Moving Object Detection and Tracking under Different Occlusion Conditions in Video Sequences

Abstract

This paper proposed a novel method for the detection and tracking of hand-thrown object in a video sequence in real time sports events. This paper mainly aimed to detect regions of the object in a set of outdoor and indoor videos in different occlusion conditions, and used Kalman filter to detect object on different trajectories over a fixed time window. This approach proved that the thrown object is successfully detected in various cases under occlusion and non-occlusion conditions with different backgrounds. To evaluate the accuracy, two different types of performance evaluations metrics are used based on object detection and tracking. The results shows that significant performance, the average accuracy of the object detecting is 97.89% and tracking is 98.35%.