Multiple Passive Sensors Track State Estimates And Associated Algorithms

Compared to the active sensor systems, like radar, passive sensor systems do not radiate electromagnetic signals, thus, they have the characteristics such as good concealment, strong survival and detection ability, long sensor distance, and so on. They have been widely used in the applications of sonar, infra-red, and navigation. Tracking the target by multiple passive sensors simultaneously will avoid the problem of designing the optimal trajectory using single sensor, and at the same time, improve the tracking accuracy of the system, so it has important theoretical and applied values.Multi passive sensor system actually is a nonlinear system. While the current research focuses on tracking the single maneuvering target based on the stationary observation stations, the multiple targets tracking problems are less studied. This thesis studies the single and multiple targets tracking problems and improves the algorithm from the aspects of nonlinear estimation of track states, data association of multiple targets, and non-stationary observation stations.First, the thesis describes the basic aspects and features of the different target tracking systems. Also, it introduces several common target motion models and proposes a measuring and location method in passive tracking systems. All of this lay the foundation for the following research of the thesis.Next, the thesis builds a model of multi passive sensors based on stationary observation stations as well as a model of the single target movement. To cope with the nonlinear filtering problems, the thesis proposes classical Extended Kalman Filtering(EKF) and modern Unscented Kalman Filtering(UKF) algorithms and separately applies them to passive tracking systems. The simulation results show that EKF and UKF both track the target well, and have similar accuracy level under the hypothesis of this thesis.Then, in order to solve the problem of data association of multiple targets tracking, this thesis proposes a method that applying Probability Nearest Neighbor Filtering(PNNF) to multi passive sensor systems. Based on the analysis and the comparison of several data association algorithms, it compares the tracking performance of PNNF and Joint Probability Data Association(JPDA). The results show that although PNNF is inferior to JPDA in the aspect of tracking accuracy, PNNF can realize multiple targets tracking when the targets are widely distributed, and its calculation burden is reduced by about 50% compared with JPDA.In the end, this thesis constucts a model of the translational motion and attitude motion of the non-stationary observation stations, and calculates the integrated observation errors. The following simulation experiments analyze the influence of the parameters-like sample interval and motion errors-on the system accuracy, which making the research have certain reference value for the relating engineering applications.