Studies On Target Tracking In Passive Radar Based On Opportunity Illuminator

The passive radar technology that uses the signal from an opportunity transmitter is very important, as its concealment and anti-jamming performance, has received considerable attention from researchers at home and abroad. In the passive radar operation, one, two or all of the bearing, time-delay, and Doppler frequency parameters are exploited to locate and track the moving target. The target tracking algorithms of passive radar are focused in this dissertation. The main contributions of this dissertation concentrate on a few aspects as follows:1. The target observations before the target tracking are considered. Because high effective estimation and tracking of the carried frequency of illuminator signal in the target observation and parameter estimations is very important, an effective frequency estimator based on autocorrelation is presented. It can estimate the single frequency of a complex tone in white noise fast and effectively, especially with low SNR, and whose performance can reach the Cramér-Rao low bound at high SNR.2. Almost all classical Bayes based tracking algorithms in passive radar are introduced. The characteristics of those algorithms are analyzed respectively, and their application situations are given.To deal with the nonlinear and non-Gaussian noise situations of passive radar, particle filters (PFs) based on sequential Monte Carlo method and recursive Bayes estimation have been highly focused due to its excellent performance. Aiming at the famous problems of particles degeneration and sample impoverishment when using PFs, two improved algorithms are proposed as follows:To deal with the sample degradation problem of PFs, we propose a novel resampling algorithm framework. It incorporates the latest measurements into the resampling procedure to drive particles with high enough weight to the highly likelihood regions. Methods of local linearization (LL), e.g. the extended Kalman filter (EKF) and the unscented Kalman filter (UKF), are adopted to achieve it. The proposed algorithms show improved performance but less computational cost over such algorithms that incorporate latest measurements in the proposal importance distribution using the same LL methods through simulations of two different nonlinear systems.To deal with the sample impoverishment problem of PFs, a novel resampling algorithm is proposed, which introduces the sigma points, in the form of a small set of points generated using simple deterministic rule, into the generic resampling procedure to enrich the representation of the posterior distribution characterized by particles with weight high enough. Compared to other related resampling algorithms, such as the exquisite resampling (ER) algorithm reported recently, the proposed algorithm showed an improved estimation performance and reduced computational complexity in the simulation of two different nonlinear systems.3. The issues of multi-target tracking are studied. The probability hypothesis density (PHD) filter has emerged as a promising tool for dealing with the multi-target tracking problem in recent years. However, except in some special situations, closed-form recursive update equations for the PHD filter do not exist and the PF approaches have to be used. The output of the PF at each step is the particle clouds approximation of the PHD. Thus, some special algorithms are needed to extract the target states from those particles. Utilising the information of both particles'weight and their spatial distribution, an improved algorithm named C-Clean is proposed. Simulation results demonstrate that its performance is better than those algorithms using the information of particles'spatial distribution or weight only.4. Real-time implementations of PFs for target tracking are considered in this dissertation. Firstly, hardware implementation of PF applied to bearings-only tracking (BOT) problem is studied on FPGA platform. A simplified PF-LLR-EKF algorithm, which highly reduces the complexity for hardware implementation, is proposed. Experimental study indicates that this algorithm has better performance than the traditional PF implementation when the root mean square error (RMSE) is considered, but with approximately equal time cost. Secondly, parallel implementation of PF applied to the target tracking of passive radar is studied on Graphics Processing Units (GPU) platform. An interacting multiple model PF (IMMPF) algorithm is experimented. Experimental results indicate that the performance improvement on GPU platform over the serial implementation on PC is over 10 times.