Learning OpenCV 4 Computer Vision with Python 3 - Third Edition

Planning the flow of the application

The application will adhere to the following logic:

1. Capture frames from a video file.
2. Use the first 20 frames to populate the history of a background subtractor.
3. Based on background subtraction, use the 21st frame to identify moving foreground objects. We will treat these as pedestrians. For each pedestrian, assign an ID and an initial tracking window, and then calculate a histogram.
4. For each subsequent frame, track each pedestrian using a Kalman filter and MeanShift.

If this were a real-world application, you would probably store a record of each pedestrian's route through the scene so that a user could analyze it later. However, this type of record-keeping is beyond the remit of this example.

Additionally, ...