Football Player Detection and Tracking by Coupled Kalman Filter, Training Algorithm and Object Detection Methods---

This paper describes a camera-based observation system for football games that is used for the automatic analysis of football games and reasoning about multi-agent activity. The algorithm runs on video streams produced by cameras set up for TV broadcasting. Our approach consisted of read video files, convert video files to frame, player detection using coupled object detection methods and training algorithms and tracking with Kalman filter. The result showed that the accuracy of player detection and tracking of the movies with low speed was more than high speed. Furthermore when the region of filed was nearly constant the accuracy of detection and tracking increased, because of training algorithm effects on detection. When the players run with constant speed, the accuracy of detection and tracking was very high until the players run acceleratory. However Kalman filter coupled with training algorithm and object detection methods is efficient for detecting and tracking the football players very well.