Occluded Moving Objects Extraction And Reconstruction In Video Monitoring System

There may be one case in video monitoring:We can't see the whole moving objects because of the occluder,but parts of the moving objects appear in the video images one after the other,and how to use these video images to get the whole moving objects? Namely,how to realize occluded moving objects extraction and reconstruction in video monitoring system? This is a very useful and significant problem.Our idea to solve this problem is as follows:Firstly,detecting and extracting the moving objects;Secondly,using image stitch algorithm to reconstruct the whole moving object. Therefore,we designed a video monitoring system,and then studied the two core problem of this idea. Firstly,how to detect and extract the moving objects,namely the problem about the moving objects detection and extraction. Secondly,how to use the uncovered parts of the moving objects images to reconstruct the whole moving objects,namely the problem about image stitching.To the problem of moving objects detection and extraction.Firstly, We used a moving objects detection algorithm combining frame-difference and background-difference to detect and segment moving objects. Secondly,we used the three-frame difference to extract the moving objects from the background,and then one new image restoration algorithm was proposed to restore the moving objects image,therefore this can make the moving objects extraction good visual effect.To the problem of moving objects stitching,one new image stitching algorithm based on corners matching is proposed. The new algorithm was achieved by the pyramidal corner filtering through the corner value constraint, the corner number constraint and the corner distance constraint. Moreover, consistency checking method was used to make sure the stitching accuracy. The heavy computation of template search, which often occurs in classical image stitching algorithms based on corners matching, was avoided by the proposed algorithm.At last,we introduced the hardware and the software of this video monitoring sysytem detailedly. The moving objects detection , moving objects extraction and images reconstruction were tested a lot. The results show that the system we proposed can detect and extract the moving objects accurately and effectively in camera-fixed scenes. Moreover, if the whole moving objects can't be seen because of occlusion, but parts of the moving objects appear in the video images one after the other, we can also extract the parts of the moving objects and then stitch these parts to reconstruct the whole moving objects image.