Multiple view soccer players tracking using area-of-coverage

Multiview provides correlated and redundant information which introduces advantages and obstacles over single view in application such as object tracking. In this paper, a novel system is proposed for parallel tracking of soccer players across multiple cameras using the concept of Area-of-Coverage. The system constructs and analyses the homography of each camera views(Area-of-Coverage) and based on their relationship selects the set of the camera for process which could greatly reduce redundancy and complexity on data fusion between individual views. The Area-of-Coverage also helps to increase accuracy in the detection process as well as provides information for automatic occlusion reasoning. Results and quantitative measurement are demonstrated on the TricTrac multiview soccer dataset.