Such methods are often used to segment moving objects in sequence images, and the mainstream methods are background subtraction method, frame difference method and optical flow method. The background subtraction method is to subtract the image frame from the reference background, and then perform foreground detection on the absolute difference image to determine the motion area. It requires the background of the scene to be constant or change slowly. The frame difference method is to make a difference between consecutive frame images, and then detect the moving area in the difference image to obtain the foreground target. The frame difference method responds to scene changes in a timely manner, but the segmented motion regions are usually very incomplete, and generally require more post-processing steps to repair them. The optical flow method assumes that the image change is completely caused by the movement of the target or the background, and realizes the detection of the moving target by calculating the two-dimensional instantaneous velocity field of the optical flow field. The optical flow method can obtain complete motion information, and has good robustness to scene changes and camera motion, but its main defect is poor anti-noise ability. Most optical flow calculation methods are very complex and difficult to run in real time. In practice, optical flow calculation methods with lower complexity are often used. Due to the poor quality of thermal infrared imaging, complex and changeable human body shapes, and the inherent defects of motion segmentation methods, motion segmentation based on the combination of spatiotemporal information can improve the segmentation quality at the cost of high computational complexity.