Instructions/Guidelines

[markmorris]

Hi, discovered this looking for something open-source that does mocap. Reading through the code it's quite impressive how close this is. However, getting it to function without diving deep into code is what I'm struggling with. I've managed to get the software working along with getting the cameras connected, however, I'm unsure how to use the actual software after starting the calibration process. Would you mind posting a guide on how to use it the way you did on your video (https://www.youtube.com/watch?v=4ye_lzx6UWw)? I would be willing to put many hours into this to make it more of a complete open source mocap solution as well as develop a readme and promote to other developers to get on board with this.

I've developed a couple of mocap systems using multiple Xbox Kinects along with a blender plugin as the main software, but the results are not the greatest and cleanup is more trouble than its worth, but something like this would work really well.

[markmorris]

Some progression on usage, manage to get it to find a single marker. Although I'm still figuring out how this all works, it's quite hard when you have to read code to figure this all out.

[markmorris]

[Rob2048]

Hi @markmorris, I'm honestly quite impressed that you've got it running to this stage.

I'm going to try and remember exactly how the calibration process worked, but it has been a few years so I may be off.

Find extrinsic parameters (Relative position and orientation of cameras to each other)

1. You need to record a single reflective marker from all cameras. This marker should be moved around to cover as much screen space of each camera as possible. (It would be so cool to have the program tell you how much coverage you have achieved for each camera). There isn't any limit on length of these recordings. I think it is also lenient on the marker disappearing from view from some cameras some of the time. When you have the take you can use the 'Find extrinsic Parameters' button.

2. Bundle adjust: After you have the initial extrinsics you can then run the bundle adjust which will refine the positions even more.

Find world basis (Scale and orientation of the world)

1. Record a new take with a 3 markers arranged in a right angle on the floor. Each edge of the right angle should be 755mm long. (This can be adjusted in the code, I picked this length because I had steel rods this long that I connected the markers with).

2. Use 'Build 3D markers' to create the triangulated positions of the calibration markers.

3. Click 'Select World Basis' to enter selection mode. Select each marker in the scene view by clicking on it. Make sure the middle marker is selected second.

4. Use 'Assign World Basis' to use the selected markers as the floor plane, X/Y axis, and world scale.

5. Click 'Select World Basis' again to disable selections.

[markmorris]

Hi, Thanks for this, I will let you know all my findings.

[HighMasterGogo]

Great instructions, thanks Rob.

What's the process for individual camera calibration?

[Rob2048]

I used a program called camera-calib to find the intrinsic parameters for each camera. Then copied the calibration values into the config file that is stored on the Pi.

The values that need to be filled in the config are these ones:

'intrinsic': { 'camera': [0, 0, 0, 0, 0, 0, 0, 0, 0], 'distortion': [0, 0, 0, 0, 0] },

The 'camera' values are the 3x3 intrinsic matrix. (I honestly don't remember if it is packed row first or column first).
The 'distortion' values are as follows: k1, k2, p1, p2, k3

[HighMasterGogo]

Ah this makes perfect sense! Awesome. Thanks for confirming.

[dpredie]

@markmorris are you successful in this? im also looking for a more affordable solution for IR tracking, Rob's video demo is really great

[ninj99]

@markmorris, have you made any progress on this project?

[markmorris]

Although we made significant progress, we ultimately decided to go with an off-the-shelf solution, as we managed to get offered a system at half the cost. We may revisit this project in the future. Kudos to the initial developer for the excellent work done.