Research On Visual Tracking Of Moving Target Based On Kalman Filter

After the concept of moving target tracking presented in the last mid-century, thetechnology of moving target detection and tracking have developed rapidly and wasapplied into production extensively. From1990s, the research of target tracking basedon visual data has gradually increased. Owing to the real environments of trackingtarget are complicated and the model of being tracked moving target is uncertain, thevisual tracking technology of moving target has become a quite challenging researchaspect of computer vision. This paper, based on the theory of Kalman filter, givesseveral research methods of visual tracking technology of moving target, and presents anew method of moving target detection and shadow detection and elimination, andsome reasonable experimental results are obtained. The mainly tasks in this paperinclude:In the aspect of moving target detection, in order to solve the disadvantages ofthree-frame difference method, i.e. the images after detection by this method havenoises and partial region may be undetected, we utilize mathematical morphology toprocess the detection results, remove the noises in detection results and fill up thepartial undetected region, and present an edge linking method which is based ongradient to optimize the results of edge detection of moving target, and detect themoving target by combining three-frame difference method and the improved edgedetection method. This method can not only enhance the accuracy of target detection,but also preclude the occurrence of undetection of partial region of the moving target.In the aspect of moving target tracking, we apply the classical Kalman filteralgorithm to realize the detection of moving target. Because of the classical Kalmanfilter is linear filter algorithm and the detection effect of nonlinear problem is notsatisfactory, we apply adaptive unscented Kalman filter to detect single or multiplemoving target in sequence images. The experimental results demonstrate that the effectof detection of moving target by adaptive unscented Kalman filter is better.In the aspect of shadow detection and elimination, we analyze the reason ofoccurrence and features of shadow, and then present a new algorithm of shadow detection and elimination based on the features and textures of shadow in HSV colourspace. The experimental results show that this new algorithm can eliminate shadoweffectively, and enhance the accuracy of moving target detection and tracking.