Occluded Moving Objects Extraction And Reconstruction In Video Monitoring System

Video monitoring system has been widely used in our daily life, and now its use value has been shown day by day. But there may be some moving objects occluded in the video, and we need to know the whole target, so the moving object must be extracted and reconstructed to recover the original object.This paper mainly studies the extraction and the reconstruction of the occluded moving object, but there are two difficulties in this paper. Firstly, in the case of the moving object in the video are all partly occluded, how to detect the moving object accurately; secondly, how to use the object’s parts to get a fine splicing result with a shorter time. To solve these two questions, an improved moving object detection method has been proposed in the detection section to improve the image quality. In the section of reconstruction, combines the SURF algorithm and the RANSAC algorithm to splice the image, its function is to speed up the method. This paper includes two parts, the first part is detection and extraction and the second part is splicing and reconstruction.In the moving object extraction’s section, using the moving object detection method improved by improving the update rate in the Gaussian mixture model to enhance the precision, and the motion history image was united with the frame difference method to overcome the frame difference method’s defect that the result prone to make holes. Finally, combines the improved background subtraction with the improved frame difference method to detect and extract the moving object to get a better effect. Experimental results show that the algorithm in this paper can make a higher quality image than using the normal ones.In the moving object splicing and reconstruction section, do preprocessing with the detected target to prepare for the splice process subsequently. In order to get a splice result with better quality and save the time, the SURF algorithm and the RANSAC algorithm are both used in the splicing experiment. Experimental results show that the moving object reconstruction effect is good with a shorter time.