Method and apparatus for detecting abnormal behavior

The present invention provides a method and apparatus for detecting abnormal behavior. The method comprising: moving a plurality of targets to be detected behavior of real-time acquisition of the three-dimensional video to be detected is not determined that the type of behavior analysis, behavior characteristics acquired for each moving object to be detected, including normal and abnormal behavior type behavior, behavior characteristics comprising a three-dimensional space information and the time point information; construct to be detected behavioral characteristics corresponding to the model based on the behavior characteristic detected moving object to be; to be detected behavioral characteristics model in the database has been determined that the sample behavior behavior type of feature model determined by comparing the database model behavior feature to be detected maximum similarity model sample corresponding behavior characteristic behavior behavior type to be detected moving object type. Abnormal behavior detection apparatus of the present invention provides a method and avoids the analytical error due to occlusion between pedestrians caused severe behavioral characteristics of the moving object is large, the lower the accuracy of problem detection results.