Skip to content

README.md

Latest commit

 

History

History
77 lines (51 loc) · 2.76 KB

README.md

File metadata and controls

77 lines (51 loc) · 2.76 KB

Detection and Tracking of Pantograph and Cable Contact Points with Spark Identification

Overview

This project implements a computer vision-based system to detect, track, and analyze pantographs in railway videos. Using a trained YOLOv8 model, the system identifies pantograph components, tracks their movement, detects sparks, and logs all relevant information for performance and safety analysis. The solution is optimized for real-world deployment in railway monitoring systems.

Features

  • YOLOv8 Object Detection: Detects pantograph, pantobar, and cables in each video frame.
  • Movement Tracking: Calculates horizontal and vertical displacement of the pantograph between frames.
  • Spark Detection: Identifies sparks in the pantobar region by analyzing high-intensity (bright) spots.
  • Contact Point Analysis: Computes distances between pantobar and cables to check for proper alignment.
  • CSV Logging: Records timestamped movement data, detected sparks, and contact distances.
  • Annotated Video Output: Saves the processed video with bounding boxes, annotations, and overlays.

Setup

Requirements

  • Python 3.8+

  • Required Libraries:

    pip install opencv-python-headless numpy ultralytics

YOLO Model

  • Ensure you have a trained YOLOv8 model for pantograph detection.
  • Update the model_path variable in the script with the path to your weights file.

File Paths

  • Input Video: Set input_video_path to the location of your raw video.
  • Output Video: Set output_video_path where the processed annotated video will be saved.
  • CSV File: Set csv_output_path to define where the CSV log should be stored.

Script Workflow

  1. Frame Capture: Reads each frame from the input video.
  2. Object Detection: Runs YOLOv8 inference to detect pantograph components.
  3. Movement Analysis: Compares pantograph position across consecutive frames.
  4. Spark Detection: Identifies bright pixel clusters within the pantobar region.
  5. Contact Analysis: Measures distances between pantobar and overhead cables.
  6. Logging: Writes structured data to a CSV file.
  7. Video Rendering: Saves annotated video with bounding boxes, spark highlights, and metrics overlay.

Usage

  1. Update paths for model, input video, output video, and CSV inside the script.

  2. Run the script:

    python pantograph_tracking.py

  3. Outputs:

    • Processed video with annotations.
    • CSV log file containing movement, spark count, and distances.

Notes

  • Ensure YOLO is properly trained for pantograph, pantobar, and cable classes.
  • Detection parameters may require tuning for different lighting or video conditions.
  • Output files will be saved in paths configured in the script.