The mining safety problem of coal bed methane is becoming more and more widely concerned. In order to solve the problem of video monitoring anomaly detection in the coal bed gas mining field, which contains the reciprocating motion of the pump, a new anomaly detection algorithm based on the regular variation background is proposed. In this method, the segmentation algorithm based on the three-frame-difference method is used to divide the image into the static background and the dynamic background of the reciprocating movement of the pump. At the same time, in order to set up the background model, the Surendra algorithm based on the three frame difference is used to extract the static background area information. Then different abnormal detection algorithms are used in different background areas, which can be better to eliminate the normal pump reciprocating movement interference of anomaly detection in the scene. In the static background, the three frame difference method and the background subtraction algorithm are used to divide the foreground. For the dynamic background, the three frame differential method is used to divide the foreground. Experiments show that this algorithm can accurately detect the foreground in coal-bed methane scenes, and meet the requirement of real-time video monitoring.