Study On Detection And Tracking Algorithm Based On Sensor Network

With the concepts of"Internet of Things"and"Experience China"put forward, the vast potential for the sensor network is emphasized again in many domains. Detection and estimation technologies based on sensor networks are more satisfied with practical engineering requirements, but they also suffer many new problems and challenges.First, how to save energy is the primary problem in the detection and fusion algorithm of sensor network. Second, the sensor network's communication bandwidth is very limited, and generally it is only a hundred bits per second. Third, due to the living environment of sensor networks is usually very harsh, they will inevitably be disturbed by channel bandwidth and channel fading and other factors, so these radio channels are usually non-ideal.The multi-target tracking approaches based on multi-sensor and passive pure azimuth can achieve high imperceptibility and safety and they are important parts in modern tracking and defense system. They contain a lot of respects, but most of existing research is concerned with a single function or a combination of several functions. The research including all the functions about the pure azimuth multi-target multi-sensor targets tracking method is seldom, especially the study with the practical application performance and value is fewer.For these reasons, the thesis conducted the following research work:(1) In this thesis, we present a hybrid detection method for sensor network. It can greatly reduce energy consumption compared with centralized testing method and can improve the accuracy of detection results compared with distributed detection method.(2) By presenting the topology of any sensor network with a matrix, we propose the optimal detection method of this form when the channel is not reliable. We also give the comparison of detection performance when the channel parameters are different.(3) We give the solution to the pure azimuth multi-target tracking problems in multi-sensor platform. In our method, we use the algorithm of JPDA to solve the problem of data association, the UKF to solve the problem of nonlinear filtering, an intuitive method to solve the problem of track initiation and a sliding window method to detect changes of the targets'number .Our method overcomes the limitation for JPDA that it can only be used to the scenarios in which the number of the targets is invariable.