How to track vehicle accuratly using opencv python?

  kalman-filter, object-detection, opencv, python-3.x, yolo

I am working on a project where I have to detect vehicle from cctv and check whether the vehicle violate any rules. There are several use cases- red-light violation, U-turn violation, without-helmet violation etc…

  • For vehicle detection I have used YOLOv3 algorithm. The challange I am facing, during use_cases that is for every use case I have to track each vehicle.

  • I have used deepsort (Kalman-Filter) and other tracking algorithm.Among them deepsort is better but for non-linear transformation trac-id changes for same car, and also not accuratly tracking when more vehicle(more than 4) detected in video frame.

  • Used haar-cascade classifier for tracking …but result is not satisfactory.

Can anyone give me an idea of altarnate tracking process or idea (not GPS tracking)??

Used : python==3.8, opencv==4.2, tensorflow-gpu==2.4.1 etc.

Source: Python-3x Questions