Investigation On Detection And Segmetation Of Moving Object In Video Sequences

The intelligent video surveillance system has found many applications, such as traffic, electric power, bank, public security, military and so on. The detection and segmentation of the moving object have attracted many attentions because it is the basis and key point in intelligent video surveillance system. Two typical algorithms including the symmetrical frame difference and background subtraction are investigated. Besides, we also analyzed the adaptive threshold, shadow cancellation and some key morphological operations. The main works are given as follows:On the analysis of the advantages and disadvantages of symmetrical frame difference and background subtraction, several improvements are made to the process of the detection and segmentation of moving object. Firstly, we extracted the relatively accurate motion region from the video sequence through marking points between any pairs on the contour the mask detected through symmetrical frame difference as moving points. Secondly, moving object is detected with the background subtraction which is carried out in the acquired region and the accurate background achieved via symmetrical frame difference and region growth. Thirdly, adaptive threshold is adopted in the process and we select the optimal threshold algorithm for the symmetrical frame difference and background subtraction separately via implementing and comparing different threshold algorithms.Based on the research about the algorithm of the detection and segmentation of moving object, a system is developed to detect and extract moving object from video sequences with C++in Visual Studio 2008, with the help of API of Open CV 1.0 supplied by the Intel and MS SQL Server 2000. And the result shows that the system is robust and effective.