A Study On Target Tracking In Wireless Sensor Networks Algorithm Based On Improved Particle Filtering

Wireless Sensor Network (WSN), which integrates the technologies of sensor, Micro-electro-mechanism system (MEMS), wireless communication and distributed computing, is a new mode of computing and a hot spot information technology which influences the lifestyle of human beings profoundly after Internet in 21st century.Based on the characteristics of low cost, low power, small size, high redundancy and multi-function, it is more convenient, more reliable and more effective for WSN to apply to target tracking than traditional system. The main purpose of WSN target tracking is to fix the location, the number, the velocity, the direction of targets and so on, its research aims to cost the low power consumption of WSN and obtain the high accuracy of target tracking. In the WSN target tracking, sensor nodes acquire the surrounding informations through mutual collaboration and send them to the central node for processing, while the problem of target tracking is an issue of nonlinear/non-Gaussian noise where Particle filter (PF) performs admirably. Although PF has a lot of advantages, its degrading will lead to the inaccuracy of target tracking.When the particles degrade, the target tracking algorithm of PF may reduce the accuracy of target tracking. Thus, this thesis aims to study the issue of WSN target tracking with improved PF. First, the thesis intends to summarize the current research situation of WSN target tracking and studies three typical methods of target tracking, focusing more on the comparison and analysis of target tracking research based on PF, including PF, unscented particle filter (UPF) and distributed particle filter (DPF), and then puts a target tracking algorithm of improved PF, namely the WSN target tracking algorithm based on cost reference particle filter. Next, to the thesis introduces the two improved particle filters, UPF, and cost reference particle filter (CRPF), after the description of the conventional research methods, the Bayesian theory, Monte Carlo method, the importance of sampling. And then, from the energy point of view, in order to solve the collaboration among nodes in WSN target tracking, the dynamic network model tree model is designed. In the network model, the cluster head (CH) may be dynamically generated in target tracking by some criterion, and receives data from other nodes. When the target leaves beyond the CH monitoring area, a new CH may be generated and the original CH restores to monitoring state.In this thesis, the target tracking algorithm of CRPF is designed and achieved by integrating CRPF and dynamic tree network model. After making the comparison of CRPF with PF, and UPF through the MAFLAB simulation, CRPF algorithm turns out to possess better tracking performance.