A Research On Motion Analysis Of Ping-pang Ball Based On Digital Video Image Processing

Digital video is a serial of images constituted by frames, which contains rich information. With the improvements of the capabilities of computing and the technology of digital image processing, it has been the focus of researching to abstract the images from every frame and acquire useful information after analyzing and processing. So the technology to achieve this goal is called visual tracking. It is a new researching area that introduces the technology of visual tracking to physical training to enhance the training efficiency and science. The machine vision has better accuracy and memory than human eyes, so it can catch fast moving objects and record the moving statistics which provide a more concrete form to judge the training effects. In the area of contemporary sports, we can track the moving objects and analyze the conditions and features of the video by the technology of visual tracking, so as to assist the sports training.A good moving objects tracking and detecting algorithm plays a very important role in the whole visual tracking system. It needs appropriate methods of detecting and tracking objects in ground of different applications. This issue presents a new method of tracking the moving ping-pang ball which mainly studies a algorithm of object detecting and tracking according to the video of table tennis matches, finally it applies the digital image processing to the studies and analyses of ball sporting.First, the algorithms decomposed every frame from the AVI video, and analyze every frame. Then it analyzes the center of the ball by the method of centroid, and gets the 3D coordinates by the algorithm of binocular matching to track the ping-pang ball. During the process it firstly reduces the impacts of noise by filtering, and then it eliminates the background disturbance of the match. During the process of image abstracting and partition, the region-growing is used to avoid plenty of computing and increase the accuracy of recognition. It proposed a algorithm of artificial mutual selection at the first frame and automatic searching at the rest frames during the process of the selection of seed. The principle of region-growing is also determined on the concern of every kind of features of color images.The contents of the issue are as follows:(1) It concludes and analyzes the theory and method of image partition and object tracking based on binocular vision.(2) It presents a method of tracking ping-pang ball by binocularity, based on video sequence.(3) It proposes a method of artificial mutual selection at the first frame and automatic searching at the rest frames, during the process of object abstracting by region-growing algorithm.(4) It uses the centroid of ping-pang ball as the matching point by binocularity, making the binocular tracking work out of marking, so it is useful in sports video tracking.This paper concludes the contemporary detecting and tracking methods. Moreover, it proposes a concrete tracking algorithm according to the special application in the video tracking of table tennis match, which introduces HSI color space. The algorithm is realized by C++ Builder, running on the platform of windows XP. The results of the experiments prove that the algorithm this issue proposes has a good effect.