Abnormal behavior detection has a wide application prospect in the field of self-service banking intelligent monitoring system. In this paper, an anomaly detection method based on regional optical flow feature is proposed. Firstly, the mixed Gaussian model is used to represent the change of the background pixels and the background model is updated. The motion foreground is extracted from the video sequence with the background difference method. The optical flow information in the moving region is calculated with the lucas-kanade optical flow method. The weight-oriented histogram is used to describe the behavior, and the motion anomaly region of the histogram is calculated by using the motion entropy of the histogram. Then the SVM is used to classify the anomaly regions. From the experimental results, it can be seen that the abnormal events can be identified better and the real-time performance is better, which can meet the practical application requirements.