Crowd Behaviour Recognition Methods in Video Images

With the development of the information age and the acceleration of the process of urban intelligence, the analysis and recognition of video crowd behaviour has become a hot spot in computer vision research. Crowd behaviour recognition provides an effective technical means for video surveillance and video retrieval in public places. The aim of this paper is to investigate methods for crowd behaviour recognition in video images. The full paper begins by describing the current state of research and the research significance of this topic, arguing that the recognition of crowd behaviour is an essential component of intelligent monitoring and development. Secondly, an improved Vibe foreground detection model is proposed, and finally a behaviour recognition method based on anchor point selection with the KLT feature point extraction algorithm is proposed, showing that the algorithm can accurately estimate the number of people in the surveillance area.