Human Limb Motion Tracking Based On Blocking And Change Detection

During the past few years, human motion analysis has become a hot topic in the fields of image processing and computer vision. It has various applications in human animation, computer game, virtual reality and augmented reality, human-computer interaction, video surveillance, sports video analysis, computer-aided clinical diagnosis, and so on. The focus of human motion analysis research is detecting and tracking human in video images, capturing the human motion figures, reconstructing, and then describing and interpreting the human motion. Tracking human motion is a key and important step. Therefore tracking human motion in video images has large research value and application.In this thesis, we mainly do research on the technique of tracking human motion. The research is focus on tracking human limb (include trunk) from normal, non-mark and including human motion monocular video images, gaining 2D coordinates of joints. First, we make a summarization of the domestic and overseas status of the research in this field, contrast various methods and analyzing the difficult problems on tracking human motion. On the basis of this, we propose two new methods on tracking human limb motion based on blocking and change detection.The main research achievement is as follows:(1) A new human limb motion tracking method with blocking and region detection is brought forward. Firstly, detect the human moving region with the change detection between frames based on blocking and 2D human block model, and construct complete background region from the accumulated frame difference information by using background statistical technology. Then pick up the head and limb regions from the moving region, define the location information of the limb region with the analysis of region shape, and judge the location of the joints. The experimental results show that, the method can track large human limb motion, and it has low time cost and good usability.(2) A new human limb motion tracking method with model block sub-regions motion analysis is proposed. The basic idea of this method is, optical flow computation on motion region of object from change detection, dividing model block into some sub-regions on the method of optical flow clustering analysis, and motion analysis on them. On the basis of the motion of sub-regions, estimate the movement parameters of model block, calculate the motion information of model block on the algorithm of matching model block and change region, and calculating color model similarity, then obtain the position of the joints. This method improves accuracy of optical flow, has good usability.These two methods both initialize through manual intervention on the first frame of video images. The experimental results show that, the methods can exactly track human limb motion, solve the problem of self-occlusion in some extent, and it has low time and cost, high efficiency and good usability.