Group Behavior Pattern Recognition Algorithm Based on Spatio-Temporal Graph Convolutional Networks

With the rapid growth of population, more diverse crowd activities, and development socialization process, group scenes are becoming common, so demand for modeling, analyzing, understanding behavior data in video is increasing. Compared with previous work on content analysis, factors such as increasing number people complex scene make analysis face great challenges. Therefore, a pattern recognition algorithm based spatio-temporal graph convolutional network proposed this paper, aiming at density video. A detection location method map regression-guided classification was designed. Finally, grade division designed to complete detection. In addition, paper also proposes extract features posture by using double-flow model, effectively capture differentiated movement information among different groups. Experimental results public datasets show that has high accuracy can predict behavior.