The Algorithms Research Of Moving Human Detection And Tracking In Video Surveillance

In modern society, security problems are becoming increasingly subject to people's attention. Video surveillance is the most important technical means to protect the national, social and people's lives and property. It is also one of the popular problems in the computer vision and pattern recognition research fields recentely. Video surveillance consists of moving object detection, target classification, target tracking, target recognition, behavior analysis,and many other topics. In these moving targets, we mostly concerned about human body. This article from the video surveillance of the theory, put forward some of their predecessors on the human body detection and tracking algorithms were some of taxonomic studies, and on this basis, to enhance understanding of the algorithm, as well as suggest improvements.In this paper, the contents of the following points:1,In the aspect of human detection, we proposed a shadow remove method based on improved LBP feature in this paper. It would suffer light, shadows and other environmental impacts, especially the shadow in the human body detection. As the shadows and the human body's movements have similar motion characteristics, many moving target detection algorithm look the moving shadow as a part of the moving target or a new target. These will impact on the follow-up operations. Therefore, this paper introduce a new method the name is moving shadow removal base on improved LBP features. Based on the local binary patterns, an improved-new method for extract local texture is introduced, namely local ternary patterns. And this extraction method can be used to remove the shadow of movement. The background is modeled to use adaptive gaussian mixture models, which can get the background and foreground object. Thus, we can use the intensity property to obtain the probable-shadow blocks. The shadow detected was improved based on the similarity of texture represented by LTP between shadow region and corresponding region in the background. Finally, we have accuracy of the shadow regions and remove the shadow.The experimental results show that the algorithm is able to inhibit partition noise, and accurately remove the shadow of the movement, with good experimental results.2,In the aspect of human tracking, first, we introduce the Mean Shift tracking algorithm and its application in object tracking detailed. After the deep study of the Mean Shift algorithm, we find it would lose the tracking target if the human bodies move too fast. Otherwise, it would get the wrong tracking position in the complicated environment. According to these disadvantages of Mean Shift algorithm, we propose a target tracking method combines with Mean Shift and Kalman algorithm. In our method, we use the foregone information of the moving targets, that is the target moving information and space color information. We use the Kalman filte for the moving human's probable location's estimation. This location is a new starting point. And it use the Mean Shift algorithm to search the best match position of the moving human body. The combined algorithm can accurately track human targets, and it have very good handling in the external disturbances. The experimental result is satisfactory.These are the main contents of this paper. The experimental is based on OpenCv. And the results is good, it is also proved the methods correctly in this paper.