A Reliability Point and Kalman Filter - based Vehicle Tracking Technique

This paper introduces a technique for tracking the movement of vehicles in consecutive video frames. The technique is based on a Kalman filter and a reliability point system. The Kalman filter predicts the most probable location of a detected vehicle in the subsequent video frame. This information is used by the tracking function to narrow down the search area for re-detecting a vehicle. The Kalman filter also helps to smooth out the irregularities due to the measurement error. The reliability point system provides a simple and fast mechanism to monitor the quality of tracking for the vehicles in the tracking list. Each vehicle is assigned with a reliability point, which can be increased or reduced at every tracking cycle depending on how consistent the vehicle is being re-detected. Experiment on some pre-recorded videos showed that the proposed technique can successfully track the preceding and overtaking vehicles in consecutive video frames. Keywords—Driver assistance system, intelligent vehicle, image processing, vehicle tracking.