| In recent years, it is one of the most important research subjects that the detecting, recognizing and tracking of the interested part from persons' video images in the field of image processing and computer visualization. And it has wide applications, such as in the clinical field, it is very useful in the diagnosing and healing of the disabled people by analyzing the their kinetic information; and in the athletic sports, it is helpful in the prompting achievements also by analyzing the kinetic status of the athletes. The process of kinetic analysis includes the followings. Firstly, the capture of the images from the video camera; secondly, the proper tracking algorithms are adopted for the tracking of the persons' joints; thirdly, the three dimensional reconstruction of joints' coordinates; and eventually the kinetic analysis by the obtained parameters.The main content is how to confirm the information of joints fast and accurately under the guarantee of the matching precision in this paper. A better algorithm of Adaptive Size Search (ASS) depended on Four Step Search (FSS) is adopted by a new matching criterion relied on texture in the paper. Besides, the Mean Shift Algorithm (MSA) is mainly studied, and an improved algorithm based on MSA is proposed, which is called Dynamic Mean Shift Algorithm (DMSA), and it adapts for the great moving objects. Comparing the experimental results of MSA with DMSA, DMSA is better than MSA because of using the spatial information of objects. And the reliability and validity of tracking objects by DMSA is improved. In order to improve the exactness of tracking objects like human body joints, the new tracking algorithms based on several kinds of moments, which are embedded in dynamic mean shift, are proposed in the paper. The experimental results show that the tracking veracity and reliability are more improved and satisfactory results can be obtained when these algorithms are applied in the tracking of human joint points.
|
PDF Full Text Request