Target Tracking Algorithm Based On Information Fusion Research

The target tracking technology in video sequence images, which is a very significant branch in the computer vision, is becoming more widely used in the construction of national defense and economic spheres. The traditional target tracking technology which is based on visible light, would easily lost its target under the influence from illumination, obstruction or other factors. Hence, this article was based on the target tracking of visible light bringing up2viewpoints of target tracking. One is based on the fusion of infrared/visible light and the other is based on the fusion of video and sound information. The main work includes:1) In the research of the tracking algorithms of moving objects video,The Mean Shift Algorithm is used to track along the visible light moving and achieve the goal of basic target tracking;Regarding the problem that the size of the tracking window are not able to adjust along the size changes of the moving target under the Mean Shift Algorithm, the CamShift Algorithm is studied and achieve the goal to make the size of the tracking window adjust along the size changes of the moving target and enhance the Robustness of the target tracking in order to prevent the target from getting lost;2) Based on the study of image fusion method adequately, improving the tracking quality of sequence image as to better the quality of the tracking result, it brought up a fusion measure based on the sequence image of super-resolution restoration and the infrared/visible light images of moving objects testing. Experiments show that the target tracking after the fusion of sequence image is more stable than a single visible image or infrared image;3) Based on the circumstance that the fusion of infrared/visible light improve the tracking quality, this article brought up a fusion tracking method consisting of video and sound information based on Mac phone array. The fusion of sound information and video information adopts particle filter algorithm, in order to reduce the influence of particle degeneration, the measure of Mean Shift is utilized to converge the sample particles to the mean value. This theory shows that the fusion information tracking is more stable and more accurate than single information tracking.