Tracking Multiple Moving Objects Based on the Kalman Filter

Object tracking is an important subject in computer vision. The tracking process includes two steps. The first step is background subtraction to detect the moving objects. The second step is using the filters, such as the Kalman filter or particle filter to track the detected objects. In this thesis, the Kalman filter and particle filter algorithms are simulated using a simple test video. An algorithm for tracking multiple moving birds in sky using the Kalman filter is presented. The objective of tracking is to analyse video sequences which is obtained from a fixed camera monitoring and track the birds in every frame and predict the birds’ location in the next frame. 15 videos are tested using the Kalman filter in this thesis. Finally, the tracking system should be implemented in GPU.