Research On Moving Targets Collaborative Tracking In Wireless Sensor Networks

With the rapid development of embedded technology, communication technology and computer vision technology, wireless sensor network (WSN), with its advanced idea and the broad prospect of application is increasingly attracting the attention of scholars and related technologies have become one of the highlights in agro-scientific research in the world. Collaborative tracking of moving targets as one of the typical applications has always been the hot spot of the study, but most of the current study was to high-flying targets, and track of moving target in daily life such as the human body is rarely involved. Intelligent video surveillance system is on the computer vision and image processing technology, combined with other related technology and theory, developed a technology designed to use a computer or data analysis ability of intelligent processing unit, automatic video scene of static and dynamic physical awareness, description and analysis, intelligent security, intelligent transportation, and satisfy the social life wisdom city construction needs. So applying wireless sensor network in intelligent video monitoring system, can realize the wireless sensor network analysis of suspicious target and collaborative tracking, has certain practical significance and application value.Collaborative target tracking based on wireless sensor network involved in a wider range of knowledge according to the working process involves the following several important technical problems and key steps:optimized allocation of network resources, moving target detection and tracking, target behavior analysis, feature extraction and matching and tracking algorithm, etc. Although existing video image analysis technology can solve the above problem in some application scenarios, but for the wireless sensor network’s own limitations, such as energy consumption of wireless sensor node is sensitive and limited computing power, so this article in view of the above mentioned problems and key techniques of research, the concrete content is as follows:(1) Focus on the efficiency of wireless sensor network deployment optimization problem, puts forward a kind of based on functional division and mobile network optimization deployment of point to point method. Intelligent monitoring field synergy of target tracking are often a target for suspicious, so in the network initialization phase, first of all, to distinguish the function of wireless sensor nodes, will be divide into two categories: behavior recognition monitor and collaborative tracking monitor. Then in view of the system is part of the movable collaborative tracking monitor, setting behavior recognition monitor to initialize clustering center, a dynamic fuzzy clustering algorithm for network optimization deployment, for there is only a static system of point to point with the improved particle swarm optimization algorithm for network deployment. The function division of monitors and deployment optimization algorithm contribute of the solution, to give full play to the inherent advantages of wireless sensor network, for the determination of a suspected object feature extraction and lay the foundation.(2) For suspicious target selection problems and the limitations of wireless sensor nodes, is suitable for wireless sensor network of moving human behavior recognition method. Determined by suspicious behavior identify target positioning points a few steps:moving object detection, tracking and behavior recognition. For moving target detection in the practical application of scene and changeable interference and wireless sensor node operation ability is limited, is put forward using background subtraction division and the method of combining local generalized Hoff vote can be more effective to extract the moving object regions. Moving target tracking method based on detection technique was used to realize, through continuous motion target detection, achieve the goal of single node tracking. In view of the suspicious target decision problem, and puts forward the behavior template library, and through the moving target contour wavelet moment and speed, with the method of wavelet moment and behavior library template matching, if the behavior is sure to stay coordinated tracking target. Only to collaborative tracking the suspicious target, is better for the actual system application requirements.(3) Aim at the problem of the influence from different background between monitors for moving target feature extraction and feature extraction algorithm of complex practical problems cannot be applied to wireless sensor nodes, we put forward the multi-angle suspicious target feature extraction and matching algorithm of information fusion. First has repeated monitoring covering multiple will be super suspicious target pixel point to break up, for the target contour within the external rectangular super pixel area color feature expression, then multiple perspectives of pixel region color features for data fusion. In collaborative tracking monitor feature matching, similar characteristics of the moving targets are extracted, and then adopt two layers of matching feature matching method, according to the matching results determine whether the target is collaborative tracking target. The method can improve the different scenarios for suspicious target feature extraction, feature matching accuracy.(4) Focus on the energy consumption of wireless sensor network problem, the thesis proposes a geometry monitoring area based on the mechanism of sleep and wake up and track prediction algorithm, behavior recognition monitor is always in working status, and collaborative tracking monitor is default dormant, through to the suspicious target trajectory prediction, by behavior identify monitory point send command will involve tracking point to wake up together. In addition, in view of the problem of computing power, especially the multi-objective tracking pressure for behavior recognition monitor and the operation pressure of subnet management problems, this paper proposes a model based on the DOT of the parallel computing ideas, finally established the energy consumption model of collaborative tracking system, through the experiment of the prototype system and performance simulation, combined with similarity algorithm of data contrast, the research of collaborative tracking algorithm presented in this paper has certain advancement.