Abstract: Real time detection of croud abnormal behavior in public places is of great significance to maintain public safety and protect people's lives and property. Research shows that when the crowd's behavior changes according to certain rules, the crowd is normal behavior, otherwise it is abnormal behavior. For this reason, this paper mined the rules of crowd motion-changed in the video, established the abnormal behavior detection algorithm, and identified and located the abnormal behavior. The performance of the algorithm was evaluated on UMN data set and self-built data set, and compared with other algorithms. The experimental results show that the algorithm is effective in abnormal behavior detection.