The Key Technique Studies On Moving Human-Body Detection And Tracking Based On Indoor Environment

Robot vision spans many subjects including computer science, image processing, pattern recognition, artificial intelligence, etc. Object detection and object tracking is one of the most important technique in Robot vision. Object detection and tracking technology has already been applied in military, medicine, traffic and industrial fields, etc. It is currently a hot subject of research. Although the traditional video monitor has been used more widely, it lacks the property of intelligence. In this paper, object detection and tracking technology is used in video monitor, it make the system has the ability of access information, automatic analysis and processing, realize the intelligence of video surveillance system.As to human-body video object detection and tracking , there are still many problems no matter in theory research and in applications. In this paper, we major study the image segmentation, object detection and tracking technology. Image sequence of moving object can be divided into two types by which background is static or dynamic. According to the indoor environment, we major study the human-body object detection and tracking based on static background. At first, we make some research in image segmentation, Non-extensive is a property of Tsallis entropy, its segmentation performance is superior to thresholding methods using Shannon entropy. But traditional Tsallis entropy image thresholding segmentation method is based on point segmentation, the disadvantage is neglect the information of borders. In this paper , we present a two-dimensional Tsallis cross-entropy liner-type threshold segmentation method, then the Clustering-Based Niching Particle Swarm Optimization(CBNPSO) is used to search the best two-dimensional threshold vector, experimental results show that the proposed method can give better segmentation results than traditional Tsallis entropy thresholding image segmentation method. Frame difference method and background subtraction method were studied to join the merit, an improved method for object detection is proposed in this paper, join the method for shadow detection and noise detection, experimental results show that the proposed method can give better moving object detection results. In the aspect of object tracking, the methods for object tracking are summarized in the paper, the research on the algorithm of Mean-Shift is deeply discussed. The object oriented programming be used in system design, the human-body object detecting and tracking methods are tested in some realistic applications and results imply that those methods will be effective in the video surveillance system.