Research On Video Dynamic Target Detection Method

Video dynamic target detection, which aims to extract moving object in video sequences, is the basis of the field of computer vision. Fast and accurately extract the moving region, is a prerequisite task for the following research, such as target classification, video tracking, behavior recognition. In recent years, video dynamic target detection technology is widely used in intelligent surveillance, video retrieval, pattern recognition and other related fields; at same time, it is also an important tool in video coding, content of the interactive multimedia applications.Because of its strong temporal continuity of the video frame and spatial relationship with highly structured frame, based on the traditional target detection background difference method, time and space are fused in this video frame, three detection method is proposed:(1) detection method based on adaptive blocking and gray-value correlation;(2) detection method based on adaptive blocking and structure similarity;(3) detection method based on adaptive blocking and Hausdorff distance.The three methods are based on image adaptive segmentation, making full use of the relationship between intra pixel neighbor-hood, avoiding operating on single pixel, reducing the interference with single noise and a large amount of computation. Among them, first method refers to gray matching and gray correlation within adjacent frame blocks, calculate the correlation with corresponding blocks by an improved formula to distinct background; second method refers image quality assessment rules and structure similarity principle, calculates the similarity between corresponding blocks to judge background; third method, according to the basic principle of background difference method, refers image distance measurement and the Hausdorff distance, calculates the size between corresponding blocks to select background.Compared with traditional background subtraction method, the test results show that these methods have better detection effect and have a great increase in speed.