Research On Mobile Target Tracking Based On Data Fusion In Wireless Multimedia Sensor Networks

With the technologies of multimedia sensors and embedded Processors development,the research of Wireless Multimedia Sensor Networks(WMSNs)has been a gradual increase.Because multimedia data can often provide a wealth of information about surroundings of a network,wireless multimedia sensor networks can enhance the traditional sensor network applications and enable several new applications.For examples,the applications for military intelligence,surveillance,smart home care and target tracking are researched.Intelligent tracking system with wireless multimedia sensor networks has become a hot topic today now.In this thesis,the relevant references on the basis of wireless multimedia sensor networks based on target tracking are read and studied,and a systematic research is made.The principle and basic methods of calibration,target detection and target tracking have been described and deeply researched.Calibration correction method based on least squares method is explained,and the target detection algorithm based on Gaussian mixture model is studied,and Kalman filtering method for tracking is also carefully studied.And we made the experiment in the semi-physical simulation platform built one by one,and get a good tracking result.According to the characteristics of wireless multimedia sensor networks is proposed particle filter algorithm based on fuzzy C-means clustering algorithm based on data fusion algorithm as the different trajectories obtained tracking data fusion.This algorithm first by fuzzy C-means clustering algorithm will track every time coordinate value of the target was divided into two categories,each of which generates a cluster center,then assigned a weight of two cluster centers based on particle filter algorithm,the final goal of more accurate coordinates of the current track.Experimental results show that the particle filter algorithm Fuzzy C-means clustering algorithm can effectively remove the extreme error,considering the information given in each node,the effective data integration.Through analysis,the algorithm has achieved the expected goals,and the experiment has a good results.Finally,the thesis apply the 3D detecting model of multimedia sensor nodes and the routing algorithm proposed for video streaming to semi-physical simulation platform of target tracking,to build a more comprehensive wireless multimedia sensor network system of target tracking based on data fusion,and completed the process from video capture to track the results of a series of working process,after several tests,tracking and data fusion experiments have achieved good results.