Global Motion Pattern Based Event Recognition in Multi-person Videos.

Event analysis in multi-person videos is a great challenge-able task, especially in the sports videos due to the intensive and fast movement of players with serious occlusions. In this paper, we propose a global motion pattern (GMP) based event recognition algorithm. In order to avoid obstacles from complex background noise, GMP of video frames in sports video is extracted by optical flow. In particular, both spatial and temporal features of GMP are devoted to event recognition by sequential CNN and LSTM. Experimental analysis demonstrates that the proposed algorithm is capable to take benefit from spatial and temporal characteristics of GMP for effective event recognition.