Research On Moving Object And Shadow Detection Algorithms Based On Video Image

In the field of computer vision,moving object detection methods are the basis for object recognition,tracking,and other visual applications.It can provide real-time,rich and effective information for intelligent video surveillance systems in areas such as transportation to complete target recognition and tracking in video images.However,in practical applications,illumination changes,background changes,and moving shadows bring many challenges to the detection of moving target detection algorithms.Therefore,it is of great research value and practical significance to study effective moving target detection algorithms.Based on this,the research work in this paper can be summarized as follows:Firstly,three kinds of moving object detection algorithms are studied,including optical flow method,inter-frame difference method and background difference method,and the advantages and disadvantages of visual background extraction model(ViBe)algorithm are mostly analyzed.Secondly,for problems of the ViBe algorithm that are vulnerable to the effects of illumination and background changes and the phenomenon of “ghost”,a ViBe algorithm based on adaptive updating of pixel states is proposed;The algorithm adaptively adjusts the background model update rate by introducing a state function determined by the foreground and background switching frequencies of the pixel points.Experimental results show that the algorithm can better overcome the effects of illumination and background changes,and can quickly eliminate "ghost".Compared with the traditional ViBe algorithm and other typical moving object detection algorithms,the algorithm has better object detection effects and objective evaluation indicators,and it can meet the real-time requirements of monitoring.However,in the process of moving object detection,there is still a problem that moving shadows are misdetected as objects.Therefore,by analyzing the principle of shadow formation and on the basis of studying the moving shadow detection algorithm based on the shadow feature,and for the high rate of error detection and missed detection and other issues of the motion shadow detection by using a single feature,this paper proposes a moving shadow detection algorithm based on multi-feature fusion.The algorithm firstly fuses the results of shadow detection based on HSV color space,improved local binary pattern(CLBP)texture feature and Gabor filter texture feature,and then performs connectivity domain consistency correction and morphological operations to further modify the test results.Experimental results show that compared with other typical motion shadow detection algorithms,this algorithm can detect motion shadows more effectively and improve the accuracy of moving object detection.