Abstract: Human group activities analysis has become a new research interest of computer vision. It has broad application prospects and great economic value in many aspects, such as intelligent video surveillance, virtual reality, video retrieval and so on. This paper made a survey on the human group activities analysis from group databases and analysis algorithms. Firstly, the group databases mainly consist of behavior databases and surveillance databases. This paper summarized the typical databases of these two classes. Secondly, we sumed up the human group activities analysis algorithms from decomposition methods and recognition methods of group activities. `hhe decomposition methods were divided into three kinds and their strengths and weaknesses were pointed out, separately. The recognition algorithms of group activities were divided into statistical approaches and description-based ones. We compared them in detail and pointed out the development tendency. At last, we summarized the potential problems of human group activities analysis and looked forward to its future.

Key words: Human group activities analysis,Databases,Decomposition methods,Human group activities recognition