + point_history: 轨迹点列表 + Returns: + 绘制后的画面 + """ for i, (x, y) in enumerate(point_history): - if x != 0 and y != 0: + if x != 0 and y != 0: # 跳过无效点 + # 轨迹点大小随索引递增（视觉层次感） cv.circle(image, (x, y), 2 + i // 2, (0, 255, 0), -1) return image def draw_info(image, fps, mode, number): + """绘制全局信息（FPS、模式、数字） + Args: + image: 待绘制的画面 + fps: 当前帧率 + mode: 当前模式 + number: 当前数字 + Returns: + 绘制后的画面 + """ + # 绘制FPS cv.putText(image, f"FPS: {fps}", (10, 30), cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255), 3) + # 绘制模式 mode_text = ["Idle", "Log Keypoint", "Log Point History"][mode] if 0 <= mode <= 2 else "Idle" cv.putText(image, f"Mode: {mode_text}", (10, 90), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) + # 绘制数字 if 0 <= number <= 9: cv.putText(image, f"Num: {number}", (10, 120), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) return image -# ========== 主函数 ========== +# ===================== 主程序入口 ===================== def main(): + # 1. 初始化参数和资源 args = get_args() - # 初始化摄像头 - cap = cv.VideoCapture(args.device) - cap.set(cv.CAP_PROP_FRAME_WIDTH, args.width) - cap.set(cv.CAP_PROP_FRAME_HEIGHT, args.height) + cap = cv.VideoCapture(args.device) # 打开摄像头 + cap.set(cv.CAP_PROP_FRAME_WIDTH, args.width) # 设置宽度 + cap.set(cv.CAP_PROP_FRAME_HEIGHT, args.height) # 设置高度 # 初始化工具类 cvFpsCalc = CvFpsCalc(buffer_len=10) keypoint_classifier = KeyPointClassifier() point_history_classifier = PointHistoryClassifier() - # 标签和历史数据 + # 初始化标签和历史数据 keypoint_labels = ["None", "Point", "Fist", "OK", "Peace", "ThumbUp", "ThumbDown", "PointGesture"] point_history_labels = ["None", "MoveUp", "MoveDown", "MoveLeft", "MoveRight"] - history_length = 16 - point_history = deque(maxlen=history_length) - finger_gesture_history = deque(maxlen=history_length) - mode = 0 + history_length = 16 # 轨迹点缓存长度 + point_history = deque(maxlen=history_length) # 指尖轨迹缓存 + finger_gesture_history = deque(maxlen=history_length) # 手势分类结果缓存 + mode = 0 # 初始模式：空闲 + # 2. 主循环（摄像头帧处理） while True: - # FPS计算 + # 计算当前FPS fps = cvFpsCalc.get() - # 按键处理 + # 按键处理（ESC退出） key = cv.waitKey(1) & 0xFF - if key == 27: # ESC退出 + if key == 27: break number, mode = select_mode(key, mode) # 读取摄像头帧 ret, frame = cap.read() - if not ret: + if not ret: # 帧读取失败则退出 break - frame = cv.flip(frame, 1) # 镜像显示 - debug_frame = copy.deepcopy(frame) - - # 【核心修改】跳过真实检测，直接模拟手部数据 - brect = calc_bounding_rect(debug_frame) - landmark_list = calc_landmark_list(debug_frame) + frame = cv.flip(frame, 1) # 镜像翻转（符合视觉习惯） + debug_frame = copy.deepcopy(frame) # 用于绘制的帧副本 - # 预处理 - pre_landmark = pre_process_landmark(landmark_list) - pre_point_history = pre_process_point_history(debug_frame, point_history) + # 3. 核心逻辑：模拟手部数据生成 + 预处理 + 分类 + brect = calc_bounding_rect(debug_frame) # 生成边界框 + landmark_list = calc_landmark_list(debug_frame) # 生成关键点 + pre_landmark = pre_process_landmark(landmark_list) # 关键点预处理 + pre_point_history = pre_process_point_history(debug_frame, point_history) # 轨迹点预处理 # 手势分类 - hand_sign_id = keypoint_classifier(pre_landmark) + hand_sign_id = keypoint_classifier(pre_landmark) # 关键点分类 + # 记录指尖轨迹（点手势时记录8号点，否则记录无效点） point_history.append(landmark_list[8] if hand_sign_id == 7 else [0, 0]) - # 手指手势分类 + # 轨迹手势分类（缓存满16*2个点时分类） finger_gesture_id = 0 if len(pre_point_history) == history_length * 2: finger_gesture_id = point_history_classifier(pre_point_history) finger_gesture_history.append(finger_gesture_id) + # 取最频繁的手势分类结果（防抖） most_common = Counter(finger_gesture_history).most_common(1) - # 绘制UI - debug_frame = draw_bounding_rect(debug_frame, brect) - debug_frame = draw_landmarks(debug_frame, landmark_list) - debug_frame = draw_info_text( + # 4. UI绘制 + debug_frame = draw_bounding_rect(debug_frame, brect) # 绘制边界框 + debug_frame = draw_landmarks(debug_frame, landmark_list) # 绘制关键点 + debug_frame = draw_info_text( # 绘制手势信息 debug_frame, brect, keypoint_labels[hand_sign_id] if hand_sign_id < len(keypoint_labels) else "Unknown", point_history_labels[most_common[0][0]] if most_common else "Unknown" ) + debug_frame = draw_point_history(debug_frame, point_history) # 绘制轨迹 + debug_frame = draw_info(debug_frame, fps, mode, number) # 绘制全局信息 - # 绘制辅助信息 - debug_frame = draw_point_history(debug_frame, point_history) - debug_frame = draw_info(debug_frame, fps, mode, number) - - # 显示窗口 + # 显示画面 cv.imshow('Hand Gesture Recognition (ESC to exit)', debug_frame) - # 释放资源 + # 3. 资源释放 cap.release() cv.destroyAllWindows() From c6eea875f5e71cdf1ec50c38fd5f79b0bd7be137 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Sun, 21 Dec 2025 14:20:12 +0800 Subject: [PATCH 31/45] =?UTF-8?q?=E6=96=B0=E5=A2=9EDrone=5Fgesture=5Fopera?= =?UTF-8?q?tion/keypoint=5Fclassifier.py=EF=BC=9A=E6=97=A0=E4=BA=BA?= =?UTF-8?q?=E6=9C=BA=E6=89=8B=E5=8A=BF=E5=85=B3=E9=94=AE=E7=82=B9=E5=88=86?= =?UTF-8?q?=E7=B1=BB=E5=99=A8?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../keypoint_classifier.py | 36 +++++++++++++++++++ 1 file changed, 36 insertions(+) create mode 100644 src/Drone_gesture_operation/keypoint_classifier.py diff --git a/src/Drone_gesture_operation/keypoint_classifier.py b/src/Drone_gesture_operation/keypoint_classifier.py new file mode 100644 index 0000000000..3d055414e8 --- /dev/null +++ b/src/Drone_gesture_operation/keypoint_classifier.py @@ -0,0 +1,36 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +import numpy as np +import tensorflow as tf + + +class KeyPointClassifier(object): + def __init__( + self, + model_path='model/keypoint_classifier/keypoint_classifier.tflite', + num_threads=1, + ): + self.interpreter = tf.lite.Interpreter(model_path=model_path, + num_threads=num_threads) + + self.interpreter.allocate_tensors() + self.input_details = self.interpreter.get_input_details() + self.output_details = self.interpreter.get_output_details() + + def __call__( + self, + landmark_list, + ): + input_details_tensor_index = self.input_details[0]['index'] + self.interpreter.set_tensor( + input_details_tensor_index, + np.array([landmark_list], dtype=np.float32)) + self.interpreter.invoke() + + output_details_tensor_index = self.output_details[0]['index'] + + result = self.interpreter.get_tensor(output_details_tensor_index) + + result_index = np.argmax(np.squeeze(result)) + + return result_index From f2dc1afafb922ab284726927d662b4b5d49edfcc Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Mon, 22 Dec 2025 00:32:24 +0800 Subject: [PATCH 32/45] =?UTF-8?q?=E4=BC=98=E5=8C=96Drone=5Fgesture=5Fopera?= =?UTF-8?q?tion/main.py=EF=BC=9A=E5=8F=AF=E5=88=87=E6=8D=A2=E4=B8=89?= =?UTF-8?q?=E7=A7=8D=E6=A8=A1=E5=BC=8F?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 350 +++++++++++----------------- 1 file changed, 133 insertions(+), 217 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index b5da6dc69d..c64b2e7b55 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -1,7 +1,5 @@ #!/usr/bin/env python # -*- coding: utf-8 -*- -# 功能：无人机手势识别模拟程序（适配Python3.13+Windows） -# 说明：移除MediaPipe/TensorFlow依赖，通过模拟手部数据实现核心逻辑展示 import copy import argparse import itertools @@ -13,219 +11,150 @@ import numpy as np -# ===================== 工具类：FPS计算 ===================== +# ========== FPS计算（还原初始版本，无多余逻辑） ========== class CvFpsCalc: - """帧率计算类，基于时间队列滑动平均计算FPS""" - def __init__(self, buffer_len=10): - self.buffer_len = buffer_len # 缓存长度（计算最近N帧的平均FPS） - self.times = deque(maxlen=buffer_len) # 时间戳队列 + self.buffer_len = buffer_len + self.times = deque(maxlen=buffer_len) def get(self): - """获取当前FPS值""" - self.times.append(time.perf_counter()) # 记录当前时间戳 - if len(self.times) < 2: # 至少需要2个时间戳才能计算 + self.times.append(time.perf_counter()) + if len(self.times) < 2: return 0 - # 计算平均FPS：帧数 / 总时间 - fps = len(self.times) / (self.times[-1] - self.times[0]) - return int(fps) + return int(len(self.times) / (self.times[-1] - self.times[0])) -# ===================== 模拟分类器（适配原逻辑） ===================== +# ========== 手势分类器（还原初始版本） ========== class KeyPointClassifier: - """关键点分类器（简化版）""" - def __call__(self, landmark_list): - # 固定返回7（对应PointGesture点手势，模拟分类结果） - return 7 + return 7 # 模拟点手势 class PointHistoryClassifier: - """轨迹点分类器（简化版）""" - def __call__(self, point_history): - # 固定返回0（对应None，模拟分类结果） return 0 - # ===================== 参数解析 ===================== - +# ========== 参数解析（还原初始版本） ========== def get_args(): - """解析命令行参数""" - parser = argparse.ArgumentParser(description="手部手势识别模拟程序") - parser.add_argument("--device", type=int, default=0, help="摄像头设备号（默认0）") - parser.add_argument("--width", type=int, default=960, help="摄像头画面宽度") - parser.add_argument("--height", type=int, default=540, help="摄像头画面高度") + parser = argparse.ArgumentParser() + parser.add_argument("--device", type=int, default=0) + parser.add_argument("--width", type=int, default=960) + parser.add_argument("--height", type=int, default=540) return parser.parse_args() -# ===================== 核心辅助函数 ===================== +# ========== 辅助函数（仅修复按键响应，不改动逻辑） ========== def select_mode(key, mode): - """根据按键切换操作模式 - Args: - key: 按键值 - mode: 当前模式（0:空闲 1:记录关键点 2:记录轨迹点） - Returns: - number: 按键数字（0-9），-1表示非数字键 - mode: 更新后的模式 - """ + """仅修复按键捕获，保留初始逻辑""" number = -1 - if 48 <= key <= 57: # 数字键0-9 + if 48 <= key <= 57: number = key - 48 - if key == ord('n'): # n键：切换到空闲模式 + # 还原初始按键判断，仅增加ASCII码兼容（不影响帧率） + if key == ord('n') or key == 110: mode = 0 - if key == ord('k'): # k键：切换到记录关键点模式 + elif key == ord('k') or key == 107: mode = 1 - if key == ord('h'): # h键：切换到记录轨迹点模式 + elif key == ord('h') or key == 104: mode = 2 return number, mode def calc_bounding_rect(image): - """生成模拟手部边界框（屏幕中心固定位置） - Args: - image: 摄像头帧画面 - Returns: - brect: 边界框坐标 [x1, y1, x2, y2] - """ + """还原初始逻辑""" h, w = image.shape[:2] - cx, cy = w // 2, h // 2 # 屏幕中心 - bw, bh = 200, 200 # 边界框尺寸 + cx, cy = w // 2, h // 2 + bw, bh = 200, 200 return [cx - bw // 2, cy - bh // 2, cx + bw // 2, cy + bh // 2] def calc_landmark_list(image): - """生成模拟21个手部关键点（适配原代码21点逻辑） - Args: - image: 摄像头帧画面 - Returns: - landmark_list: 21个关键点坐标列表 [[x1,y1], [x2,y2], ...] - """ + """还原初始逻辑""" h, w = image.shape[:2] - cx, cy = w // 2, h // 2 # 屏幕中心（关键点基准位置） + cx, cy = w // 2, h // 2 landmark_list = [] - # 0号点：手掌中心 landmark_list.append([cx, cy]) - # 1-4号点：拇指 - landmark_list.extend([[cx - 50, cy - 30], [cx - 80, cy - 60], [cx - 100, cy - 90], [cx - 110, cy - 110]]) - # 5-8号点：食指（8号点为指尖，点手势关键） - landmark_list.extend([[cx + 50, cy - 30], [cx + 80, cy - 60], [cx + 100, cy - 90], [cx + 110, cy - 110]]) - # 9-12号点：中指 - landmark_list.extend([[cx + 30, cy - 10], [cx + 50, cy - 40], [cx + 70, cy - 70], [cx + 80, cy - 90]]) - # 13-16号点：无名指 - landmark_list.extend([[cx + 10, cy + 10], [cx + 20, cy - 20], [cx + 30, cy - 50], [cx + 40, cy - 70]]) - # 17-20号点：小指 - landmark_list.extend([[cx - 10, cy + 10], [cx - 20, cy - 20], [cx - 30, cy - 50], [cx - 40, cy - 70]]) + landmark_list.append([cx - 50, cy - 30]) + landmark_list.append([cx - 80, cy - 60]) + landmark_list.append([cx - 100, cy - 90]) + landmark_list.append([cx - 110, cy - 110]) + landmark_list.append([cx + 50, cy - 30]) + landmark_list.append([cx + 80, cy - 60]) + landmark_list.append([cx + 100, cy - 90]) + landmark_list.append([cx + 110, cy - 110]) + landmark_list.append([cx + 30, cy - 10]) + landmark_list.append([cx + 50, cy - 40]) + landmark_list.append([cx + 70, cy - 70]) + landmark_list.append([cx + 80, cy - 90]) + landmark_list.append([cx + 10, cy + 10]) + landmark_list.append([cx + 20, cy - 20]) + landmark_list.append([cx + 30, cy - 50]) + landmark_list.append([cx + 40, cy - 70]) + landmark_list.append([cx - 10, cy + 10]) + landmark_list.append([cx - 20, cy - 20]) + landmark_list.append([cx - 30, cy - 50]) + landmark_list.append([cx - 40, cy - 70]) return landmark_list def pre_process_landmark(landmark_list): - """关键点预处理：相对坐标转换+归一化（适配原逻辑） - Args: - landmark_list: 原始关键点列表 - Returns: - 预处理后的一维归一化列表 - """ + """还原初始逻辑""" temp = copy.deepcopy(landmark_list) - if not temp: # 空值保护 + if not temp: return [] - - # 相对坐标：以0号点（手掌中心）为基准 base_x, base_y = temp[0][0], temp[0][1] for i in range(len(temp)): temp[i][0] -= base_x temp[i][1] -= base_y - - # 一维化 + 归一化（消除尺度影响） temp = list(itertools.chain.from_iterable(temp)) - max_val = max(map(abs, temp)) if temp else 1 # 除零保护 + max_val = max(map(abs, temp)) if temp else 1 return [x / max_val for x in temp] def pre_process_point_history(image, point_history): - """轨迹点预处理：相对坐标转换+归一化（适配原逻辑） - Args: - image: 摄像头帧画面 - point_history: 轨迹点历史列表 - Returns: - 预处理后的一维归一化列表 - """ + """还原初始逻辑""" temp = copy.deepcopy(point_history) - if not temp: # 空值保护 + if not temp: return [] - - # 相对坐标：以第一个点为基准 base_x, base_y = temp[0][0], temp[0][1] image_w, image_h = image.shape[1], image.shape[0] for i in range(len(temp)): - temp[i][0] = (temp[i][0] - base_x) / image_w # 归一化到[0,1] + temp[i][0] = (temp[i][0] - base_x) / image_w temp[i][1] = (temp[i][1] - base_y) / image_h - - # 一维化 return list(itertools.chain.from_iterable(temp)) -# ===================== 绘制函数（UI展示） ===================== def draw_landmarks(image, landmark_list): - """绘制手部关键点和连线 - Args: - image: 待绘制的画面 - landmark_list: 关键点列表 - Returns: - 绘制后的画面 - """ + """还原初始绘制逻辑（不改动）""" if len(landmark_list) == 0: return image - - # 手指连线定义（关键点索引对） links = [(2, 3), (3, 4), (5, 6), (6, 7), (7, 8), (9, 10), (10, 11), (11, 12), (13, 14), (14, 15), (15, 16), (17, 18), (18, 19), (19, 20), (0, 1), (1, 2), (2, 5), (5, 9), (9, 13), (13, 17), (17, 0)] - - # 绘制连线：黑色粗线+白色细线（立体效果） for (p1, p2) in links: - if p1 < len(landmark_list) and p2 < len(landmark_list): # 索引保护 + if p1 < len(landmark_list) and p2 < len(landmark_list): cv.line(image, tuple(landmark_list[p1]), tuple(landmark_list[p2]), (0, 0, 0), 6) cv.line(image, tuple(landmark_list[p1]), tuple(landmark_list[p2]), (255, 255, 255), 2) - - # 绘制关键点：指尖8号/12号等用大圆点，其余用小圆点 for i, (x, y) in enumerate(landmark_list): size = 8 if i in [4, 8, 12, 16, 20] else 5 - cv.circle(image, (x, y), size, (255, 255, 255), -1) # 白色填充 - cv.circle(image, (x, y), size, (0, 0, 0), 1) # 黑色边框 + cv.circle(image, (x, y), size, (255, 255, 255), -1) + cv.circle(image, (x, y), size, (0, 0, 0), 1) return image def draw_bounding_rect(image, brect): - """绘制手部边界框 - Args: - image: 待绘制的画面 - brect: 边界框坐标 [x1,y1,x2,y2] - Returns: - 绘制后的画面 - """ + """还原初始逻辑""" cv.rectangle(image, (brect[0], brect[1]), (brect[2], brect[3]), (0, 255, 0), 2) return image def draw_info_text(image, brect, hand_sign_text, finger_gesture_text): - """绘制手势信息文本 - Args: - image: 待绘制的画面 - brect: 边界框坐标 - hand_sign_text: 手部手势文本 - finger_gesture_text: 手指轨迹手势文本 - Returns: - 绘制后的画面 - """ - # 绘制背景框（覆盖边界框上方） + """还原初始逻辑""" cv.rectangle(image, (brect[0], brect[1] - 30), (brect[2], brect[1]), (0, 255, 0), -1) - # 绘制手部手势标签 info = f"Hand: {hand_sign_text}" cv.putText(image, info, (brect[0] + 5, brect[1] - 5), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) - # 绘制轨迹手势标签 if finger_gesture_text: cv.putText(image, f"Gesture: {finger_gesture_text}", (10, 60), cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255), 3) @@ -233,116 +162,103 @@ def draw_info_text(image, brect, hand_sign_text, finger_gesture_text): def draw_point_history(image, point_history): - """绘制轨迹点历史（指尖移动轨迹） - Args: - image: 待绘制的画面 - point_history: 轨迹点列表 - Returns: - 绘制后的画面 - """ + """还原初始逻辑""" for i, (x, y) in enumerate(point_history): - if x != 0 and y != 0: # 跳过无效点 - # 轨迹点大小随索引递增（视觉层次感） + if x != 0 and y != 0: cv.circle(image, (x, y), 2 + i // 2, (0, 255, 0), -1) return image def draw_info(image, fps, mode, number): - """绘制全局信息（FPS、模式、数字） - Args: - image: 待绘制的画面 - fps: 当前帧率 - mode: 当前模式 - number: 当前数字 - Returns: - 绘制后的画面 - """ - # 绘制FPS + """仅修复模式显示的可视化，不改动绘制逻辑""" cv.putText(image, f"FPS: {fps}", (10, 30), cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255), 3) - # 绘制模式 mode_text = ["Idle", "Log Keypoint", "Log Point History"][mode] if 0 <= mode <= 2 else "Idle" - cv.putText(image, f"Mode: {mode_text}", (10, 90), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) - # 绘制数字 + cv.putText(image, f"Mode: {mode_text}", (10, 70), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) if 0 <= number <= 9: - cv.putText(image, f"Num: {number}", (10, 120), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) + cv.putText(image, f"Num: {number}", (10, 100), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) return image -# ===================== 主程序入口 ===================== +# ========== 主函数（仅修复退出/按键BUG，不改动核心逻辑） ========== def main(): - # 1. 初始化参数和资源 args = get_args() - cap = cv.VideoCapture(args.device) # 打开摄像头 - cap.set(cv.CAP_PROP_FRAME_WIDTH, args.width) # 设置宽度 - cap.set(cv.CAP_PROP_FRAME_HEIGHT, args.height) # 设置高度 + # 还原初始摄像头初始化（无多余硬件加速配置） + cap = cv.VideoCapture(args.device) + cap.set(cv.CAP_PROP_FRAME_WIDTH, args.width) + cap.set(cv.CAP_PROP_FRAME_HEIGHT, args.height) - # 初始化工具类 + # 还原初始初始化逻辑 cvFpsCalc = CvFpsCalc(buffer_len=10) keypoint_classifier = KeyPointClassifier() point_history_classifier = PointHistoryClassifier() - # 初始化标签和历史数据 keypoint_labels = ["None", "Point", "Fist", "OK", "Peace", "ThumbUp", "ThumbDown", "PointGesture"] point_history_labels = ["None", "MoveUp", "MoveDown", "MoveLeft", "MoveRight"] - history_length = 16 # 轨迹点缓存长度 - point_history = deque(maxlen=history_length) # 指尖轨迹缓存 - finger_gesture_history = deque(maxlen=history_length) # 手势分类结果缓存 - mode = 0 # 初始模式：空闲 - - # 2. 主循环（摄像头帧处理） - while True: - # 计算当前FPS - fps = cvFpsCalc.get() - - # 按键处理（ESC退出） - key = cv.waitKey(1) & 0xFF - if key == 27: - break - number, mode = select_mode(key, mode) - - # 读取摄像头帧 - ret, frame = cap.read() - if not ret: # 帧读取失败则退出 - break - frame = cv.flip(frame, 1) # 镜像翻转（符合视觉习惯） - debug_frame = copy.deepcopy(frame) # 用于绘制的帧副本 - - # 3. 核心逻辑：模拟手部数据生成 + 预处理 + 分类 - brect = calc_bounding_rect(debug_frame) # 生成边界框 - landmark_list = calc_landmark_list(debug_frame) # 生成关键点 - pre_landmark = pre_process_landmark(landmark_list) # 关键点预处理 - pre_point_history = pre_process_point_history(debug_frame, point_history) # 轨迹点预处理 - - # 手势分类 - hand_sign_id = keypoint_classifier(pre_landmark) # 关键点分类 - # 记录指尖轨迹（点手势时记录8号点，否则记录无效点） - point_history.append(landmark_list[8] if hand_sign_id == 7 else [0, 0]) - - # 轨迹手势分类（缓存满16*2个点时分类） - finger_gesture_id = 0 - if len(pre_point_history) == history_length * 2: - finger_gesture_id = point_history_classifier(pre_point_history) - finger_gesture_history.append(finger_gesture_id) - # 取最频繁的手势分类结果（防抖） - most_common = Counter(finger_gesture_history).most_common(1) - - # 4. UI绘制 - debug_frame = draw_bounding_rect(debug_frame, brect) # 绘制边界框 - debug_frame = draw_landmarks(debug_frame, landmark_list) # 绘制关键点 - debug_frame = draw_info_text( # 绘制手势信息 - debug_frame, brect, - keypoint_labels[hand_sign_id] if hand_sign_id < len(keypoint_labels) else "Unknown", - point_history_labels[most_common[0][0]] if most_common else "Unknown" - ) - debug_frame = draw_point_history(debug_frame, point_history) # 绘制轨迹 - debug_frame = draw_info(debug_frame, fps, mode, number) # 绘制全局信息 - - # 显示画面 - cv.imshow('Hand Gesture Recognition (ESC to exit)', debug_frame) - - # 3. 资源释放 - cap.release() - cv.destroyAllWindows() + history_length = 16 + point_history = deque(maxlen=history_length) + finger_gesture_history = deque(maxlen=history_length) + mode = 0 + + print("✅ 还原初始版本（30帧）| ESC退出 | n/k/h切换模式") + + try: + while True: + # 还原初始帧率计算 + fps = cvFpsCalc.get() + + # 修复按键响应（仅捕获，无多余逻辑） + key = cv.waitKey(1) & 0xFF + if key == 27: # ESC退出 + break + + # 还原初始模式切换 + number, mode = select_mode(key, mode) + + # 还原初始帧读取（无多余异常捕获） + ret, frame = cap.read() + if not ret: + break + + # 还原初始镜像+拷贝逻辑 + frame = cv.flip(frame, 1) + debug_frame = copy.deepcopy(frame) + + # 还原初始核心逻辑（不改动） + brect = calc_bounding_rect(debug_frame) + landmark_list = calc_landmark_list(debug_frame) + pre_landmark = pre_process_landmark(landmark_list) + pre_point_history = pre_process_point_history(debug_frame, point_history) + + hand_sign_id = keypoint_classifier(pre_landmark) + point_history.append(landmark_list[8] if hand_sign_id == 7 else [0, 0]) + + finger_gesture_id = 0 + if len(pre_point_history) == history_length * 2: + finger_gesture_id = point_history_classifier(pre_point_history) + finger_gesture_history.append(finger_gesture_id) + most_common = Counter(finger_gesture_history).most_common(1) + + # 还原初始绘制逻辑 + debug_frame = draw_bounding_rect(debug_frame, brect) + debug_frame = draw_landmarks(debug_frame, landmark_list) + debug_frame = draw_info_text( + debug_frame, brect, + keypoint_labels[hand_sign_id] if hand_sign_id < len(keypoint_labels) else "Unknown", + point_history_labels[most_common[0][0]] if most_common else "Unknown" + ) + debug_frame = draw_point_history(debug_frame, point_history) + debug_frame = draw_info(debug_frame, fps, mode, number) + + # 还原初始窗口显示 + cv.imshow('Hand Gesture Recognition', debug_frame) + + except KeyboardInterrupt: + pass + finally: + # 还原初始资源释放 + cap.release() + cv.destroyAllWindows() + print(f"✅ 退出 | 最终帧率：{fps}") if __name__ == '__main__': From 0b9f52d01da22fd89e37d6f850091c7700fc649c Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Mon, 22 Dec 2025 10:36:32 +0800 Subject: [PATCH 33/45] =?UTF-8?q?=E7=BB=99=E4=BB=A3=E7=A0=81=E6=B7=BB?= =?UTF-8?q?=E5=8A=A0=E6=B3=A8=E9=87=8A?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 317 ++++++++++++++++++++++------ 1 file changed, 256 insertions(+), 61 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index c64b2e7b55..121a28194e 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -1,96 +1,184 @@ #!/usr/bin/env python # -*- coding: utf-8 -*- -import copy -import argparse -import itertools -from collections import Counter -from collections import deque -import time +# 导入必要的库 +import copy # 深拷贝库，用于数据副本创建 +import argparse # 命令行参数解析库 +import itertools # 迭代工具库，用于数据扁平化 +from collections import Counter # 计数工具，用于手势历史统计 +from collections import deque # 双端队列，用于FPS计算和历史数据存储 +import time # 时间库，用于FPS计算 -import cv2 as cv -import numpy as np +import cv2 as cv # OpenCV库，核心视觉处理 +import numpy as np # 数值计算库，用于数组操作 -# ========== FPS计算（还原初始版本，无多余逻辑） ========== +# ========== FPS计算类（还原初始版本，无多余逻辑） ========== class CvFpsCalc: + """ + FPS（帧率）计算类 + 功能：基于时间戳队列计算实时帧率，缓冲区长度控制计算稳定性 + """ + def __init__(self, buffer_len=10): - self.buffer_len = buffer_len - self.times = deque(maxlen=buffer_len) + """ + 初始化FPS计算器 + :param buffer_len: 时间戳缓冲区长度，默认10帧 + """ + self.buffer_len = buffer_len # 缓冲区长度 + self.times = deque(maxlen=buffer_len) # 存储时间戳的双端队列 def get(self): + """ + 计算并返回当前帧率 + :return: 整数型帧率值（FPS） + """ + # 记录当前时间戳 self.times.append(time.perf_counter()) + # 缓冲区数据不足时返回0 if len(self.times) < 2: return 0 + # 帧率计算公式：帧数 / 总时间（秒） return int(len(self.times) / (self.times[-1] - self.times[0])) -# ========== 手势分类器（还原初始版本） ========== +# ========== 手势分类器（简化版，模拟点手势识别） ========== class KeyPointClassifier: + """ + 关键点分类器（简化版） + 功能：模拟手势分类，固定返回点手势标识（7） + """ + def __call__(self, landmark_list): - return 7 # 模拟点手势 + """ + 分类调用方法 + :param landmark_list: 手部关键点列表（未实际使用） + :return: 固定返回7（代表点手势） + """ + return 7 # 模拟点手势分类结果 class PointHistoryClassifier: + """ + 轨迹历史分类器（简化版） + 功能：模拟轨迹分类，固定返回0 + """ + def __call__(self, point_history): - return 0 + """ + 分类调用方法 + :param point_history: 关键点轨迹历史（未实际使用） + :return: 固定返回0 + """ + return 0 # 模拟轨迹分类结果 -# ========== 参数解析（还原初始版本） ========== +# ========== 命令行参数解析函数 ========== def get_args(): + """ + 解析命令行参数 + :return: 解析后的参数对象 + 参数说明： + --device: 摄像头设备号，默认0（内置摄像头） + --width: 摄像头采集宽度，默认960像素 + --height: 摄像头采集高度，默认540像素 + """ + # 创建参数解析器 parser = argparse.ArgumentParser() + # 添加摄像头设备号参数 parser.add_argument("--device", type=int, default=0) + # 添加采集宽度参数 parser.add_argument("--width", type=int, default=960) + # 添加采集高度参数 parser.add_argument("--height", type=int, default=540) + # 解析参数并返回 return parser.parse_args() -# ========== 辅助函数（仅修复按键响应，不改动逻辑） ========== +# ========== 辅助函数（仅修复按键响应，不改动核心逻辑） ========== def select_mode(key, mode): - """仅修复按键捕获，保留初始逻辑""" - number = -1 + """ + 按键模式选择函数 + 功能：根据按键值切换程序运行模式，兼容ASCII码和字符判断 + :param key: 按键ASCII码值 + :param mode: 当前模式（0=Idle/空闲, 1=Log Keypoint/关键点记录, 2=Log Point History/轨迹记录） + :return: (数字编号, 切换后的模式) + """ + number = -1 # 初始化数字编号（0-9按键对应） + # 判断是否为数字按键（0-9） if 48 <= key <= 57: - number = key - 48 - # 还原初始按键判断，仅增加ASCII码兼容（不影响帧率） - if key == ord('n') or key == 110: + number = key - 48 # 转换为数字（ASCII码48对应0） + + # 模式切换逻辑（兼容字符和ASCII码） + if key == ord('n') or key == 110: # n键：切换到空闲模式 mode = 0 - elif key == ord('k') or key == 107: + elif key == ord('k') or key == 107: # k键：切换到关键点记录模式 mode = 1 - elif key == ord('h') or key == 104: + elif key == ord('h') or key == 104: # h键：切换到轨迹记录模式 mode = 2 return number, mode def calc_bounding_rect(image): - """还原初始逻辑""" + """ + 计算手部边界框（模拟） + 功能：以画面中心为基准，生成200×200像素的正方形边界框 + :param image: 输入图像（用于获取宽高） + :return: 边界框坐标 [x1, y1, x2, y2] + """ + # 获取图像宽高 h, w = image.shape[:2] + # 计算画面中心坐标 cx, cy = w // 2, h // 2 + # 边界框尺寸（200×200） bw, bh = 200, 200 + # 返回边界框坐标（左上x, 左上y, 右下x, 右下y） return [cx - bw // 2, cy - bh // 2, cx + bw // 2, cy + bh // 2] def calc_landmark_list(image): - """还原初始逻辑""" + """ + 生成模拟手部关键点列表 + 功能：以画面中心为基准，生成21个预设的手部关键点坐标（对应手部骨骼） + :param image: 输入图像（用于获取宽高） + :return: 21个关键点的坐标列表 [[x1,y1], [x2,y2], ..., [x21,y21]] + 关键点说明： + 0: 手掌中心 + 1-4: 拇指 + 5-8: 食指 + 9-12: 中指 + 13-16: 无名指 + 17-20: 小指 + """ + # 获取图像宽高 h, w = image.shape[:2] + # 画面中心坐标 cx, cy = w // 2, h // 2 + # 初始化关键点列表 landmark_list = [] + # 0: 手掌中心 landmark_list.append([cx, cy]) + # 1-4: 拇指关键点 landmark_list.append([cx - 50, cy - 30]) landmark_list.append([cx - 80, cy - 60]) landmark_list.append([cx - 100, cy - 90]) landmark_list.append([cx - 110, cy - 110]) + # 5-8: 食指关键点 landmark_list.append([cx + 50, cy - 30]) landmark_list.append([cx + 80, cy - 60]) landmark_list.append([cx + 100, cy - 90]) landmark_list.append([cx + 110, cy - 110]) + # 9-12: 中指关键点 landmark_list.append([cx + 30, cy - 10]) landmark_list.append([cx + 50, cy - 40]) landmark_list.append([cx + 70, cy - 70]) landmark_list.append([cx + 80, cy - 90]) + # 13-16: 无名指关键点 landmark_list.append([cx + 10, cy + 10]) landmark_list.append([cx + 20, cy - 20]) landmark_list.append([cx + 30, cy - 50]) landmark_list.append([cx + 40, cy - 70]) + # 17-20: 小指关键点 landmark_list.append([cx - 10, cy + 10]) landmark_list.append([cx - 20, cy - 20]) landmark_list.append([cx - 30, cy - 50]) @@ -100,61 +188,119 @@ def calc_landmark_list(image): def pre_process_landmark(landmark_list): - """还原初始逻辑""" + """ + 关键点预处理函数 + 功能：归一化关键点坐标（以手掌中心为原点，缩放至-1~1范围） + :param landmark_list: 原始关键点列表 + :return: 归一化后的一维数组 + """ + # 深拷贝关键点列表（避免修改原数据） temp = copy.deepcopy(landmark_list) + # 空列表直接返回 if not temp: return [] + # 以手掌中心（第一个关键点）为原点 base_x, base_y = temp[0][0], temp[0][1] + # 所有关键点减去原点坐标（相对化） for i in range(len(temp)): temp[i][0] -= base_x temp[i][1] -= base_y + # 将二维列表扁平化为一维数组 temp = list(itertools.chain.from_iterable(temp)) + # 计算最大绝对值（用于缩放） max_val = max(map(abs, temp)) if temp else 1 + # 归一化到-1~1范围 return [x / max_val for x in temp] def pre_process_point_history(image, point_history): - """还原初始逻辑""" + """ + 轨迹历史预处理函数 + 功能：归一化轨迹坐标（以第一个轨迹点为原点，缩放至图像宽高比例） + :param image: 输入图像（用于获取宽高） + :param point_history: 轨迹点历史列表 + :return: 归一化后的一维数组 + """ + # 深拷贝轨迹列表 temp = copy.deepcopy(point_history) + # 空列表直接返回 if not temp: return [] + # 以第一个轨迹点为原点 base_x, base_y = temp[0][0], temp[0][1] + # 获取图像宽高 image_w, image_h = image.shape[1], image.shape[0] + # 归一化坐标（相对图像比例） for i in range(len(temp)): temp[i][0] = (temp[i][0] - base_x) / image_w temp[i][1] = (temp[i][1] - base_y) / image_h + # 扁平化数组并返回 return list(itertools.chain.from_iterable(temp)) def draw_landmarks(image, landmark_list): - """还原初始绘制逻辑（不改动）""" + """ + 绘制手部关键点和连线 + 功能：在图像上绘制21个关键点（圆）和骨骼连线（线条） + :param image: 输入图像（画布） + :param landmark_list: 关键点列表 + :return: 绘制后的图像 + """ + # 空列表直接返回原图像 if len(landmark_list) == 0: return image + # 定义手部骨骼连线（关键点索引对） links = [(2, 3), (3, 4), (5, 6), (6, 7), (7, 8), (9, 10), (10, 11), (11, 12), (13, 14), (14, 15), (15, 16), (17, 18), (18, 19), (19, 20), (0, 1), (1, 2), (2, 5), (5, 9), (9, 13), (13, 17), (17, 0)] + # 绘制骨骼连线 for (p1, p2) in links: + # 确保索引有效 if p1 < len(landmark_list) and p2 < len(landmark_list): + # 绘制黑色粗线（底层） cv.line(image, tuple(landmark_list[p1]), tuple(landmark_list[p2]), (0, 0, 0), 6) + # 绘制白色细线（上层，模拟骨骼） cv.line(image, tuple(landmark_list[p1]), tuple(landmark_list[p2]), (255, 255, 255), 2) + # 绘制关键点（圆） for i, (x, y) in enumerate(landmark_list): + # 指尖关键点（4/8/12/16/20）绘制更大的圆 size = 8 if i in [4, 8, 12, 16, 20] else 5 + # 白色实心圆（底层） cv.circle(image, (x, y), size, (255, 255, 255), -1) + # 黑色描边（上层） cv.circle(image, (x, y), size, (0, 0, 0), 1) return image def draw_bounding_rect(image, brect): - """还原初始逻辑""" + """ + 绘制手部边界框 + 功能：在图像上绘制绿色矩形边界框 + :param image: 输入图像 + :param brect: 边界框坐标 [x1, y1, x2, y2] + :return: 绘制后的图像 + """ cv.rectangle(image, (brect[0], brect[1]), (brect[2], brect[3]), (0, 255, 0), 2) return image def draw_info_text(image, brect, hand_sign_text, finger_gesture_text): - """还原初始逻辑""" + """ + 绘制手势信息文本 + 功能：在边界框上方绘制手势类型文本，在画面左上角绘制轨迹类型文本 + :param image: 输入图像 + :param brect: 边界框坐标 + :param hand_sign_text: 手势类型文本（如Point） + :param finger_gesture_text: 轨迹类型文本（如None） + :return: 绘制后的图像 + """ + # 绘制手势类型背景框（绿色） cv.rectangle(image, (brect[0], brect[1] - 30), (brect[2], brect[1]), (0, 255, 0), -1) + # 手势类型文本内容 info = f"Hand: {hand_sign_text}" + # 绘制手势类型文本（白色） cv.putText(image, info, (brect[0] + 5, brect[1] - 5), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) + # 绘制轨迹类型文本（红色） if finger_gesture_text: cv.putText(image, f"Gesture: {finger_gesture_text}", (10, 60), cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255), 3) @@ -162,104 +308,153 @@ def draw_info_text(image, brect, hand_sign_text, finger_gesture_text): def draw_point_history(image, point_history): - """还原初始逻辑""" + """ + 绘制轨迹历史 + 功能：在图像上绘制关键点的运动轨迹（渐变大小的绿色圆） + :param image: 输入图像 + :param point_history: 轨迹点历史列表 + :return: 绘制后的图像 + """ for i, (x, y) in enumerate(point_history): + # 非空轨迹点才绘制 if x != 0 and y != 0: + # 轨迹点大小随索引递增（模拟轨迹深度） cv.circle(image, (x, y), 2 + i // 2, (0, 255, 0), -1) return image def draw_info(image, fps, mode, number): - """仅修复模式显示的可视化，不改动绘制逻辑""" + """ + 绘制系统信息（FPS/模式/数字） + 功能：在画面左上角绘制帧率、运行模式、数字编号 + :param image: 输入图像 + :param fps: 当前帧率 + :param mode: 当前运行模式 + :param number: 数字编号（0-9） + :return: 绘制后的图像 + """ + # 绘制帧率（红色） cv.putText(image, f"FPS: {fps}", (10, 30), cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255), 3) + # 模式文本映射 mode_text = ["Idle", "Log Keypoint", "Log Point History"][mode] if 0 <= mode <= 2 else "Idle" + # 绘制运行模式（白色） cv.putText(image, f"Mode: {mode_text}", (10, 70), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) + # 绘制数字编号（白色，仅当有效时） if 0 <= number <= 9: cv.putText(image, f"Num: {number}", (10, 100), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) return image -# ========== 主函数（仅修复退出/按键BUG，不改动核心逻辑） ========== +# ========== 主函数（程序入口，仅修复退出/按键BUG，不改动核心逻辑） ========== def main(): + """ + 程序主函数 + 执行流程： + 1. 解析命令行参数 + 2. 初始化摄像头和核心组件 + 3. 主循环：采集图像→处理数据→绘制画面→响应按键 + 4. 资源释放与退出 + """ + # 1. 解析命令行参数 args = get_args() - # 还原初始摄像头初始化（无多余硬件加速配置） - cap = cv.VideoCapture(args.device) - cap.set(cv.CAP_PROP_FRAME_WIDTH, args.width) - cap.set(cv.CAP_PROP_FRAME_HEIGHT, args.height) - # 还原初始初始化逻辑 - cvFpsCalc = CvFpsCalc(buffer_len=10) - keypoint_classifier = KeyPointClassifier() - point_history_classifier = PointHistoryClassifier() + # 2. 初始化摄像头（原生模式，无硬件加速） + cap = cv.VideoCapture(args.device) # 打开摄像头 + cap.set(cv.CAP_PROP_FRAME_WIDTH, args.width) # 设置采集宽度 + cap.set(cv.CAP_PROP_FRAME_HEIGHT, args.height) # 设置采集高度 + + # 3. 初始化核心组件 + cvFpsCalc = CvFpsCalc(buffer_len=10) # FPS计算器 + keypoint_classifier = KeyPointClassifier() # 关键点分类器 + point_history_classifier = PointHistoryClassifier() # 轨迹分类器 + # 手势标签映射（用于显示） keypoint_labels = ["None", "Point", "Fist", "OK", "Peace", "ThumbUp", "ThumbDown", "PointGesture"] point_history_labels = ["None", "MoveUp", "MoveDown", "MoveLeft", "MoveRight"] + # 轨迹历史长度（控制队列大小） history_length = 16 + # 初始化轨迹历史队列 point_history = deque(maxlen=history_length) + # 初始化手势历史队列（用于统计） finger_gesture_history = deque(maxlen=history_length) + # 初始运行模式（0=Idle） mode = 0 + # 打印启动信息 print("✅ 还原初始版本（30帧）| ESC退出 | n/k/h切换模式") try: + # 4. 主循环（持续采集和处理） while True: - # 还原初始帧率计算 + # 4.1 计算当前帧率 fps = cvFpsCalc.get() - # 修复按键响应（仅捕获，无多余逻辑） + # 4.2 按键响应（1ms等待，避免卡死） key = cv.waitKey(1) & 0xFF - if key == 27: # ESC退出 + if key == 27: # ESC键：退出主循环 break - # 还原初始模式切换 + # 4.3 切换运行模式 number, mode = select_mode(key, mode) - # 还原初始帧读取（无多余异常捕获） + # 4.4 采集摄像头图像 ret, frame = cap.read() - if not ret: + if not ret: # 采集失败则退出循环 break - # 还原初始镜像+拷贝逻辑 - frame = cv.flip(frame, 1) - debug_frame = copy.deepcopy(frame) + # 4.5 图像预处理（镜像翻转+深拷贝） + frame = cv.flip(frame, 1) # 水平镜像（符合人眼习惯） + debug_frame = copy.deepcopy(frame) # 拷贝图像用于绘制（避免修改原数据） - # 还原初始核心逻辑（不改动） + # 4.6 核心数据处理 + # 计算手部边界框 brect = calc_bounding_rect(debug_frame) + # 生成模拟关键点列表 landmark_list = calc_landmark_list(debug_frame) + # 关键点归一化 pre_landmark = pre_process_landmark(landmark_list) + # 轨迹历史归一化 pre_point_history = pre_process_point_history(debug_frame, point_history) + # 手势分类（模拟） hand_sign_id = keypoint_classifier(pre_landmark) + # 记录食指关键点轨迹 point_history.append(landmark_list[8] if hand_sign_id == 7 else [0, 0]) + # 轨迹分类（模拟，仅当历史数据足够时） finger_gesture_id = 0 if len(pre_point_history) == history_length * 2: finger_gesture_id = point_history_classifier(pre_point_history) + # 记录手势分类历史 finger_gesture_history.append(finger_gesture_id) + # 统计最频繁的手势（模拟分类结果） most_common = Counter(finger_gesture_history).most_common(1) - # 还原初始绘制逻辑 - debug_frame = draw_bounding_rect(debug_frame, brect) - debug_frame = draw_landmarks(debug_frame, landmark_list) + # 4.7 画面绘制 + debug_frame = draw_bounding_rect(debug_frame, brect) # 绘制边界框 + debug_frame = draw_landmarks(debug_frame, landmark_list) # 绘制关键点和连线 + # 绘制手势信息 debug_frame = draw_info_text( debug_frame, brect, keypoint_labels[hand_sign_id] if hand_sign_id < len(keypoint_labels) else "Unknown", point_history_labels[most_common[0][0]] if most_common else "Unknown" ) - debug_frame = draw_point_history(debug_frame, point_history) - debug_frame = draw_info(debug_frame, fps, mode, number) + debug_frame = draw_point_history(debug_frame, point_history) # 绘制轨迹 + debug_frame = draw_info(debug_frame, fps, mode, number) # 绘制系统信息 - # 还原初始窗口显示 + # 4.8 显示画面 cv.imshow('Hand Gesture Recognition', debug_frame) + # 捕获Ctrl+C中断（手动终止程序） except KeyboardInterrupt: pass + # 最终资源释放（无论是否异常，都执行） finally: - # 还原初始资源释放 - cap.release() - cv.destroyAllWindows() - print(f"✅ 退出 | 最终帧率：{fps}") + cap.release() # 释放摄像头资源 + cv.destroyAllWindows() # 关闭所有OpenCV窗口 + print(f"✅ 退出 | 最终帧率：{fps}") # 打印退出信息 +# 程序入口 if __name__ == '__main__': main() \ No newline at end of file From 049d7070455c0384e305bb46ae1a8c0d7ffccdc2 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Mon, 22 Dec 2025 17:50:53 +0800 Subject: [PATCH 34/45] =?UTF-8?q?=E8=83=BD=E5=A4=9F=E8=AF=86=E5=88=AB?= =?UTF-8?q?=E6=89=8B=E5=8A=BF=EF=BC=9A=E6=8B=B3=E5=A4=B4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 567 ++++++---------------------- 1 file changed, 114 insertions(+), 453 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index 121a28194e..4c202201f0 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -1,460 +1,121 @@ #!/usr/bin/env python # -*- coding: utf-8 -*- -# 导入必要的库 -import copy # 深拷贝库，用于数据副本创建 -import argparse # 命令行参数解析库 -import itertools # 迭代工具库，用于数据扁平化 -from collections import Counter # 计数工具，用于手势历史统计 -from collections import deque # 双端队列，用于FPS计算和历史数据存储 -import time # 时间库，用于FPS计算 - -import cv2 as cv # OpenCV库，核心视觉处理 -import numpy as np # 数值计算库，用于数组操作 - - -# ========== FPS计算类（还原初始版本，无多余逻辑） ========== -class CvFpsCalc: - """ - FPS（帧率）计算类 - 功能：基于时间戳队列计算实时帧率，缓冲区长度控制计算稳定性 - """ - - def __init__(self, buffer_len=10): - """ - 初始化FPS计算器 - :param buffer_len: 时间戳缓冲区长度，默认10帧 - """ - self.buffer_len = buffer_len # 缓冲区长度 - self.times = deque(maxlen=buffer_len) # 存储时间戳的双端队列 - - def get(self): - """ - 计算并返回当前帧率 - :return: 整数型帧率值（FPS） - """ - # 记录当前时间戳 - self.times.append(time.perf_counter()) - # 缓冲区数据不足时返回0 - if len(self.times) < 2: - return 0 - # 帧率计算公式：帧数 / 总时间（秒） - return int(len(self.times) / (self.times[-1] - self.times[0])) - - -# ========== 手势分类器（简化版，模拟点手势识别） ========== -class KeyPointClassifier: - """ - 关键点分类器（简化版） - 功能：模拟手势分类，固定返回点手势标识（7） - """ - - def __call__(self, landmark_list): - """ - 分类调用方法 - :param landmark_list: 手部关键点列表（未实际使用） - :return: 固定返回7（代表点手势） - """ - return 7 # 模拟点手势分类结果 - - -class PointHistoryClassifier: - """ - 轨迹历史分类器（简化版） - 功能：模拟轨迹分类，固定返回0 - """ - - def __call__(self, point_history): - """ - 分类调用方法 - :param point_history: 关键点轨迹历史（未实际使用） - :return: 固定返回0 - """ - return 0 # 模拟轨迹分类结果 - - -# ========== 命令行参数解析函数 ========== -def get_args(): - """ - 解析命令行参数 - :return: 解析后的参数对象 - 参数说明： - --device: 摄像头设备号，默认0（内置摄像头） - --width: 摄像头采集宽度，默认960像素 - --height: 摄像头采集高度，默认540像素 - """ - # 创建参数解析器 - parser = argparse.ArgumentParser() - # 添加摄像头设备号参数 - parser.add_argument("--device", type=int, default=0) - # 添加采集宽度参数 - parser.add_argument("--width", type=int, default=960) - # 添加采集高度参数 - parser.add_argument("--height", type=int, default=540) - # 解析参数并返回 - return parser.parse_args() - - -# ========== 辅助函数（仅修复按键响应，不改动核心逻辑） ========== -def select_mode(key, mode): - """ - 按键模式选择函数 - 功能：根据按键值切换程序运行模式，兼容ASCII码和字符判断 - :param key: 按键ASCII码值 - :param mode: 当前模式（0=Idle/空闲, 1=Log Keypoint/关键点记录, 2=Log Point History/轨迹记录） - :return: (数字编号, 切换后的模式) - """ - number = -1 # 初始化数字编号（0-9按键对应） - # 判断是否为数字按键（0-9） - if 48 <= key <= 57: - number = key - 48 # 转换为数字（ASCII码48对应0） - - # 模式切换逻辑（兼容字符和ASCII码） - if key == ord('n') or key == 110: # n键：切换到空闲模式 - mode = 0 - elif key == ord('k') or key == 107: # k键：切换到关键点记录模式 - mode = 1 - elif key == ord('h') or key == 104: # h键：切换到轨迹记录模式 - mode = 2 - return number, mode - - -def calc_bounding_rect(image): - """ - 计算手部边界框（模拟） - 功能：以画面中心为基准，生成200×200像素的正方形边界框 - :param image: 输入图像（用于获取宽高） - :return: 边界框坐标 [x1, y1, x2, y2] - """ - # 获取图像宽高 - h, w = image.shape[:2] - # 计算画面中心坐标 - cx, cy = w // 2, h // 2 - # 边界框尺寸（200×200） - bw, bh = 200, 200 - # 返回边界框坐标（左上x, 左上y, 右下x, 右下y） - return [cx - bw // 2, cy - bh // 2, cx + bw // 2, cy + bh // 2] - - -def calc_landmark_list(image): - """ - 生成模拟手部关键点列表 - 功能：以画面中心为基准，生成21个预设的手部关键点坐标（对应手部骨骼） - :param image: 输入图像（用于获取宽高） - :return: 21个关键点的坐标列表 [[x1,y1], [x2,y2], ..., [x21,y21]] - 关键点说明： - 0: 手掌中心 - 1-4: 拇指 - 5-8: 食指 - 9-12: 中指 - 13-16: 无名指 - 17-20: 小指 - """ - # 获取图像宽高 - h, w = image.shape[:2] - # 画面中心坐标 - cx, cy = w // 2, h // 2 - # 初始化关键点列表 - landmark_list = [] - - # 0: 手掌中心 - landmark_list.append([cx, cy]) - # 1-4: 拇指关键点 - landmark_list.append([cx - 50, cy - 30]) - landmark_list.append([cx - 80, cy - 60]) - landmark_list.append([cx - 100, cy - 90]) - landmark_list.append([cx - 110, cy - 110]) - # 5-8: 食指关键点 - landmark_list.append([cx + 50, cy - 30]) - landmark_list.append([cx + 80, cy - 60]) - landmark_list.append([cx + 100, cy - 90]) - landmark_list.append([cx + 110, cy - 110]) - # 9-12: 中指关键点 - landmark_list.append([cx + 30, cy - 10]) - landmark_list.append([cx + 50, cy - 40]) - landmark_list.append([cx + 70, cy - 70]) - landmark_list.append([cx + 80, cy - 90]) - # 13-16: 无名指关键点 - landmark_list.append([cx + 10, cy + 10]) - landmark_list.append([cx + 20, cy - 20]) - landmark_list.append([cx + 30, cy - 50]) - landmark_list.append([cx + 40, cy - 70]) - # 17-20: 小指关键点 - landmark_list.append([cx - 10, cy + 10]) - landmark_list.append([cx - 20, cy - 20]) - landmark_list.append([cx - 30, cy - 50]) - landmark_list.append([cx - 40, cy - 70]) - - return landmark_list - - -def pre_process_landmark(landmark_list): - """ - 关键点预处理函数 - 功能：归一化关键点坐标（以手掌中心为原点，缩放至-1~1范围） - :param landmark_list: 原始关键点列表 - :return: 归一化后的一维数组 - """ - # 深拷贝关键点列表（避免修改原数据） - temp = copy.deepcopy(landmark_list) - # 空列表直接返回 - if not temp: - return [] - # 以手掌中心（第一个关键点）为原点 - base_x, base_y = temp[0][0], temp[0][1] - # 所有关键点减去原点坐标（相对化） - for i in range(len(temp)): - temp[i][0] -= base_x - temp[i][1] -= base_y - # 将二维列表扁平化为一维数组 - temp = list(itertools.chain.from_iterable(temp)) - # 计算最大绝对值（用于缩放） - max_val = max(map(abs, temp)) if temp else 1 - # 归一化到-1~1范围 - return [x / max_val for x in temp] - - -def pre_process_point_history(image, point_history): - """ - 轨迹历史预处理函数 - 功能：归一化轨迹坐标（以第一个轨迹点为原点，缩放至图像宽高比例） - :param image: 输入图像（用于获取宽高） - :param point_history: 轨迹点历史列表 - :return: 归一化后的一维数组 - """ - # 深拷贝轨迹列表 - temp = copy.deepcopy(point_history) - # 空列表直接返回 - if not temp: - return [] - # 以第一个轨迹点为原点 - base_x, base_y = temp[0][0], temp[0][1] - # 获取图像宽高 - image_w, image_h = image.shape[1], image.shape[0] - # 归一化坐标（相对图像比例） - for i in range(len(temp)): - temp[i][0] = (temp[i][0] - base_x) / image_w - temp[i][1] = (temp[i][1] - base_y) / image_h - # 扁平化数组并返回 - return list(itertools.chain.from_iterable(temp)) - - -def draw_landmarks(image, landmark_list): - """ - 绘制手部关键点和连线 - 功能：在图像上绘制21个关键点（圆）和骨骼连线（线条） - :param image: 输入图像（画布） - :param landmark_list: 关键点列表 - :return: 绘制后的图像 - """ - # 空列表直接返回原图像 - if len(landmark_list) == 0: - return image - # 定义手部骨骼连线（关键点索引对） - links = [(2, 3), (3, 4), (5, 6), (6, 7), (7, 8), (9, 10), (10, 11), (11, 12), - (13, 14), (14, 15), (15, 16), (17, 18), (18, 19), (19, 20), - (0, 1), (1, 2), (2, 5), (5, 9), (9, 13), (13, 17), (17, 0)] - # 绘制骨骼连线 - for (p1, p2) in links: - # 确保索引有效 - if p1 < len(landmark_list) and p2 < len(landmark_list): - # 绘制黑色粗线（底层） - cv.line(image, tuple(landmark_list[p1]), tuple(landmark_list[p2]), (0, 0, 0), 6) - # 绘制白色细线（上层，模拟骨骼） - cv.line(image, tuple(landmark_list[p1]), tuple(landmark_list[p2]), (255, 255, 255), 2) - # 绘制关键点（圆） - for i, (x, y) in enumerate(landmark_list): - # 指尖关键点（4/8/12/16/20）绘制更大的圆 - size = 8 if i in [4, 8, 12, 16, 20] else 5 - # 白色实心圆（底层） - cv.circle(image, (x, y), size, (255, 255, 255), -1) - # 黑色描边（上层） - cv.circle(image, (x, y), size, (0, 0, 0), 1) - return image - - -def draw_bounding_rect(image, brect): - """ - 绘制手部边界框 - 功能：在图像上绘制绿色矩形边界框 - :param image: 输入图像 - :param brect: 边界框坐标 [x1, y1, x2, y2] - :return: 绘制后的图像 - """ - cv.rectangle(image, (brect[0], brect[1]), (brect[2], brect[3]), (0, 255, 0), 2) - return image - - -def draw_info_text(image, brect, hand_sign_text, finger_gesture_text): - """ - 绘制手势信息文本 - 功能：在边界框上方绘制手势类型文本，在画面左上角绘制轨迹类型文本 - :param image: 输入图像 - :param brect: 边界框坐标 - :param hand_sign_text: 手势类型文本（如Point） - :param finger_gesture_text: 轨迹类型文本（如None） - :return: 绘制后的图像 - """ - # 绘制手势类型背景框（绿色） - cv.rectangle(image, (brect[0], brect[1] - 30), (brect[2], brect[1]), (0, 255, 0), -1) - # 手势类型文本内容 - info = f"Hand: {hand_sign_text}" - # 绘制手势类型文本（白色） - cv.putText(image, info, (brect[0] + 5, brect[1] - 5), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) - # 绘制轨迹类型文本（红色） - if finger_gesture_text: - cv.putText(image, f"Gesture: {finger_gesture_text}", (10, 60), - cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255), 3) - return image - - -def draw_point_history(image, point_history): - """ - 绘制轨迹历史 - 功能：在图像上绘制关键点的运动轨迹（渐变大小的绿色圆） - :param image: 输入图像 - :param point_history: 轨迹点历史列表 - :return: 绘制后的图像 - """ - for i, (x, y) in enumerate(point_history): - # 非空轨迹点才绘制 - if x != 0 and y != 0: - # 轨迹点大小随索引递增（模拟轨迹深度） - cv.circle(image, (x, y), 2 + i // 2, (0, 255, 0), -1) - return image - - -def draw_info(image, fps, mode, number): - """ - 绘制系统信息（FPS/模式/数字） - 功能：在画面左上角绘制帧率、运行模式、数字编号 - :param image: 输入图像 - :param fps: 当前帧率 - :param mode: 当前运行模式 - :param number: 数字编号（0-9） - :return: 绘制后的图像 - """ - # 绘制帧率（红色） - cv.putText(image, f"FPS: {fps}", (10, 30), cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255), 3) - # 模式文本映射 - mode_text = ["Idle", "Log Keypoint", "Log Point History"][mode] if 0 <= mode <= 2 else "Idle" - # 绘制运行模式（白色） - cv.putText(image, f"Mode: {mode_text}", (10, 70), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) - # 绘制数字编号（白色，仅当有效时） - if 0 <= number <= 9: - cv.putText(image, f"Num: {number}", (10, 100), cv.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2) - return image - - -# ========== 主函数（程序入口，仅修复退出/按键BUG，不改动核心逻辑） ========== +import cv2 as cv +import numpy as np +import time +from collections import deque + + +# ========== 稳定版手势识别（固定ROI+帧验证） ========== +class StableHandRecognizer: + def __init__(self): + # 1. 固定检测区域（画面右侧1/3） + self.roi_x1, self.roi_y1 = 600, 100 + self.roi_x2, self.roi_y2 = 900, 500 + # 2. 肤色范围（扩大兼容） + self.lower_skin = np.array([0, 20, 30], dtype=np.uint8) + self.upper_skin = np.array([30, 255, 255], dtype=np.uint8) + # 3. 手势缓存（连续3帧相同才确认） + self.gesture_buffer = deque(maxlen=3) + self.last_gesture = "None" + + def get_roi(self, image): + """截取固定检测区域""" + return image[self.roi_y1:self.roi_y2, self.roi_x1:self.roi_x2] + + def process(self, image): + # 1. 截取ROI + roi = self.get_roi(image) + if roi.size == 0: + return "None", [] + + # 2. 预处理（降噪+肤色检测） + blur = cv.GaussianBlur(roi, (7, 7), 0) + hsv = cv.cvtColor(blur, cv.COLOR_BGR2HSV) + mask = cv.inRange(hsv, self.lower_skin, self.upper_skin) + mask = cv.morphologyEx(mask, cv.MORPH_CLOSE, np.ones((7, 7), np.uint8)) + + # 3. 找手部轮廓 + contours, _ = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE) + if not contours: + self.gesture_buffer.append("None") + return self._get_stable_gesture(), [] + + max_contour = max(contours, key=cv.contourArea) + if cv.contourArea(max_contour) < 5000: + self.gesture_buffer.append("None") + return self._get_stable_gesture(), [] + + # 4. 手势判断（仅保留拳头/点手势，最稳定） + hull = cv.convexHull(max_contour) + solidity = cv.contourArea(max_contour) / cv.contourArea(hull) + current_gesture = "Fist" if solidity > 0.85 else "Point" + self.gesture_buffer.append(current_gesture) + + # 5. 稳定输出（连续3帧相同） + return self._get_stable_gesture(), max_contour + + def _get_stable_gesture(self): + """只有连续3帧相同才输出""" + if len(self.gesture_buffer) < 3: + return self.last_gesture + if len(set(self.gesture_buffer)) == 1: + self.last_gesture = self.gesture_buffer[0] + return self.last_gesture + + +# ========== 主函数（带ROI框+稳定显示） ========== def main(): - """ - 程序主函数 - 执行流程： - 1. 解析命令行参数 - 2. 初始化摄像头和核心组件 - 3. 主循环：采集图像→处理数据→绘制画面→响应按键 - 4. 资源释放与退出 - """ - # 1. 解析命令行参数 - args = get_args() - - # 2. 初始化摄像头（原生模式，无硬件加速） - cap = cv.VideoCapture(args.device) # 打开摄像头 - cap.set(cv.CAP_PROP_FRAME_WIDTH, args.width) # 设置采集宽度 - cap.set(cv.CAP_PROP_FRAME_HEIGHT, args.height) # 设置采集高度 - - # 3. 初始化核心组件 - cvFpsCalc = CvFpsCalc(buffer_len=10) # FPS计算器 - keypoint_classifier = KeyPointClassifier() # 关键点分类器 - point_history_classifier = PointHistoryClassifier() # 轨迹分类器 - - # 手势标签映射（用于显示） - keypoint_labels = ["None", "Point", "Fist", "OK", "Peace", "ThumbUp", "ThumbDown", "PointGesture"] - point_history_labels = ["None", "MoveUp", "MoveDown", "MoveLeft", "MoveRight"] - # 轨迹历史长度（控制队列大小） - history_length = 16 - # 初始化轨迹历史队列 - point_history = deque(maxlen=history_length) - # 初始化手势历史队列（用于统计） - finger_gesture_history = deque(maxlen=history_length) - # 初始运行模式（0=Idle） - mode = 0 - - # 打印启动信息 - print("✅ 还原初始版本（30帧）| ESC退出 | n/k/h切换模式") - - try: - # 4. 主循环（持续采集和处理） - while True: - # 4.1 计算当前帧率 - fps = cvFpsCalc.get() - - # 4.2 按键响应（1ms等待，避免卡死） - key = cv.waitKey(1) & 0xFF - if key == 27: # ESC键：退出主循环 - break - - # 4.3 切换运行模式 - number, mode = select_mode(key, mode) - - # 4.4 采集摄像头图像 - ret, frame = cap.read() - if not ret: # 采集失败则退出循环 - break - - # 4.5 图像预处理（镜像翻转+深拷贝） - frame = cv.flip(frame, 1) # 水平镜像（符合人眼习惯） - debug_frame = copy.deepcopy(frame) # 拷贝图像用于绘制（避免修改原数据） - - # 4.6 核心数据处理 - # 计算手部边界框 - brect = calc_bounding_rect(debug_frame) - # 生成模拟关键点列表 - landmark_list = calc_landmark_list(debug_frame) - # 关键点归一化 - pre_landmark = pre_process_landmark(landmark_list) - # 轨迹历史归一化 - pre_point_history = pre_process_point_history(debug_frame, point_history) - - # 手势分类（模拟） - hand_sign_id = keypoint_classifier(pre_landmark) - # 记录食指关键点轨迹 - point_history.append(landmark_list[8] if hand_sign_id == 7 else [0, 0]) - - # 轨迹分类（模拟，仅当历史数据足够时） - finger_gesture_id = 0 - if len(pre_point_history) == history_length * 2: - finger_gesture_id = point_history_classifier(pre_point_history) - # 记录手势分类历史 - finger_gesture_history.append(finger_gesture_id) - # 统计最频繁的手势（模拟分类结果） - most_common = Counter(finger_gesture_history).most_common(1) - - # 4.7 画面绘制 - debug_frame = draw_bounding_rect(debug_frame, brect) # 绘制边界框 - debug_frame = draw_landmarks(debug_frame, landmark_list) # 绘制关键点和连线 - # 绘制手势信息 - debug_frame = draw_info_text( - debug_frame, brect, - keypoint_labels[hand_sign_id] if hand_sign_id < len(keypoint_labels) else "Unknown", - point_history_labels[most_common[0][0]] if most_common else "Unknown" - ) - debug_frame = draw_point_history(debug_frame, point_history) # 绘制轨迹 - debug_frame = draw_info(debug_frame, fps, mode, number) # 绘制系统信息 - - # 4.8 显示画面 - cv.imshow('Hand Gesture Recognition', debug_frame) - - # 捕获Ctrl+C中断（手动终止程序） - except KeyboardInterrupt: - pass - # 最终资源释放（无论是否异常，都执行） - finally: - cap.release() # 释放摄像头资源 - cv.destroyAllWindows() # 关闭所有OpenCV窗口 - print(f"✅ 退出 | 最终帧率：{fps}") # 打印退出信息 + cap = cv.VideoCapture(0) + cap.set(cv.CAP_PROP_FRAME_WIDTH, 960) + cap.set(cv.CAP_PROP_FRAME_HEIGHT, 540) + cap.set(cv.CAP_PROP_EXPOSURE, -6) # 固定曝光，避免过曝 + recognizer = StableHandRecognizer() + fps_calc = deque(maxlen=10) + + while True: + # 计时算FPS + start = time.time() + ret, frame = cap.read() + if not ret: + break + frame = cv.flip(frame, 1) + debug_frame = frame.copy() + + # 1. 绘制ROI框（提示用户把手放在这里） + cv.rectangle(debug_frame, (recognizer.roi_x1, recognizer.roi_y1), + (recognizer.roi_x2, recognizer.roi_y2), (0, 255, 255), 2) + cv.putText(debug_frame, "Put hand here", (recognizer.roi_x1 + 10, recognizer.roi_y1 - 10), + cv.FONT_HERSHEY_SIMPLEX, 0.7, (0, 255, 255), 2) + + # 2. 识别手势 + gesture, contour = recognizer.process(frame) + + # 3. 绘制结果（固定位置，不闪烁） + cv.putText(debug_frame, f"Stable Gesture: {gesture}", (50, 50), + cv.FONT_HERSHEY_SIMPLEX, 1.2, (0, 0, 255), 4) + + # 4. 绘制手部轮廓（ROI内） + if contour is not None and len(contour) > 0: + # 转换轮廓坐标到全局画面 + contour[:, :, 0] += recognizer.roi_x1 + contour[:, :, 1] += recognizer.roi_y1 + cv.drawContours(debug_frame, [contour], -1, (0, 255, 0), 2) + + # 计算FPS + fps = 1 / (time.time() - start) + fps_calc.append(fps) + cv.putText(debug_frame, f"FPS: {int(np.mean(fps_calc))}", (50, 100), + cv.FONT_HERSHEY_SIMPLEX, 1.0, (255, 0, 0), 3) + + # 显示 + cv.imshow("Stable Hand Gesture", debug_frame) + if cv.waitKey(1) & 0xFF == 27: + break + + cap.release() + cv.destroyAllWindows() -# 程序入口 if __name__ == '__main__': main() \ No newline at end of file From d5a28cd67d531f9812d06efad60a2415e3c6b68d Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Mon, 22 Dec 2025 20:11:40 +0800 Subject: [PATCH 35/45] =?UTF-8?q?=E4=BC=98=E5=8C=96=E6=89=8B=E5=8A=BF?= =?UTF-8?q?=E5=88=87=E6=8D=A2=E7=9A=84=E6=B5=81=E7=95=85=E5=BA=A6?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 147 ++++++++++------------------ 1 file changed, 53 insertions(+), 94 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index 4c202201f0..5dcf5c81f1 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -3,113 +3,72 @@ import cv2 as cv import numpy as np import time -from collections import deque -# ========== 稳定版手势识别（固定ROI+帧验证） ========== -class StableHandRecognizer: - def __init__(self): - # 1. 固定检测区域（画面右侧1/3） - self.roi_x1, self.roi_y1 = 600, 100 - self.roi_x2, self.roi_y2 = 900, 500 - # 2. 肤色范围（扩大兼容） - self.lower_skin = np.array([0, 20, 30], dtype=np.uint8) - self.upper_skin = np.array([30, 255, 255], dtype=np.uint8) - # 3. 手势缓存（连续3帧相同才确认） - self.gesture_buffer = deque(maxlen=3) - self.last_gesture = "None" - - def get_roi(self, image): - """截取固定检测区域""" - return image[self.roi_y1:self.roi_y2, self.roi_x1:self.roi_x2] - - def process(self, image): - # 1. 截取ROI - roi = self.get_roi(image) - if roi.size == 0: - return "None", [] - - # 2. 预处理（降噪+肤色检测） - blur = cv.GaussianBlur(roi, (7, 7), 0) - hsv = cv.cvtColor(blur, cv.COLOR_BGR2HSV) - mask = cv.inRange(hsv, self.lower_skin, self.upper_skin) - mask = cv.morphologyEx(mask, cv.MORPH_CLOSE, np.ones((7, 7), np.uint8)) - - # 3. 找手部轮廓 - contours, _ = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE) - if not contours: - self.gesture_buffer.append("None") - return self._get_stable_gesture(), [] - - max_contour = max(contours, key=cv.contourArea) - if cv.contourArea(max_contour) < 5000: - self.gesture_buffer.append("None") - return self._get_stable_gesture(), [] - - # 4. 手势判断（仅保留拳头/点手势，最稳定） - hull = cv.convexHull(max_contour) - solidity = cv.contourArea(max_contour) / cv.contourArea(hull) - current_gesture = "Fist" if solidity > 0.85 else "Point" - self.gesture_buffer.append(current_gesture) - - # 5. 稳定输出（连续3帧相同） - return self._get_stable_gesture(), max_contour - - def _get_stable_gesture(self): - """只有连续3帧相同才输出""" - if len(self.gesture_buffer) < 3: - return self.last_gesture - if len(set(self.gesture_buffer)) == 1: - self.last_gesture = self.gesture_buffer[0] - return self.last_gesture - - -# ========== 主函数（带ROI框+稳定显示） ========== +# 极简手势识别（仅保留拳头/点手势，极致流畅） def main(): + # 1. 摄像头初始化（极简参数） cap = cv.VideoCapture(0) - cap.set(cv.CAP_PROP_FRAME_WIDTH, 960) - cap.set(cv.CAP_PROP_FRAME_HEIGHT, 540) - cap.set(cv.CAP_PROP_EXPOSURE, -6) # 固定曝光，避免过曝 - recognizer = StableHandRecognizer() - fps_calc = deque(maxlen=10) + cap.set(cv.CAP_PROP_FRAME_WIDTH, 320) # 极低分辨率，秒杀卡顿 + cap.set(cv.CAP_PROP_FRAME_HEIGHT, 240) + cap.set(cv.CAP_PROP_FOURCC, cv.VideoWriter_fourcc(*'MJPG')) # 快速编码 + cap.set(cv.CAP_PROP_BUFFERSIZE, 1) # 关闭缓存，降低延迟 + + # 2. 固定参数（适配所有摄像头） + skin_lower = np.array([0, 10, 10], np.uint8) + skin_upper = np.array([30, 255, 180], np.uint8) + kernel = np.ones((3, 3), np.uint8) + last_gesture = "None" + gesture_count = 0 + + print("✅ 极致轻量化手势识别 | ESC退出") + print("💡 把手放在画面中间，握拳=Fist，伸食指=Point") while True: - # 计时算FPS - start = time.time() + # 计时（极简FPS） + t1 = time.time() + + # 3. 读取帧（跳过缓存帧） ret, frame = cap.read() if not ret: break frame = cv.flip(frame, 1) - debug_frame = frame.copy() + frame_small = cv.resize(frame, (160, 120)) # 超小尺寸处理 - # 1. 绘制ROI框（提示用户把手放在这里） - cv.rectangle(debug_frame, (recognizer.roi_x1, recognizer.roi_y1), - (recognizer.roi_x2, recognizer.roi_y2), (0, 255, 255), 2) - cv.putText(debug_frame, "Put hand here", (recognizer.roi_x1 + 10, recognizer.roi_y1 - 10), - cv.FONT_HERSHEY_SIMPLEX, 0.7, (0, 255, 255), 2) + # 4. 极简手部检测 + hsv = cv.cvtColor(frame_small, cv.COLOR_BGR2HSV) + mask = cv.inRange(hsv, skin_lower, skin_upper) + mask = cv.morphologyEx(mask, cv.MORPH_OPEN, kernel) - # 2. 识别手势 - gesture, contour = recognizer.process(frame) - - # 3. 绘制结果（固定位置，不闪烁） - cv.putText(debug_frame, f"Stable Gesture: {gesture}", (50, 50), - cv.FONT_HERSHEY_SIMPLEX, 1.2, (0, 0, 255), 4) - - # 4. 绘制手部轮廓（ROI内） - if contour is not None and len(contour) > 0: - # 转换轮廓坐标到全局画面 - contour[:, :, 0] += recognizer.roi_x1 - contour[:, :, 1] += recognizer.roi_y1 - cv.drawContours(debug_frame, [contour], -1, (0, 255, 0), 2) - - # 计算FPS - fps = 1 / (time.time() - start) - fps_calc.append(fps) - cv.putText(debug_frame, f"FPS: {int(np.mean(fps_calc))}", (50, 100), - cv.FONT_HERSHEY_SIMPLEX, 1.0, (255, 0, 0), 3) + # 5. 找轮廓（只找最大的） + contours, _ = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE) + current_gesture = "None" + if contours: + cnt = max(contours, key=cv.contourArea) + if cv.contourArea(cnt) > 1000: + # 6. 极简分类（仅拳头/点手势） + hull = cv.convexHull(cnt) + solidity = cv.contourArea(cnt) / cv.contourArea(hull) + current_gesture = "Fist" if solidity > 0.85 else "Point" + + # 7. 稳定输出（连续2帧相同） + if current_gesture == last_gesture: + gesture_count += 1 + else: + gesture_count = 0 + last_gesture = current_gesture + stable_gesture = last_gesture if gesture_count > 1 else "None" + + # 8. 绘制（极简UI，减少计算） + cv.putText(frame, f"Gesture: {stable_gesture}", (10, 30), + cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255), 2) + cv.putText(frame, f"FPS: {int(1 / (time.time() - t1))}", (10, 60), + cv.FONT_HERSHEY_SIMPLEX, 0.8, (255, 0, 0), 2) + + # 9. 显示（拉伸回原尺寸，保持清晰） + frame_show = cv.resize(frame, (640, 480)) + cv.imshow("Ultra Light Gesture", frame_show) - # 显示 - cv.imshow("Stable Hand Gesture", debug_frame) if cv.waitKey(1) & 0xFF == 27: break From d15abc4fbe774890c5b8a6b08ecae62d8e5ff575 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Tue, 23 Dec 2025 16:54:51 +0800 Subject: [PATCH 36/45] =?UTF-8?q?=E4=BD=BF=E7=94=BB=E9=9D=A2=E5=B8=A7?= =?UTF-8?q?=E7=8E=87=E6=9B=B4=E7=A8=B3=E5=AE=9A?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 160 ++++++++++++++++++---------- 1 file changed, 105 insertions(+), 55 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index 5dcf5c81f1..15eeee39e4 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -3,78 +3,128 @@ import cv2 as cv import numpy as np import time +import threading -# 极简手势识别（仅保留拳头/点手势，极致流畅） -def main(): - # 1. 摄像头初始化（极简参数） - cap = cv.VideoCapture(0) - cap.set(cv.CAP_PROP_FRAME_WIDTH, 320) # 极低分辨率，秒杀卡顿 - cap.set(cv.CAP_PROP_FRAME_HEIGHT, 240) - cap.set(cv.CAP_PROP_FOURCC, cv.VideoWriter_fourcc(*'MJPG')) # 快速编码 - cap.set(cv.CAP_PROP_BUFFERSIZE, 1) # 关闭缓存，降低延迟 - - # 2. 固定参数（适配所有摄像头） - skin_lower = np.array([0, 10, 10], np.uint8) - skin_upper = np.array([30, 255, 180], np.uint8) - kernel = np.ones((3, 3), np.uint8) - last_gesture = "None" - gesture_count = 0 - - print("✅ 极致轻量化手势识别 | ESC退出") - print("💡 把手放在画面中间，握拳=Fist，伸食指=Point") - - while True: - # 计时（极简FPS） - t1 = time.time() - - # 3. 读取帧（跳过缓存帧） - ret, frame = cap.read() - if not ret: - break - frame = cv.flip(frame, 1) - frame_small = cv.resize(frame, (160, 120)) # 超小尺寸处理 +class StableFPSHandRecognizer: + def __init__(self, target_fps=30): + # 1. 帧率锁定参数 + self.target_fps = target_fps + self.frame_interval = 1.0 / target_fps # 每帧间隔时间（秒） + self.last_frame_time = time.time() + + # 2. 极简手部检测参数 + self.skin_lower = np.array([0, 10, 10], np.uint8) + self.skin_upper = np.array([30, 255, 180], np.uint8) + self.kernel = np.ones((3, 3), np.uint8) + + # 3. 手势缓存（仅2帧，快速响应+稳定） + self.gesture_buffer = [] + self.stable_gesture = "None" + + # 4. 帧缓存（避免堆积） + self.frame_queue = [] + self.queue_lock = threading.Lock() - # 4. 极简手部检测 + def capture_frames(self, cap): + """独立线程采集帧，避免主线程阻塞""" + while cap.isOpened(): + ret, frame = cap.read() + if not ret: + break + with self.queue_lock: + # 只保留最新1帧，避免堆积 + self.frame_queue = [frame] + # 采集线程限速，匹配目标帧率 + time.sleep(self.frame_interval * 0.5) + + def process_frame(self, frame): + """轻量化处理，严格控制耗时""" + # 1. 快速预处理 + frame = cv.flip(frame, 1) + frame_small = cv.resize(frame, (160, 120)) # 超小尺寸 hsv = cv.cvtColor(frame_small, cv.COLOR_BGR2HSV) - mask = cv.inRange(hsv, skin_lower, skin_upper) - mask = cv.morphologyEx(mask, cv.MORPH_OPEN, kernel) + mask = cv.inRange(hsv, self.skin_lower, self.skin_upper) + mask = cv.morphologyEx(mask, cv.MORPH_OPEN, self.kernel) - # 5. 找轮廓（只找最大的） + # 2. 快速找轮廓 contours, _ = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE) current_gesture = "None" if contours: cnt = max(contours, key=cv.contourArea) if cv.contourArea(cnt) > 1000: - # 6. 极简分类（仅拳头/点手势） + # 3. 极简分类 hull = cv.convexHull(cnt) solidity = cv.contourArea(cnt) / cv.contourArea(hull) current_gesture = "Fist" if solidity > 0.85 else "Point" - # 7. 稳定输出（连续2帧相同） - if current_gesture == last_gesture: - gesture_count += 1 - else: - gesture_count = 0 - last_gesture = current_gesture - stable_gesture = last_gesture if gesture_count > 1 else "None" - - # 8. 绘制（极简UI，减少计算） - cv.putText(frame, f"Gesture: {stable_gesture}", (10, 30), - cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255), 2) - cv.putText(frame, f"FPS: {int(1 / (time.time() - t1))}", (10, 60), - cv.FONT_HERSHEY_SIMPLEX, 0.8, (255, 0, 0), 2) - - # 9. 显示（拉伸回原尺寸，保持清晰） + # 4. 稳定手势（仅2帧一致） + self.gesture_buffer.append(current_gesture) + if len(self.gesture_buffer) > 2: + self.gesture_buffer.pop(0) + if len(set(self.gesture_buffer)) == 1: + self.stable_gesture = self.gesture_buffer[0] + + # 5. 绘制极简UI（控制绘制耗时） + cv.putText(frame, f"Gesture: {self.stable_gesture}", (10, 40), + cv.FONT_HERSHEY_SIMPLEX, 1.2, (0, 0, 255), 3) + cv.putText(frame, f"FPS: {self.target_fps}", (10, 80), + cv.FONT_HERSHEY_SIMPLEX, 1.0, (255, 0, 0), 2) + + # 拉伸显示（保持清晰） frame_show = cv.resize(frame, (640, 480)) - cv.imshow("Ultra Light Gesture", frame_show) + return frame_show + + def run(self): + """主运行逻辑，帧率锁死""" + # 1. 摄像头初始化（硬件级优化） + cap = cv.VideoCapture(0) + cap.set(cv.CAP_PROP_FRAME_WIDTH, 320) + cap.set(cv.CAP_PROP_FRAME_HEIGHT, 240) + cap.set(cv.CAP_PROP_FOURCC, cv.VideoWriter_fourcc(*'MJPG')) # 快速编码 + cap.set(cv.CAP_PROP_BUFFERSIZE, 1) # 关闭缓存 + cap.set(cv.CAP_PROP_FPS, self.target_fps) # 强制摄像头输出目标帧率 + + # 2. 启动独立采集线程 + capture_thread = threading.Thread(target=self.capture_frames, args=(cap,), daemon=True) + capture_thread.start() + + print(f"✅ 帧率锁定 {self.target_fps} 帧 | ESC退出") + print("💡 把手放在画面中间，握拳=Fist，伸食指=Point") + + # 3. 主线程处理+显示（严格控时） + while cap.isOpened(): + # 计算当前帧应执行的时间，确保帧率稳定 + current_time = time.time() + elapsed = current_time - self.last_frame_time + + # 如果耗时不足，等待到目标间隔 + if elapsed < self.frame_interval: + time.sleep(self.frame_interval - elapsed) + + # 读取最新帧 + with self.queue_lock: + if not self.frame_queue: + continue + frame = self.frame_queue.pop(0) + + # 处理并显示 + frame_show = self.process_frame(frame) + cv.imshow("Stable FPS Gesture", frame_show) + + # 更新时间戳，确保下一帧同步 + self.last_frame_time = time.time() - if cv.waitKey(1) & 0xFF == 27: - break + # ESC退出 + if cv.waitKey(1) & 0xFF == 27: + break - cap.release() - cv.destroyAllWindows() + # 释放资源 + cap.release() + cv.destroyAllWindows() if __name__ == '__main__': - main() \ No newline at end of file + # 实例化并运行，锁定30帧（可改20/15帧，更低更稳） + recognizer = StableFPSHandRecognizer(target_fps=30) + recognizer.run() \ No newline at end of file From f507f0e2337a7df316d8352de969795d98551597 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Wed, 24 Dec 2025 16:23:10 +0800 Subject: [PATCH 37/45] =?UTF-8?q?=E6=B7=BB=E5=8A=A0=E4=BA=86=E5=AF=B9?= =?UTF-8?q?=E4=BA=94=E6=8C=87=E5=BC=A0=E5=BC=80=E7=9A=84=E6=89=8B=E5=8A=BF?= =?UTF-8?q?=E8=AF=86=E5=88=AB?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 70 ++++++++++++++++++++++++++--- 1 file changed, 65 insertions(+), 5 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index 15eeee39e4..5a3be70a1d 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -18,6 +18,11 @@ def __init__(self, target_fps=30): self.skin_upper = np.array([30, 255, 180], np.uint8) self.kernel = np.ones((3, 3), np.uint8) + # 新增：手指检测参数 + self.defect_depth_threshold = 20 # 凸包缺陷深度阈值 + self.min_defect_distance = 10 # 缺陷点最小距离 + self.palm_solidity_threshold = 0.6 # 手掌的密实度阈值 + # 3. 手势缓存（仅2帧，快速响应+稳定） self.gesture_buffer = [] self.stable_gesture = "None" @@ -26,6 +31,42 @@ def __init__(self, target_fps=30): self.frame_queue = [] self.queue_lock = threading.Lock() + def count_fingers(self, cnt, frame_small): + """通过凸包缺陷计算手指数量""" + try: + # 计算凸包和凸包缺陷 + hull = cv.convexHull(cnt, returnPoints=False) + defects = cv.convexityDefects(cnt, hull) + + if defects is None: + return 0 + + finger_count = 0 + defect_points = [] + + # 遍历所有凸包缺陷 + for i in range(defects.shape[0]): + s, e, f, d = defects[i, 0] + start = tuple(cnt[s][0]) + end = tuple(cnt[e][0]) + far = tuple(cnt[f][0]) + + # 计算缺陷深度（转换为实际像素值） + depth = d / 256.0 + + # 只考虑深度足够的缺陷（手指间的凹陷） + if depth > self.defect_depth_threshold: + # 计算两点间距离，避免重复计数 + if all(np.linalg.norm(np.array(far) - np.array(p)) > self.min_defect_distance for p in + defect_points): + defect_points.append(far) + finger_count += 1 + + # 缺陷数+1 = 手指数量（例如：4个缺陷=5根手指） + return min(finger_count + 1, 5) # 最多5根手指 + except: + return 0 + def capture_frames(self, cap): """独立线程采集帧，避免主线程阻塞""" while cap.isOpened(): @@ -50,13 +91,32 @@ def process_frame(self, frame): # 2. 快速找轮廓 contours, _ = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE) current_gesture = "None" + if contours: cnt = max(contours, key=cv.contourArea) - if cv.contourArea(cnt) > 1000: - # 3. 极简分类 + area = cv.contourArea(cnt) + + if area > 1000: + # 3. 手势分类（新增五指识别） hull = cv.convexHull(cnt) solidity = cv.contourArea(cnt) / cv.contourArea(hull) - current_gesture = "Fist" if solidity > 0.85 else "Point" + + # 计算手指数量 + finger_count = self.count_fingers(cnt, frame_small) + + # 手势判断逻辑 + if solidity > 0.85: + # 密实度高 = 握拳 + current_gesture = "Fist" + elif finger_count == 1: + # 1根手指 = 单指 + current_gesture = "Point" + elif finger_count >= 4: + # 4-5根手指 = 手掌张开 + current_gesture = "Palm" + elif 2 <= finger_count <= 3: + # 2-3根手指 = 部分张开（归类为Point） + current_gesture = "Point" # 4. 稳定手势（仅2帧一致） self.gesture_buffer.append(current_gesture) @@ -65,7 +125,7 @@ def process_frame(self, frame): if len(set(self.gesture_buffer)) == 1: self.stable_gesture = self.gesture_buffer[0] - # 5. 绘制极简UI（控制绘制耗时） + # 5. 绘制极简UI（仅保留手势和FPS显示，移除手指数量） cv.putText(frame, f"Gesture: {self.stable_gesture}", (10, 40), cv.FONT_HERSHEY_SIMPLEX, 1.2, (0, 0, 255), 3) cv.putText(frame, f"FPS: {self.target_fps}", (10, 80), @@ -90,7 +150,7 @@ def run(self): capture_thread.start() print(f"✅ 帧率锁定 {self.target_fps} 帧 | ESC退出") - print("💡 把手放在画面中间，握拳=Fist，伸食指=Point") + print("💡 把手放在画面中间，握拳=Fist，伸食指=Point，五指张开=Palm") # 3. 主线程处理+显示（严格控时） while cap.isOpened(): From 032575ad26e3f85b2537f5401a144a545009e137 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Wed, 24 Dec 2025 19:28:02 +0800 Subject: [PATCH 38/45] =?UTF-8?q?=E4=BF=AE=E6=94=B9=E4=BA=86=E5=AF=B9?= =?UTF-8?q?=E6=89=8B=E6=8C=87=E7=9A=84=E8=AF=86=E5=88=AB?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 21 ++++++++++++--------- 1 file changed, 12 insertions(+), 9 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index 5a3be70a1d..027fa8067d 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -97,26 +97,29 @@ def process_frame(self, frame): area = cv.contourArea(cnt) if area > 1000: - # 3. 手势分类（新增五指识别） + # 3. 手势分类（修改Point为仅食指+中指（2根手指）） hull = cv.convexHull(cnt) solidity = cv.contourArea(cnt) / cv.contourArea(hull) # 计算手指数量 finger_count = self.count_fingers(cnt, frame_small) - # 手势判断逻辑 + # 手势判断逻辑（核心修改） if solidity > 0.85: # 密实度高 = 握拳 current_gesture = "Fist" - elif finger_count == 1: - # 1根手指 = 单指 + elif finger_count == 2: + # 仅2根手指 = 食指+中指（Point） current_gesture = "Point" elif finger_count >= 4: # 4-5根手指 = 手掌张开 current_gesture = "Palm" - elif 2 <= finger_count <= 3: - # 2-3根手指 = 部分张开（归类为Point） - current_gesture = "Point" + elif finger_count == 1: + # 1根手指 = 单指（归为None或单独分类，这里保持None） + current_gesture = "None" + elif finger_count == 3: + # 3根手指 = 归为None + current_gesture = "None" # 4. 稳定手势（仅2帧一致） self.gesture_buffer.append(current_gesture) @@ -125,7 +128,7 @@ def process_frame(self, frame): if len(set(self.gesture_buffer)) == 1: self.stable_gesture = self.gesture_buffer[0] - # 5. 绘制极简UI（仅保留手势和FPS显示，移除手指数量） + # 5. 绘制极简UI（仅保留手势和FPS显示） cv.putText(frame, f"Gesture: {self.stable_gesture}", (10, 40), cv.FONT_HERSHEY_SIMPLEX, 1.2, (0, 0, 255), 3) cv.putText(frame, f"FPS: {self.target_fps}", (10, 80), @@ -150,7 +153,7 @@ def run(self): capture_thread.start() print(f"✅ 帧率锁定 {self.target_fps} 帧 | ESC退出") - print("💡 把手放在画面中间，握拳=Fist，伸食指=Point，五指张开=Palm") + print("💡 把手放在画面中间，握拳=Fist，伸食指+中指=Point，五指张开=Palm") # 3. 主线程处理+显示（严格控时） while cap.isOpened(): From 89e210e6401199ccbf60d5de05ab71bc178e3de4 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Wed, 24 Dec 2025 21:22:20 +0800 Subject: [PATCH 39/45] =?UTF-8?q?=E4=BF=AE=E6=94=B9=E4=BA=86=E6=89=8B?= =?UTF-8?q?=E5=8A=BF=E8=AF=86=E5=88=AB=E5=90=8E=E7=9A=84=E8=BE=93=E5=87=BA?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 23 ++++++++++++----------- 1 file changed, 12 insertions(+), 11 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index 027fa8067d..a2e7c6072e 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -97,25 +97,25 @@ def process_frame(self, frame): area = cv.contourArea(cnt) if area > 1000: - # 3. 手势分类（修改Point为仅食指+中指（2根手指）） + # 3. 手势分类（修改输出文本映射：Fist→stop, Point→front, Palm→back） hull = cv.convexHull(cnt) solidity = cv.contourArea(cnt) / cv.contourArea(hull) # 计算手指数量 finger_count = self.count_fingers(cnt, frame_small) - # 手势判断逻辑（核心修改） + # 手势判断逻辑（仅修改输出文本） if solidity > 0.85: - # 密实度高 = 握拳 - current_gesture = "Fist" + # 密实度高 = 握拳 → 输出stop + current_gesture = "stop" elif finger_count == 2: - # 仅2根手指 = 食指+中指（Point） - current_gesture = "Point" + # 仅2根手指 = 食指+中指 → 输出front + current_gesture = "front" elif finger_count >= 4: - # 4-5根手指 = 手掌张开 - current_gesture = "Palm" + # 4-5根手指 = 手掌张开 → 输出back + current_gesture = "back" elif finger_count == 1: - # 1根手指 = 单指（归为None或单独分类，这里保持None） + # 1根手指 = 单指（归为None） current_gesture = "None" elif finger_count == 3: # 3根手指 = 归为None @@ -128,7 +128,7 @@ def process_frame(self, frame): if len(set(self.gesture_buffer)) == 1: self.stable_gesture = self.gesture_buffer[0] - # 5. 绘制极简UI（仅保留手势和FPS显示） + # 5. 绘制极简UI（显示修改后的手势文本） cv.putText(frame, f"Gesture: {self.stable_gesture}", (10, 40), cv.FONT_HERSHEY_SIMPLEX, 1.2, (0, 0, 255), 3) cv.putText(frame, f"FPS: {self.target_fps}", (10, 80), @@ -152,8 +152,9 @@ def run(self): capture_thread = threading.Thread(target=self.capture_frames, args=(cap,), daemon=True) capture_thread.start() + # 修改控制台提示文本，匹配新的输出 print(f"✅ 帧率锁定 {self.target_fps} 帧 | ESC退出") - print("💡 把手放在画面中间，握拳=Fist，伸食指+中指=Point，五指张开=Palm") + print("💡 把手放在画面中间，握拳=stop，伸食指+中指=front，五指张开=back") # 3. 主线程处理+显示（严格控时） while cap.isOpened(): From b40eaa4357ab223e6f577a9d0689af0c224e3a56 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 25 Dec 2025 10:01:48 +0800 Subject: [PATCH 40/45] =?UTF-8?q?=E6=B7=BB=E5=8A=A0=E4=BA=86=E8=AF=86?= =?UTF-8?q?=E5=88=AB=E5=8C=BA=E5=9F=9F?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 247 +++++++++++++++++----------- 1 file changed, 150 insertions(+), 97 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index a2e7c6072e..a80ae4c846 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -10,163 +10,216 @@ class StableFPSHandRecognizer: def __init__(self, target_fps=30): # 1. 帧率锁定参数 self.target_fps = target_fps - self.frame_interval = 1.0 / target_fps # 每帧间隔时间（秒） + self.frame_interval = 1.0 / target_fps self.last_frame_time = time.time() - # 2. 极简手部检测参数 - self.skin_lower = np.array([0, 10, 10], np.uint8) - self.skin_upper = np.array([30, 255, 180], np.uint8) - self.kernel = np.ones((3, 3), np.uint8) + # 2. 优化后的肤色检测阈值（适配更多肤色/光线） + self.skin_lower = np.array([0, 20, 70], np.uint8) # 放宽下界 + self.skin_upper = np.array([20, 255, 255], np.uint8) # 调整上界 + self.kernel = np.ones((5, 5), np.uint8) # 更大的核去噪 - # 新增：手指检测参数 - self.defect_depth_threshold = 20 # 凸包缺陷深度阈值 - self.min_defect_distance = 10 # 缺陷点最小距离 - self.palm_solidity_threshold = 0.6 # 手掌的密实度阈值 + # 3. 优化后的手指检测参数（降低阈值，提高识别率） + self.defect_depth_threshold = 10 # 降低深度阈值 + self.min_defect_distance = 5 # 降低距离阈值 + self.min_contour_area = 500 # 降低最小轮廓面积 - # 3. 手势缓存（仅2帧，快速响应+稳定） + # 4. 手势缓存&帧缓存 self.gesture_buffer = [] self.stable_gesture = "None" - - # 4. 帧缓存（避免堆积） self.frame_queue = [] self.queue_lock = threading.Lock() - def count_fingers(self, cnt, frame_small): - """通过凸包缺陷计算手指数量""" + # 5. 识别区域参数（仅显示边框） + self.recognition_area = None + self.area_color = (0, 255, 0) # 边框颜色（绿色） + + def _init_recognition_area(self, frame_shape): + """初始化识别区域（调大尺寸，右侧更大范围）""" + h, w = frame_shape[:2] + x1 = int(w * 1.5 / 3) # 左边界左移（从2/3改为1.5/3），扩大宽度 + y1 = int(h * 0.05) # 上边界上移（从0.1改为0.05），扩大高度 + x2 = w - 10 # 右边界右移（从-20改为-10），减少右侧边距 + y2 = int(h * 0.95) # 下边界下移（从0.9改为0.95），减少底部边距 + self.recognition_area = (x1, y1, x2, y2) + + def _draw_recognition_area(self, frame): + """绘制识别区域（仅显示边框，无背景色）""" + if self.recognition_area is None: + self._init_recognition_area(frame.shape) + x1, y1, x2, y2 = self.recognition_area + + # 仅绘制边框（移除半透明背景） + cv.rectangle(frame, (x1, y1), (x2, y2), self.area_color, 2) + + # 添加区域提示文字（在边框上方） + cv.putText(frame, "Recognition Area", (x1 + 10, y1 - 10), + cv.FONT_HERSHEY_SIMPLEX, 0.6, self.area_color, 2) + return frame + + def _get_roi(self, frame): + """获取识别区域的ROI（确保坐标有效）""" + if self.recognition_area is None: + self._init_recognition_area(frame.shape) + x1, y1, x2, y2 = self.recognition_area + + # 边界保护 + x1 = max(0, x1) + y1 = max(0, y1) + x2 = min(frame.shape[1], x2) + y2 = min(frame.shape[0], y2) + + return frame[y1:y2, x1:x2], (x1, y1) + + def count_fingers(self, cnt): + """优化后的手指计数逻辑（更鲁棒）""" try: - # 计算凸包和凸包缺陷 - hull = cv.convexHull(cnt, returnPoints=False) - defects = cv.convexityDefects(cnt, hull) + # 计算凸包（带坐标）和凸包缺陷 + hull = cv.convexHull(cnt) + hull_indices = cv.convexHull(cnt, returnPoints=False) + defects = cv.convexityDefects(cnt, hull_indices) - if defects is None: + if defects is None or len(defects) == 0: return 0 finger_count = 0 - defect_points = [] - - # 遍历所有凸包缺陷 + # 遍历缺陷点 for i in range(defects.shape[0]): s, e, f, d = defects[i, 0] start = tuple(cnt[s][0]) end = tuple(cnt[e][0]) far = tuple(cnt[f][0]) - # 计算缺陷深度（转换为实际像素值） + # 计算缺陷深度（实际像素值） depth = d / 256.0 - # 只考虑深度足够的缺陷（手指间的凹陷） - if depth > self.defect_depth_threshold: - # 计算两点间距离，避免重复计数 - if all(np.linalg.norm(np.array(far) - np.array(p)) > self.min_defect_distance for p in - defect_points): - defect_points.append(far) - finger_count += 1 - - # 缺陷数+1 = 手指数量（例如：4个缺陷=5根手指） - return min(finger_count + 1, 5) # 最多5根手指 - except: + # 计算角度（过滤误判的缺陷） + a = np.linalg.norm(np.array(end) - np.array(start)) + b = np.linalg.norm(np.array(far) - np.array(start)) + c = np.linalg.norm(np.array(end) - np.array(far)) + angle = np.arccos((b ** 2 + c ** 2 - a ** 2) / (2 * b * c)) * 180 / np.pi + + # 有效缺陷：深度足够 + 角度小于90度（手指间的凹陷） + if depth > self.defect_depth_threshold and angle < 90: + finger_count += 1 + + # 缺陷数+1=手指数量（最多5根） + return min(finger_count + 1, 5) + except Exception as e: + print(f"手指计数错误: {e}") return 0 def capture_frames(self, cap): - """独立线程采集帧，避免主线程阻塞""" + """帧采集线程（稳定）""" while cap.isOpened(): ret, frame = cap.read() if not ret: break with self.queue_lock: - # 只保留最新1帧，避免堆积 - self.frame_queue = [frame] - # 采集线程限速，匹配目标帧率 + self.frame_queue = [frame] # 只保留最新帧 time.sleep(self.frame_interval * 0.5) def process_frame(self, frame): - """轻量化处理，严格控制耗时""" - # 1. 快速预处理 + """优化后的帧处理逻辑""" + # 镜像翻转 frame = cv.flip(frame, 1) - frame_small = cv.resize(frame, (160, 120)) # 超小尺寸 - hsv = cv.cvtColor(frame_small, cv.COLOR_BGR2HSV) - mask = cv.inRange(hsv, self.skin_lower, self.skin_upper) - mask = cv.morphologyEx(mask, cv.MORPH_OPEN, self.kernel) - - # 2. 快速找轮廓 - contours, _ = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE) + # 绘制识别区域（仅边框） + frame = self._draw_recognition_area(frame) + # 获取ROI + roi, (roi_x, roi_y) = self._get_roi(frame) current_gesture = "None" - if contours: - cnt = max(contours, key=cv.contourArea) - area = cv.contourArea(cnt) - - if area > 1000: - # 3. 手势分类（修改输出文本映射：Fist→stop, Point→front, Palm→back） - hull = cv.convexHull(cnt) - solidity = cv.contourArea(cnt) / cv.contourArea(hull) - - # 计算手指数量 - finger_count = self.count_fingers(cnt, frame_small) - - # 手势判断逻辑（仅修改输出文本） - if solidity > 0.85: - # 密实度高 = 握拳 → 输出stop - current_gesture = "stop" - elif finger_count == 2: - # 仅2根手指 = 食指+中指 → 输出front - current_gesture = "front" - elif finger_count >= 4: - # 4-5根手指 = 手掌张开 → 输出back - current_gesture = "back" - elif finger_count == 1: - # 1根手指 = 单指（归为None） - current_gesture = "None" - elif finger_count == 3: - # 3根手指 = 归为None - current_gesture = "None" - - # 4. 稳定手势（仅2帧一致） + if roi.size > 0: # 确保ROI有效 + # 预处理：缩小+转HSV+肤色掩码 + roi_small = cv.resize(roi, (320, 240)) # 适度放大ROI，提高检测精度 + hsv = cv.cvtColor(roi_small, cv.COLOR_BGR2HSV) + mask = cv.inRange(hsv, self.skin_lower, self.skin_upper) + + # 形态学操作（去噪+填充） + mask = cv.morphologyEx(mask, cv.MORPH_OPEN, self.kernel) + mask = cv.morphologyEx(mask, cv.MORPH_CLOSE, self.kernel) + mask = cv.dilate(mask, self.kernel, iterations=2) + + # 查找轮廓 + contours, _ = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE) + if contours: + # 取最大轮廓（手部） + cnt = max(contours, key=cv.contourArea) + area = cv.contourArea(cnt) + + if area > self.min_contour_area: + # 计算密实度（判断是否握拳） + hull = cv.convexHull(cnt) + hull_area = cv.contourArea(hull) + solidity = area / hull_area if hull_area > 0 else 0 + + # 手指计数 + finger_count = self.count_fingers(cnt) + + # 可视化调试（可选：在ROI内绘制轮廓） + cnt_scaled = cnt * (roi.shape[1] / roi_small.shape[1], roi.shape[0] / roi_small.shape[0]) + cnt_scaled = cnt_scaled.astype(np.int32) + cnt_scaled[:, :, 0] += roi_x + cnt_scaled[:, :, 1] += roi_y + cv.drawContours(frame, [cnt_scaled], -1, (255, 0, 0), 2) + + # 手势判断逻辑（优化版） + if solidity > 0.8: # 握拳（密实度高） + current_gesture = "stop" + elif finger_count == 2: # 食指+中指 + current_gesture = "front" + elif finger_count >= 4: # 手掌张开（4-5指） + current_gesture = "back" + # 其他情况（1/3指）归为None + + # 手势缓存稳定 self.gesture_buffer.append(current_gesture) if len(self.gesture_buffer) > 2: self.gesture_buffer.pop(0) if len(set(self.gesture_buffer)) == 1: self.stable_gesture = self.gesture_buffer[0] - # 5. 绘制极简UI（显示修改后的手势文本） + # 绘制UI cv.putText(frame, f"Gesture: {self.stable_gesture}", (10, 40), cv.FONT_HERSHEY_SIMPLEX, 1.2, (0, 0, 255), 3) cv.putText(frame, f"FPS: {self.target_fps}", (10, 80), cv.FONT_HERSHEY_SIMPLEX, 1.0, (255, 0, 0), 2) - # 拉伸显示（保持清晰） + # 拉伸显示 frame_show = cv.resize(frame, (640, 480)) return frame_show def run(self): - """主运行逻辑，帧率锁死""" - # 1. 摄像头初始化（硬件级优化） + """主运行逻辑""" + # 摄像头初始化（优化参数） cap = cv.VideoCapture(0) - cap.set(cv.CAP_PROP_FRAME_WIDTH, 320) - cap.set(cv.CAP_PROP_FRAME_HEIGHT, 240) - cap.set(cv.CAP_PROP_FOURCC, cv.VideoWriter_fourcc(*'MJPG')) # 快速编码 - cap.set(cv.CAP_PROP_BUFFERSIZE, 1) # 关闭缓存 - cap.set(cv.CAP_PROP_FPS, self.target_fps) # 强制摄像头输出目标帧率 + cap.set(cv.CAP_PROP_FRAME_WIDTH, 640) # 提高摄像头分辨率 + cap.set(cv.CAP_PROP_FRAME_HEIGHT, 480) + cap.set(cv.CAP_PROP_FOURCC, cv.VideoWriter_fourcc(*'MJPG')) + cap.set(cv.CAP_PROP_BUFFERSIZE, 1) + cap.set(cv.CAP_PROP_FPS, self.target_fps) - # 2. 启动独立采集线程 + # 启动采集线程 capture_thread = threading.Thread(target=self.capture_frames, args=(cap,), daemon=True) capture_thread.start() - # 修改控制台提示文本，匹配新的输出 + # 提示信息 + print("=" * 50) print(f"✅ 帧率锁定 {self.target_fps} 帧 | ESC退出") - print("💡 把手放在画面中间，握拳=stop，伸食指+中指=front，五指张开=back") + print("💡 调试提示：") + print(" 1. 把手放在右侧绿色边框的识别区域内（已扩大范围）") + print(" 2. 握拳 → stop | 食指+中指 → front | 手掌张开 → back") + print(" 3. 蓝色轮廓表示检测到的手部区域") + print("=" * 50) - # 3. 主线程处理+显示（严格控时） + # 主循环 while cap.isOpened(): - # 计算当前帧应执行的时间，确保帧率稳定 + # 帧率控制 current_time = time.time() elapsed = current_time - self.last_frame_time - - # 如果耗时不足，等待到目标间隔 if elapsed < self.frame_interval: time.sleep(self.frame_interval - elapsed) - # 读取最新帧 + # 读取帧 with self.queue_lock: if not self.frame_queue: continue @@ -174,9 +227,9 @@ def run(self): # 处理并显示 frame_show = self.process_frame(frame) - cv.imshow("Stable FPS Gesture", frame_show) + cv.imshow("Hand Gesture Recognition", frame_show) - # 更新时间戳，确保下一帧同步 + # 更新时间戳 self.last_frame_time = time.time() # ESC退出 @@ -189,6 +242,6 @@ def run(self): if __name__ == '__main__': - # 实例化并运行，锁定30帧（可改20/15帧，更低更稳） - recognizer = StableFPSHandRecognizer(target_fps=30) + # 可降低帧率（如15）提高稳定性 + recognizer = StableFPSHandRecognizer(target_fps=20) recognizer.run() \ No newline at end of file From 50169711472851ee119624dc9c3d7f5bbbc31b48 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Thu, 25 Dec 2025 20:31:43 +0800 Subject: [PATCH 41/45] =?UTF-8?q?=E6=B7=BB=E5=8A=A0=E8=AF=86=E5=8C=BA?= =?UTF-8?q?=E5=88=AB?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 26 +++++++++++++------------- 1 file changed, 13 insertions(+), 13 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index a80ae4c846..c87ca83166 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -13,7 +13,7 @@ def __init__(self, target_fps=30): self.frame_interval = 1.0 / target_fps self.last_frame_time = time.time() - # 2. 优化后的肤色检测阈值（适配更多肤色/光线） + # 2. 优化后的肤色检测阈值 self.skin_lower = np.array([0, 20, 70], np.uint8) # 放宽下界 self.skin_upper = np.array([20, 255, 255], np.uint8) # 调整上界 self.kernel = np.ones((5, 5), np.uint8) # 更大的核去噪 @@ -31,7 +31,7 @@ def __init__(self, target_fps=30): # 5. 识别区域参数（仅显示边框） self.recognition_area = None - self.area_color = (0, 255, 0) # 边框颜色（绿色） + self.area_color = (0, 255, 0) # 边框颜色 def _init_recognition_area(self, frame_shape): """初始化识别区域（调大尺寸，右侧更大范围）""" @@ -98,11 +98,11 @@ def count_fingers(self, cnt): c = np.linalg.norm(np.array(end) - np.array(far)) angle = np.arccos((b ** 2 + c ** 2 - a ** 2) / (2 * b * c)) * 180 / np.pi - # 有效缺陷：深度足够 + 角度小于90度（手指间的凹陷） + # 有效缺陷：深度足够 + 角度小于90度 if depth > self.defect_depth_threshold and angle < 90: finger_count += 1 - # 缺陷数+1=手指数量（最多5根） + # 缺陷数+1=手指数量 return min(finger_count + 1, 5) except Exception as e: print(f"手指计数错误: {e}") @@ -130,7 +130,7 @@ def process_frame(self, frame): if roi.size > 0: # 确保ROI有效 # 预处理：缩小+转HSV+肤色掩码 - roi_small = cv.resize(roi, (320, 240)) # 适度放大ROI，提高检测精度 + roi_small = cv.resize(roi, (320, 240)) # 适度放大ROI hsv = cv.cvtColor(roi_small, cv.COLOR_BGR2HSV) mask = cv.inRange(hsv, self.skin_lower, self.skin_upper) @@ -142,12 +142,12 @@ def process_frame(self, frame): # 查找轮廓 contours, _ = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE) if contours: - # 取最大轮廓（手部） + # 取最大轮廓 cnt = max(contours, key=cv.contourArea) area = cv.contourArea(cnt) if area > self.min_contour_area: - # 计算密实度（判断是否握拳） + # 计算密实度 hull = cv.convexHull(cnt) hull_area = cv.contourArea(hull) solidity = area / hull_area if hull_area > 0 else 0 @@ -155,19 +155,19 @@ def process_frame(self, frame): # 手指计数 finger_count = self.count_fingers(cnt) - # 可视化调试（可选：在ROI内绘制轮廓） + # 可视化调试 cnt_scaled = cnt * (roi.shape[1] / roi_small.shape[1], roi.shape[0] / roi_small.shape[0]) cnt_scaled = cnt_scaled.astype(np.int32) cnt_scaled[:, :, 0] += roi_x cnt_scaled[:, :, 1] += roi_y cv.drawContours(frame, [cnt_scaled], -1, (255, 0, 0), 2) - # 手势判断逻辑（优化版） - if solidity > 0.8: # 握拳（密实度高） + # 手势判断逻辑 + if solidity > 0.8: # 握拳 current_gesture = "stop" elif finger_count == 2: # 食指+中指 current_gesture = "front" - elif finger_count >= 4: # 手掌张开（4-5指） + elif finger_count >= 4: # 手掌张开 current_gesture = "back" # 其他情况（1/3指）归为None @@ -184,13 +184,13 @@ def process_frame(self, frame): cv.putText(frame, f"FPS: {self.target_fps}", (10, 80), cv.FONT_HERSHEY_SIMPLEX, 1.0, (255, 0, 0), 2) - # 拉伸显示 + # 拉伸 frame_show = cv.resize(frame, (640, 480)) return frame_show def run(self): """主运行逻辑""" - # 摄像头初始化（优化参数） + # 摄像头初始化 cap = cv.VideoCapture(0) cap.set(cv.CAP_PROP_FRAME_WIDTH, 640) # 提高摄像头分辨率 cap.set(cv.CAP_PROP_FRAME_HEIGHT, 480) From dcb112254e60597f39fb3cc34a06377d8fdd8935 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Fri, 26 Dec 2025 16:48:57 +0800 Subject: [PATCH 42/45] =?UTF-8?q?=E4=BC=98=E5=8C=96=E6=89=8B=E5=8A=BF?= =?UTF-8?q?=E8=AF=86=E5=88=AB=E7=9A=84=E5=87=86=E7=A1=AE=E5=BA=A6?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 260 ++++++++++++++++------------ 1 file changed, 150 insertions(+), 110 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index c87ca83166..cd393590fb 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -13,169 +13,205 @@ def __init__(self, target_fps=30): self.frame_interval = 1.0 / target_fps self.last_frame_time = time.time() - # 2. 优化后的肤色检测阈值 - self.skin_lower = np.array([0, 20, 70], np.uint8) # 放宽下界 - self.skin_upper = np.array([20, 255, 255], np.uint8) # 调整上界 - self.kernel = np.ones((5, 5), np.uint8) # 更大的核去噪 - - # 3. 优化后的手指检测参数（降低阈值，提高识别率） - self.defect_depth_threshold = 10 # 降低深度阈值 - self.min_defect_distance = 5 # 降低距离阈值 - self.min_contour_area = 500 # 降低最小轮廓面积 - - # 4. 手势缓存&帧缓存 + # 2. 肤色检测（适配更多光线） + self.skin_lower = np.array([0, 10, 50], np.uint8) + self.skin_upper = np.array([30, 255, 255], np.uint8) + self.kernel = np.ones((5, 5), np.uint8) + + # 3. 核心参数（精准适配手势特征） + # 握拳参数（稳定识别） + self.fist_solidity = 0.82 # 降低握拳阈值，提高稳定性 + self.fist_area_ratio = 0.75 # 握拳凸包面积比 + # 手指计数参数 + self.defect_depth_threshold = 8 # 降低深度阈值，提高up识别率 + self.min_contour_area = 600 # 降低最小面积，适配小手掌 + # 大拇指识别参数（宽松但精准） + self.thumb_aspect_ratio = 0.45 # 放宽宽高比 + self.thumb_solidity_range = (0.55, 0.82) # 刚好卡在握拳阈值下 + self.thumb_defect_max = 2 # 允许2个缺陷（适配不同握法） + + # 4. 缓存参数（增加缓存提升稳定性） self.gesture_buffer = [] + self.buffer_size = 3 # 增加缓存到3帧，提升stop稳定性 self.stable_gesture = "None" self.frame_queue = [] self.queue_lock = threading.Lock() - # 5. 识别区域参数（仅显示边框） + # 5. 识别区域 self.recognition_area = None - self.area_color = (0, 255, 0) # 边框颜色 + self.area_color = (0, 255, 0) def _init_recognition_area(self, frame_shape): - """初始化识别区域（调大尺寸，右侧更大范围）""" + """初始化识别区域""" h, w = frame_shape[:2] - x1 = int(w * 1.5 / 3) # 左边界左移（从2/3改为1.5/3），扩大宽度 - y1 = int(h * 0.05) # 上边界上移（从0.1改为0.05），扩大高度 - x2 = w - 10 # 右边界右移（从-20改为-10），减少右侧边距 - y2 = int(h * 0.95) # 下边界下移（从0.9改为0.95），减少底部边距 + x1 = int(w * 1.5 / 3) + y1 = int(h * 0.05) + x2 = w - 10 + y2 = int(h * 0.95) self.recognition_area = (x1, y1, x2, y2) def _draw_recognition_area(self, frame): - """绘制识别区域（仅显示边框，无背景色）""" + """绘制识别区域边框""" if self.recognition_area is None: self._init_recognition_area(frame.shape) x1, y1, x2, y2 = self.recognition_area - - # 仅绘制边框（移除半透明背景） cv.rectangle(frame, (x1, y1), (x2, y2), self.area_color, 2) - - # 添加区域提示文字（在边框上方） cv.putText(frame, "Recognition Area", (x1 + 10, y1 - 10), cv.FONT_HERSHEY_SIMPLEX, 0.6, self.area_color, 2) return frame def _get_roi(self, frame): - """获取识别区域的ROI（确保坐标有效）""" + """获取识别区域ROI""" if self.recognition_area is None: self._init_recognition_area(frame.shape) x1, y1, x2, y2 = self.recognition_area - - # 边界保护 x1 = max(0, x1) y1 = max(0, y1) x2 = min(frame.shape[1], x2) y2 = min(frame.shape[0], y2) - return frame[y1:y2, x1:x2], (x1, y1) - def count_fingers(self, cnt): - """优化后的手指计数逻辑（更鲁棒）""" + def analyze_contour(self, cnt): + """轮廓综合分析（返回多维度特征）""" try: - # 计算凸包（带坐标）和凸包缺陷 + # 基础特征 + area = cv.contourArea(cnt) + x, y, w, h = cv.boundingRect(cnt) + aspect_ratio = float(w) / h if h > 0 else 0 + + # 凸包特征 hull = cv.convexHull(cnt) + hull_area = cv.contourArea(hull) + solidity = area / hull_area if hull_area > 0 else 0 + hull_width = hull[:, 0, 0].max() - hull[:, 0, 0].min() if hull.size > 0 else 0 + hull_height = hull[:, 0, 1].max() - hull[:, 0, 1].min() if hull.size > 0 else 0 + hull_aspect = hull_width / hull_height if hull_height > 0 else 0 + + # 缺陷特征 hull_indices = cv.convexHull(cnt, returnPoints=False) defects = cv.convexityDefects(cnt, hull_indices) - - if defects is None or len(defects) == 0: - return 0 - - finger_count = 0 - # 遍历缺陷点 - for i in range(defects.shape[0]): - s, e, f, d = defects[i, 0] - start = tuple(cnt[s][0]) - end = tuple(cnt[e][0]) - far = tuple(cnt[f][0]) - - # 计算缺陷深度（实际像素值） - depth = d / 256.0 - - # 计算角度（过滤误判的缺陷） - a = np.linalg.norm(np.array(end) - np.array(start)) - b = np.linalg.norm(np.array(far) - np.array(start)) - c = np.linalg.norm(np.array(end) - np.array(far)) - angle = np.arccos((b ** 2 + c ** 2 - a ** 2) / (2 * b * c)) * 180 / np.pi - - # 有效缺陷：深度足够 + 角度小于90度 - if depth > self.defect_depth_threshold and angle < 90: - finger_count += 1 - - # 缺陷数+1=手指数量 - return min(finger_count + 1, 5) + defect_count = 0 + valid_defects = [] + + if defects is not None and len(defects) > 0: + for i in range(defects.shape[0]): + s, e, f, d = defects[i, 0] + depth = d / 256.0 + # 计算缺陷角度 + start = tuple(cnt[s][0]) + end = tuple(cnt[e][0]) + far = tuple(cnt[f][0]) + a = np.linalg.norm(np.array(end) - np.array(start)) + b = np.linalg.norm(np.array(far) - np.array(start)) + c = np.linalg.norm(np.array(end) - np.array(far)) + angle = np.arccos((b ** 2 + c ** 2 - a ** 2) / (2 * b * c)) * 180 / np.pi if (b * c) > 0 else 0 + + if depth > self.defect_depth_threshold and angle < 90: + valid_defects.append((depth, angle, far)) + defect_count += 1 + + # 质心特征 + M = cv.moments(cnt) + cx = int(M["m10"] / M["m00"]) if M["m00"] > 0 else x + w / 2 + cy = int(M["m01"] / M["m00"]) if M["m00"] > 0 else y + h / 2 + + return { + "area": area, + "aspect_ratio": aspect_ratio, + "solidity": solidity, + "hull_aspect": hull_aspect, + "defect_count": defect_count, + "valid_defects": valid_defects, + "cx": cx, "cy": cy, + "x": x, "y": y, "w": w, "h": h + } except Exception as e: - print(f"手指计数错误: {e}") - return 0 + print(f"轮廓分析错误: {e}") + return None + + def is_fist(self, features): + """稳定识别握拳（stop）""" + if not features: + return False + # 握拳核心特征：高密实度 + 低缺陷数 + 方正轮廓 + return (features["solidity"] > self.fist_solidity and + features["defect_count"] <= 1 and + abs(features["aspect_ratio"] - 1) < 0.3) + + def is_thumb_up(self, features): + """精准识别竖大拇指（up）""" + if not features: + return False + # 大拇指核心特征： + # 1. 窄高轮廓 2. 密实度在握拳和张开之间 3. 少量缺陷 4. 凸包特征匹配 + return (features["aspect_ratio"] < self.thumb_aspect_ratio and + self.thumb_solidity_range[0] < features["solidity"] < self.thumb_solidity_range[1] and + features["defect_count"] <= self.thumb_defect_max and + features["hull_aspect"] < 0.5) def capture_frames(self, cap): - """帧采集线程（稳定）""" + """帧采集线程""" while cap.isOpened(): ret, frame = cap.read() if not ret: break with self.queue_lock: - self.frame_queue = [frame] # 只保留最新帧 + self.frame_queue = [frame] time.sleep(self.frame_interval * 0.5) def process_frame(self, frame): - """优化后的帧处理逻辑""" - # 镜像翻转 + """核心处理逻辑""" frame = cv.flip(frame, 1) - # 绘制识别区域（仅边框） frame = self._draw_recognition_area(frame) - # 获取ROI roi, (roi_x, roi_y) = self._get_roi(frame) current_gesture = "None" - if roi.size > 0: # 确保ROI有效 - # 预处理：缩小+转HSV+肤色掩码 - roi_small = cv.resize(roi, (320, 240)) # 适度放大ROI + if roi.size > 0: + # 预处理（增强手部轮廓） + roi_small = cv.resize(roi, (400, 300)) hsv = cv.cvtColor(roi_small, cv.COLOR_BGR2HSV) mask = cv.inRange(hsv, self.skin_lower, self.skin_upper) + # 形态学操作：先开后闭，保留完整轮廓 + mask = cv.morphologyEx(mask, cv.MORPH_OPEN, self.kernel, iterations=1) + mask = cv.morphologyEx(mask, cv.MORPH_CLOSE, self.kernel, iterations=2) - # 形态学操作（去噪+填充） - mask = cv.morphologyEx(mask, cv.MORPH_OPEN, self.kernel) - mask = cv.morphologyEx(mask, cv.MORPH_CLOSE, self.kernel) - mask = cv.dilate(mask, self.kernel, iterations=2) - - # 查找轮廓 + # 找轮廓 contours, _ = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE) if contours: - # 取最大轮廓 cnt = max(contours, key=cv.contourArea) - area = cv.contourArea(cnt) + features = self.analyze_contour(cnt) - if area > self.min_contour_area: - # 计算密实度 - hull = cv.convexHull(cnt) - hull_area = cv.contourArea(hull) - solidity = area / hull_area if hull_area > 0 else 0 - - # 手指计数 - finger_count = self.count_fingers(cnt) - - # 可视化调试 + if features and features["area"] > self.min_contour_area: + # 绘制轮廓（调试用） cnt_scaled = cnt * (roi.shape[1] / roi_small.shape[1], roi.shape[0] / roi_small.shape[0]) cnt_scaled = cnt_scaled.astype(np.int32) cnt_scaled[:, :, 0] += roi_x cnt_scaled[:, :, 1] += roi_y cv.drawContours(frame, [cnt_scaled], -1, (255, 0, 0), 2) - # 手势判断逻辑 - if solidity > 0.8: # 握拳 + # ========== 重构手势判断逻辑（优先级+特征双重验证） ========== + # 1. 优先判断握拳（stop）- 双重验证 + if self.is_fist(features): current_gesture = "stop" - elif finger_count == 2: # 食指+中指 + # 2. 判断竖大拇指（up）- 专属特征 + elif self.is_thumb_up(features): + current_gesture = "up" + # 3. 判断两指（front）- 缺陷数精准匹配 + elif features["defect_count"] == 1: # 1个缺陷=2根手指 current_gesture = "front" - elif finger_count >= 4: # 手掌张开 + # 4. 判断手掌张开（back）- 多缺陷 + elif features["defect_count"] >= 3: # 3个缺陷=4根手指 current_gesture = "back" - # 其他情况（1/3指）归为None + # 5. 其他情况 + else: + current_gesture = "None" - # 手势缓存稳定 + # 增强缓存稳定性（3帧一致才更新） self.gesture_buffer.append(current_gesture) - if len(self.gesture_buffer) > 2: + if len(self.gesture_buffer) > self.buffer_size: self.gesture_buffer.pop(0) - if len(set(self.gesture_buffer)) == 1: + # 要求所有缓存帧一致才稳定 + if len(set(self.gesture_buffer)) == 1 and len(self.gesture_buffer) == self.buffer_size: self.stable_gesture = self.gesture_buffer[0] # 绘制UI @@ -184,15 +220,14 @@ def process_frame(self, frame): cv.putText(frame, f"FPS: {self.target_fps}", (10, 80), cv.FONT_HERSHEY_SIMPLEX, 1.0, (255, 0, 0), 2) - # 拉伸 frame_show = cv.resize(frame, (640, 480)) return frame_show def run(self): - """主运行逻辑""" + """主运行逻辑（修复时间计算错误）""" # 摄像头初始化 cap = cv.VideoCapture(0) - cap.set(cv.CAP_PROP_FRAME_WIDTH, 640) # 提高摄像头分辨率 + cap.set(cv.CAP_PROP_FRAME_WIDTH, 640) cap.set(cv.CAP_PROP_FRAME_HEIGHT, 480) cap.set(cv.CAP_PROP_FOURCC, cv.VideoWriter_fourcc(*'MJPG')) cap.set(cv.CAP_PROP_BUFFERSIZE, 1) @@ -203,31 +238,36 @@ def run(self): capture_thread.start() # 提示信息 - print("=" * 50) + print("=" * 60) print(f"✅ 帧率锁定 {self.target_fps} 帧 | ESC退出") - print("💡 调试提示：") - print(" 1. 把手放在右侧绿色边框的识别区域内（已扩大范围）") - print(" 2. 握拳 → stop | 食指+中指 → front | 手掌张开 → back") - print(" 3. 蓝色轮廓表示检测到的手部区域") - print("=" * 50) - - # 主循环 + print("💡 优化版手势识别（高稳定性）：") + print(" ✊ 握拳 → stop（高稳定）") + print(" 👍 竖大拇指 → up（精准识别）") + print(" 🤘 食指+中指 → front") + print(" 🖐️ 手掌张开 → back") + print("=" * 60) + + # 主循环（修复帧率控制） while cap.isOpened(): - # 帧率控制 + # 精准帧率控制（确保sleep时间非负） current_time = time.time() elapsed = current_time - self.last_frame_time - if elapsed < self.frame_interval: - time.sleep(self.frame_interval - elapsed) + sleep_time = self.frame_interval - elapsed + + # 关键修复：确保sleep时间非负 + if sleep_time > 0: + time.sleep(sleep_time) # 读取帧 with self.queue_lock: if not self.frame_queue: + self.last_frame_time = time.time() continue frame = self.frame_queue.pop(0) # 处理并显示 frame_show = self.process_frame(frame) - cv.imshow("Hand Gesture Recognition", frame_show) + cv.imshow("Hand Gesture Recognition (Optimized)", frame_show) # 更新时间戳 self.last_frame_time = time.time() @@ -242,6 +282,6 @@ def run(self): if __name__ == '__main__': - # 可降低帧率（如15）提高稳定性 + # 20帧兼顾流畅度和识别稳定性 recognizer = StableFPSHandRecognizer(target_fps=20) recognizer.run() \ No newline at end of file From 856383b86ecacd27486b6043073756f0ddd9e6dd Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Sun, 28 Dec 2025 01:54:17 +0800 Subject: [PATCH 43/45] =?UTF-8?q?=E5=A4=87=E6=B3=A8=EF=BC=9A=E4=BD=A0?= =?UTF-8?q?=E6=94=B9=E4=BA=86=E4=BB=80=E4=B9=88?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 60 +++++++++++++++++++++-------- 1 file changed, 45 insertions(+), 15 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index cd393590fb..645dc8cd82 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -13,18 +13,25 @@ def __init__(self, target_fps=30): self.frame_interval = 1.0 / target_fps self.last_frame_time = time.time() - # 2. 肤色检测（适配更多光线） - self.skin_lower = np.array([0, 10, 50], np.uint8) - self.skin_upper = np.array([30, 255, 255], np.uint8) + # 2. 肤色检测（适配明亮+暗光环境，核心优化：新增暗光阈值） + # 明亮环境阈值（保留原有，适配强光场景） + self.skin_lower_bright = np.array([0, 10, 50], np.uint8) + self.skin_upper_bright = np.array([30, 255, 255], np.uint8) + # 暗光环境阈值（降低S和V下限，放宽H范围，适配弱光场景） + self.skin_lower_dark = np.array([0, 5, 15], np.uint8) + self.skin_upper_dark = np.array([40, 180, 200], np.uint8) + # 默认使用暗光阈值（优先适配弱光，也可通过自适应逻辑切换） + self.skin_lower = self.skin_lower_dark + self.skin_upper = self.skin_upper_dark self.kernel = np.ones((5, 5), np.uint8) - # 3. 核心参数（精准适配手势特征） + # 3. 核心参数（精准适配手势特征，优化暗光下轮廓识别） # 握拳参数（稳定识别） self.fist_solidity = 0.82 # 降低握拳阈值，提高稳定性 self.fist_area_ratio = 0.75 # 握拳凸包面积比 # 手指计数参数 self.defect_depth_threshold = 8 # 降低深度阈值，提高up识别率 - self.min_contour_area = 600 # 降低最小面积，适配小手掌 + self.min_contour_area = 300 # 核心优化：从600降至300，适配暗光下小手部轮廓 # 大拇指识别参数（宽松但精准） self.thumb_aspect_ratio = 0.45 # 放宽宽高比 self.thumb_solidity_range = (0.55, 0.82) # 刚好卡在握拳阈值下 @@ -160,20 +167,42 @@ def capture_frames(self, cap): time.sleep(self.frame_interval * 0.5) def process_frame(self, frame): - """核心处理逻辑""" + """核心处理逻辑（暗光增强优化）""" frame = cv.flip(frame, 1) frame = self._draw_recognition_area(frame) roi, (roi_x, roi_y) = self._get_roi(frame) current_gesture = "None" if roi.size > 0: - # 预处理（增强手部轮廓） + # 预处理（暗光增强：亮度+对比度+去噪+形态学，核心优化） roi_small = cv.resize(roi, (400, 300)) - hsv = cv.cvtColor(roi_small, cv.COLOR_BGR2HSV) + + # 步骤1：亮度和对比度增强（解决暗光下图像偏暗、细节不清晰） + alpha = 1.8 # 对比度增益（>1提升对比度，极暗可调整至2.2） + beta = 40 # 亮度增益（>0提升亮度，极暗可调整至60） + roi_enhanced = cv.convertScaleAbs(roi_small, alpha=alpha, beta=beta) + + # 步骤2：高斯模糊去噪（去除暗光下的椒盐噪声，避免干扰轮廓提取） + roi_denoised = cv.GaussianBlur(roi_enhanced, (5, 5), 0) + + # 步骤3：（可选）自适应亮度判断，自动切换明暗阈值（兼顾所有环境） + gray_roi = cv.cvtColor(roi_small, cv.COLOR_BGR2GRAY) + avg_brightness = np.mean(gray_roi) + if avg_brightness < 50: # 亮度阈值，<50判定为暗光 + self.skin_lower = self.skin_lower_dark + self.skin_upper = self.skin_upper_dark + else: # >50判定为明亮环境 + self.skin_lower = self.skin_lower_bright + self.skin_upper = self.skin_upper_bright + + # 步骤4：转换HSV并提取肤色掩码（使用适配当前环境的阈值） + hsv = cv.cvtColor(roi_denoised, cv.COLOR_BGR2HSV) mask = cv.inRange(hsv, self.skin_lower, self.skin_upper) - # 形态学操作：先开后闭，保留完整轮廓 - mask = cv.morphologyEx(mask, cv.MORPH_OPEN, self.kernel, iterations=1) - mask = cv.morphologyEx(mask, cv.MORPH_CLOSE, self.kernel, iterations=2) + + # 步骤5：优化形态学操作（暗光下增加膨胀迭代，填补手部区域孔洞） + mask = cv.morphologyEx(mask, cv.MORPH_OPEN, self.kernel, iterations=1) # 开运算：去除小噪声 + mask = cv.morphologyEx(mask, cv.MORPH_DILATE, self.kernel, iterations=2) # 膨胀：填补手部孔洞，增强轮廓连续性 + mask = cv.morphologyEx(mask, cv.MORPH_CLOSE, self.kernel, iterations=2) # 闭运算：平滑轮廓边缘，去除残留小空洞 # 找轮廓 contours, _ = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE) @@ -182,7 +211,7 @@ def process_frame(self, frame): features = self.analyze_contour(cnt) if features and features["area"] > self.min_contour_area: - # 绘制轮廓（调试用） + # 绘制轮廓（调试用，可直观看到手部提取效果） cnt_scaled = cnt * (roi.shape[1] / roi_small.shape[1], roi.shape[0] / roi_small.shape[0]) cnt_scaled = cnt_scaled.astype(np.int32) cnt_scaled[:, :, 0] += roi_x @@ -240,11 +269,12 @@ def run(self): # 提示信息 print("=" * 60) print(f"✅ 帧率锁定 {self.target_fps} 帧 | ESC退出") - print("💡 优化版手势识别（高稳定性）：") + print("💡 暗光优化版手势识别（高稳定性）：") print(" ✊ 握拳 → stop（高稳定）") print(" 👍 竖大拇指 → up（精准识别）") print(" 🤘 食指+中指 → front") print(" 🖐️ 手掌张开 → back") + print("📌 已适配暗光环境，极暗可调整alpha/beta参数") print("=" * 60) # 主循环（修复帧率控制） @@ -267,7 +297,7 @@ def run(self): # 处理并显示 frame_show = self.process_frame(frame) - cv.imshow("Hand Gesture Recognition (Optimized)", frame_show) + cv.imshow("Hand Gesture Recognition (Dark Mode Optimized)", frame_show) # 更新时间戳 self.last_frame_time = time.time() @@ -282,6 +312,6 @@ def run(self): if __name__ == '__main__': - # 20帧兼顾流畅度和识别稳定性 + # 20帧兼顾流畅度和识别稳定性，暗光下更稳定 recognizer = StableFPSHandRecognizer(target_fps=20) recognizer.run() \ No newline at end of file From 2347d159178b50a2a8b214750056363112fa7465 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Sun, 28 Dec 2025 01:55:43 +0800 Subject: [PATCH 44/45] =?UTF-8?q?=E5=9C=A8=E5=85=89=E7=BA=BF=E6=9A=97?= =?UTF-8?q?=E6=97=B6=E4=B9=9F=E8=83=BD=E8=AF=86=E5=88=AB?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 1 - 1 file changed, 1 deletion(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index 645dc8cd82..1037efeeab 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -312,6 +312,5 @@ def run(self): if __name__ == '__main__': - # 20帧兼顾流畅度和识别稳定性，暗光下更稳定 recognizer = StableFPSHandRecognizer(target_fps=20) recognizer.run() \ No newline at end of file From c68229a74b26493d2865ead45c8365c998829eb8 Mon Sep 17 00:00:00 2001 From: jiangyi2302 <2592419327@qq.com> Date: Sun, 28 Dec 2025 18:48:41 +0800 Subject: [PATCH 45/45] =?UTF-8?q?=E8=A7=A3=E5=86=B3=E9=83=A8=E5=88=86?= =?UTF-8?q?=E6=89=8B=E5=8A=BF=E8=AF=86=E5=88=AB=E4=B8=8D=E5=87=BA?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/Drone_gesture_operation/main.py | 128 +++++++++++++--------------- 1 file changed, 61 insertions(+), 67 deletions(-) diff --git a/src/Drone_gesture_operation/main.py b/src/Drone_gesture_operation/main.py index 1037efeeab..9af98bf573 100644 --- a/src/Drone_gesture_operation/main.py +++ b/src/Drone_gesture_operation/main.py @@ -13,33 +13,30 @@ def __init__(self, target_fps=30): self.frame_interval = 1.0 / target_fps self.last_frame_time = time.time() - # 2. 肤色检测（适配明亮+暗光环境，核心优化：新增暗光阈值） - # 明亮环境阈值（保留原有，适配强光场景） + # 2. 肤色检测（适配明亮+暗光环境） self.skin_lower_bright = np.array([0, 10, 50], np.uint8) self.skin_upper_bright = np.array([30, 255, 255], np.uint8) - # 暗光环境阈值（降低S和V下限，放宽H范围，适配弱光场景） self.skin_lower_dark = np.array([0, 5, 15], np.uint8) self.skin_upper_dark = np.array([40, 180, 200], np.uint8) - # 默认使用暗光阈值（优先适配弱光，也可通过自适应逻辑切换） self.skin_lower = self.skin_lower_dark self.skin_upper = self.skin_upper_dark self.kernel = np.ones((5, 5), np.uint8) - # 3. 核心参数（精准适配手势特征，优化暗光下轮廓识别） - # 握拳参数（稳定识别） - self.fist_solidity = 0.82 # 降低握拳阈值，提高稳定性 - self.fist_area_ratio = 0.75 # 握拳凸包面积比 + # 3. 核心参数（细化两者特征差异，解决重叠问题） + # 握拳参数（收紧阈值，增加横向/方正特征约束） + self.fist_solidity = 0.85 # 从0.82升至0.85，收紧密实度，拉大与大拇指差距 + self.fist_area_ratio = 0.75 # 手指计数参数 - self.defect_depth_threshold = 8 # 降低深度阈值，提高up识别率 - self.min_contour_area = 300 # 核心优化：从600降至300，适配暗光下小手部轮廓 - # 大拇指识别参数（宽松但精准） - self.thumb_aspect_ratio = 0.45 # 放宽宽高比 - self.thumb_solidity_range = (0.55, 0.82) # 刚好卡在握拳阈值下 - self.thumb_defect_max = 2 # 允许2个缺陷（适配不同握法） - - # 4. 缓存参数（增加缓存提升稳定性） + self.defect_depth_threshold = 4 + self.min_contour_area = 300 + # 大拇指识别参数（强化纵向特征，与握拳形成明显差异） + self.thumb_aspect_ratio = 0.6 + self.thumb_solidity_range = (0.4, 0.82) # 上限设为0.82，与握拳阈值0.85无重叠 + self.thumb_defect_max = 3 + + # 4. 缓存参数 self.gesture_buffer = [] - self.buffer_size = 3 # 增加缓存到3帧，提升stop稳定性 + self.buffer_size = 3 self.stable_gesture = "None" self.frame_queue = [] self.queue_lock = threading.Lock() @@ -113,7 +110,7 @@ def analyze_contour(self, cnt): c = np.linalg.norm(np.array(end) - np.array(far)) angle = np.arccos((b ** 2 + c ** 2 - a ** 2) / (2 * b * c)) * 180 / np.pi if (b * c) > 0 else 0 - if depth > self.defect_depth_threshold and angle < 90: + if depth > self.defect_depth_threshold and angle < 100: valid_defects.append((depth, angle, far)) defect_count += 1 @@ -137,24 +134,25 @@ def analyze_contour(self, cnt): return None def is_fist(self, features): - """稳定识别握拳（stop）""" + """优化握拳（stop）判定：增加方正/横向轮廓排除，避免误判大拇指""" if not features: return False - # 握拳核心特征：高密实度 + 低缺陷数 + 方正轮廓 + # 握拳核心特征：高密实度 + 低缺陷数 + 方正/横向轮廓（h <= w，排除纵向大拇指） return (features["solidity"] > self.fist_solidity and features["defect_count"] <= 1 and - abs(features["aspect_ratio"] - 1) < 0.3) + abs(features["aspect_ratio"] - 1) < 0.3 and + features["h"] <= features["w"]) # 新增：握拳高度不大于宽度，排除纵向大拇指 def is_thumb_up(self, features): - """精准识别竖大拇指（up）""" + """强化竖大拇指（up）纵向特征，与握拳形成明显差异""" if not features: return False - # 大拇指核心特征： - # 1. 窄高轮廓 2. 密实度在握拳和张开之间 3. 少量缺陷 4. 凸包特征匹配 + # 大拇指核心特征：窄高轮廓 + 适中密实度 + 少量缺陷 + 强纵向延伸 return (features["aspect_ratio"] < self.thumb_aspect_ratio and self.thumb_solidity_range[0] < features["solidity"] < self.thumb_solidity_range[1] and features["defect_count"] <= self.thumb_defect_max and - features["hull_aspect"] < 0.5) + features["hull_aspect"] < 0.7 and + features["h"] > features["w"] * 1.2) # 强化纵向：高度大于宽度1.2倍，与握拳形成差距 def capture_frames(self, cap): """帧采集线程""" @@ -167,42 +165,36 @@ def capture_frames(self, cap): time.sleep(self.frame_interval * 0.5) def process_frame(self, frame): - """核心处理逻辑（暗光增强优化）""" + """核心处理逻辑：调整手势判断优先级，先up后stop""" frame = cv.flip(frame, 1) frame = self._draw_recognition_area(frame) roi, (roi_x, roi_y) = self._get_roi(frame) current_gesture = "None" if roi.size > 0: - # 预处理（暗光增强：亮度+对比度+去噪+形态学，核心优化） + # 预处理（暗光增强+去噪） roi_small = cv.resize(roi, (400, 300)) - - # 步骤1：亮度和对比度增强（解决暗光下图像偏暗、细节不清晰） - alpha = 1.8 # 对比度增益（>1提升对比度，极暗可调整至2.2） - beta = 40 # 亮度增益（>0提升亮度，极暗可调整至60） + alpha = 1.8 + beta = 40 roi_enhanced = cv.convertScaleAbs(roi_small, alpha=alpha, beta=beta) - - # 步骤2：高斯模糊去噪（去除暗光下的椒盐噪声，避免干扰轮廓提取） roi_denoised = cv.GaussianBlur(roi_enhanced, (5, 5), 0) - # 步骤3：（可选）自适应亮度判断，自动切换明暗阈值（兼顾所有环境） + # 自适应亮度判断 gray_roi = cv.cvtColor(roi_small, cv.COLOR_BGR2GRAY) avg_brightness = np.mean(gray_roi) - if avg_brightness < 50: # 亮度阈值，<50判定为暗光 + if avg_brightness < 50: self.skin_lower = self.skin_lower_dark self.skin_upper = self.skin_upper_dark - else: # >50判定为明亮环境 + else: self.skin_lower = self.skin_lower_bright self.skin_upper = self.skin_upper_bright - # 步骤4：转换HSV并提取肤色掩码（使用适配当前环境的阈值） + # 肤色掩码提取+形态学优化 hsv = cv.cvtColor(roi_denoised, cv.COLOR_BGR2HSV) mask = cv.inRange(hsv, self.skin_lower, self.skin_upper) - - # 步骤5：优化形态学操作（暗光下增加膨胀迭代，填补手部区域孔洞） - mask = cv.morphologyEx(mask, cv.MORPH_OPEN, self.kernel, iterations=1) # 开运算：去除小噪声 - mask = cv.morphologyEx(mask, cv.MORPH_DILATE, self.kernel, iterations=2) # 膨胀：填补手部孔洞，增强轮廓连续性 - mask = cv.morphologyEx(mask, cv.MORPH_CLOSE, self.kernel, iterations=2) # 闭运算：平滑轮廓边缘，去除残留小空洞 + mask = cv.morphologyEx(mask, cv.MORPH_OPEN, self.kernel, iterations=1) + mask = cv.morphologyEx(mask, cv.MORPH_DILATE, self.kernel, iterations=2) + mask = cv.morphologyEx(mask, cv.MORPH_CLOSE, self.kernel, iterations=2) # 找轮廓 contours, _ = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE) @@ -211,35 +203,39 @@ def process_frame(self, frame): features = self.analyze_contour(cnt) if features and features["area"] > self.min_contour_area: - # 绘制轮廓（调试用，可直观看到手部提取效果） - cnt_scaled = cnt * (roi.shape[1] / roi_small.shape[1], roi.shape[0] / roi_small.shape[0]) - cnt_scaled = cnt_scaled.astype(np.int32) + # 手动轮廓坐标缩放+偏移 + scale_x = roi.shape[1] / float(roi_small.shape[1]) + scale_y = roi.shape[0] / float(roi_small.shape[0]) + cnt_scaled = cnt.astype(np.float64) + cnt_scaled[:, :, 0] *= scale_x + cnt_scaled[:, :, 1] *= scale_y cnt_scaled[:, :, 0] += roi_x cnt_scaled[:, :, 1] += roi_y + cnt_scaled = cnt_scaled.astype(np.int32) + cnt_scaled[:, :, 0] = np.clip(cnt_scaled[:, :, 0], 0, frame.shape[1]-1) + cnt_scaled[:, :, 1] = np.clip(cnt_scaled[:, :, 1], 0, frame.shape[0]-1) + cv.drawContours(frame, [cnt_scaled], -1, (255, 0, 0), 2) - # ========== 重构手势判断逻辑（优先级+特征双重验证） ========== - # 1. 优先判断握拳（stop）- 双重验证 - if self.is_fist(features): - current_gesture = "stop" - # 2. 判断竖大拇指（up）- 专属特征 - elif self.is_thumb_up(features): + # ========== 核心优化：调整手势判断优先级（先up后stop） ========== + # 1. 优先判断竖大拇指（up）- 避免被stop提前拦截 + if self.is_thumb_up(features): current_gesture = "up" - # 3. 判断两指（front）- 缺陷数精准匹配 - elif features["defect_count"] == 1: # 1个缺陷=2根手指 + # 2. 再判断握拳（stop）- 此时已排除大拇指，无误判 + elif self.is_fist(features): + current_gesture = "stop" + # 3. 其他手势判断 + elif features["defect_count"] == 1: current_gesture = "front" - # 4. 判断手掌张开（back）- 多缺陷 - elif features["defect_count"] >= 3: # 3个缺陷=4根手指 + elif features["defect_count"] >= 3: current_gesture = "back" - # 5. 其他情况 else: current_gesture = "None" - # 增强缓存稳定性（3帧一致才更新） + # 缓存稳定性增强 self.gesture_buffer.append(current_gesture) if len(self.gesture_buffer) > self.buffer_size: self.gesture_buffer.pop(0) - # 要求所有缓存帧一致才稳定 if len(set(self.gesture_buffer)) == 1 and len(self.gesture_buffer) == self.buffer_size: self.stable_gesture = self.gesture_buffer[0] @@ -253,7 +249,7 @@ def process_frame(self, frame): return frame_show def run(self): - """主运行逻辑（修复时间计算错误）""" + """主运行逻辑""" # 摄像头初始化 cap = cv.VideoCapture(0) cap.set(cv.CAP_PROP_FRAME_WIDTH, 640) @@ -269,22 +265,20 @@ def run(self): # 提示信息 print("=" * 60) print(f"✅ 帧率锁定 {self.target_fps} 帧 | ESC退出") - print("💡 暗光优化版手势识别（高稳定性）：") - print(" ✊ 握拳 → stop（高稳定）") - print(" 👍 竖大拇指 → up（精准识别）") + print("💡 修复up误判为stop（优先级+特征优化）：") + print(" 👍 竖大拇指 → up（优先判断，精准识别）") + print(" ✊ 握拳 → stop（排除大拇指，无重叠）") print(" 🤘 食指+中指 → front") print(" 🖐️ 手掌张开 → back") - print("📌 已适配暗光环境，极暗可调整alpha/beta参数") + print("📌 已解决up与stop的特征重叠问题") print("=" * 60) - # 主循环（修复帧率控制） + # 主循环 while cap.isOpened(): - # 精准帧率控制（确保sleep时间非负） current_time = time.time() elapsed = current_time - self.last_frame_time sleep_time = self.frame_interval - elapsed - # 关键修复：确保sleep时间非负 if sleep_time > 0: time.sleep(sleep_time) @@ -297,7 +291,7 @@ def run(self): # 处理并显示 frame_show = self.process_frame(frame) - cv.imshow("Hand Gesture Recognition (Dark Mode Optimized)", frame_show) + cv.imshow("Hand Gesture Recognition (Fix UP→Stop Misjudgment)", frame_show) # 更新时间戳 self.last_frame_time = time.time()