Research On Target Tracking In Wireless Sensor Network

Dynamic target tracking in WSNs is a hot research topic in recent years. It has very important academic value and can be used in civil and military area, such as air traffic control and battle field surveillance. The traditional filter algorithms for target tracking cannot be used in multiple target tracking and nonlinear non-Gaussian noise system. But particle filter can solve such problems.We first do thorough research work on single-object and multiple target tracking in WSNs. Based on the existing algorithms, this paper brings forward some new algorithms which consider the system energy consumption and tracking accuracy. In order to reduce the energy consumption, dynamic clustering has been used to balance the energy consumption of the system. On the basis of clustering, in order to use observed value in prediction, this paper puts forward an improved Kalman dynamic clustering particle filter algorithm(DDCPF) which can improve the tracking accuracy.Since of the tracking accuracy of EPF will decrease or even diverge greatly in nonlinear environment and UPF will appear particle degradation in the environment when the absolute differences between process noise covariance and estimated error is big. In order to solve the above problems, a new algorithm based on EPF and UPF named EUPF is proposed. In order to adapt to the system change, this EUPF algorithm alternately uses EPF or UPF to update particles to improve the system stability and tracking accuracy. Since kalman filtering is optimal in linear and Gaussian system, based on the EUPF algorithm, we put forward a dynamic clustering FCM marginal particle filter algorithm to adapt to the changing environment. According to the characteristics of the system, we choose kalman or EUPF to track the multiple targets.Finally, we set up the simulation platform to evaluate the performance of the proposed algorithms. Through simulation analysis, we find that the new algorithms can reduce energy consumption, improve the network lifetime and estimation accuracy.