Learning group activity in soccer videos from local motion

摘要: This paper proposes a local motion-based approach for recognizing group activities in soccer videos. Given the SIFT keypoint matches on two successive frames, we propose simple but effective method to these keypoints into background point set and foreground set. The former one is used estimate camera motion latter applied represent actions. After compensation, apply descriptor characterize relative between corresponding consecutive frames. novel representing since it focuses of individuals excludes noise such as caused by inaccurate compensation. Experimental results show that our achieves high recognition rates videos robust compensation results.