Study On Algorithms Of Recognition And Tracking Of Moving Objects

The thesis is focused on the study of algorithm for moving target recognition and tracking in video sequence images. Video sequence analysis is task-dependent. According to different appllications, the corresponding algorithm will be seclected. In order to solve how to detect and track moving objects in video sequence, the method based on image difference combined with analysis of image connectivity and the method based on Mean Shift segementation algorithm combined with the adaptive thresholding segementation are studied according to the characteristic of video images. The program of above mentioned methods are coded by using VC++6.0,analysis and comparison are made and experiment result are given. For the first method, pre-processing such as filtering,noise reduction and smothing must be made before differencing, then differencing method on two sequential frames or every two frames is implemented by the speed of object ,then images including moving information are filted by morphological filtering to eliminate background noise, lastly the object contour is abstracted from the filted image by means of the analysis of image connectivity.The Mean Shift algorthm for segmentation is mainly used in static images, and has the characteristic of fast and effective iteration. In this thesis, the mehod of segmentation based on the Mean Shift combined with the adaptive thresholding segmentation algorithm is applied to segement moving sequence images. The result indicates that the algorithm meets the requirement of real-time recogintion and its performance is much better than the method based on difference. Also study and analysis of the present general situation of target tracking are made and the trial study of the centroid tracking method based on the Mean Shift algorithm is made.Results show that the method can effectively track moving targets.