Research On Track Before Detect For Low Pulse Repetition Frequency Radars

Track before detect(TBD) is a power efficient signal processing technique which can be used in target detection and tracking. When the signal-to-noise ratio(SNR) is low, TBD can achieve superior performance over the traditional methods via accumulating the energy of consecutive scans of signal. Currently, TBD is a international hot area of research. Dynamic programming(DP) is an efficient implementation for TBD, which can avoid some degree of velocity mismatch and has linear complexity in the number of integrated frames. This thesis focuses on the utilization of DP based TBD in low pulse repetition frequency(PRF) radar. Some problems are considered in this thesis, e.g., the velocity ambiguity problem in low PRF radar, additional processing needed for the DP based TBD detections, and the estimation accuracy limited by the quantization size.Regarding the problems mentioned above, the main contents of the thesis are as follows:1. With reference to single target and multi-target scenario respectively, the basic theory of DP based TBD is studied. Since the low SNR targets are difficult to detect and the velocity ambiguity exists in low PRF radar, the single target DP based TBD algorithm for low PRF radar is studied.2. For multi-target scenario, the challenges of the utilization of DP based TBD in low PRF radar are analyzed. In the light of the challenges, a computationally efficient multi-target DP based TBD algorithm for low PRF radar is proposed. The proposed algorithm can solve the velocity ambiguity in the multi-target scenario, and make valid estimation of both the real states and the number of the targets in the scenario.3. The outputs of DP based TBD are not the filtered target trajectories. And, DP based is a grid based method whose estimation accuracy is limited by the quantization size. In order to solve the problems above, a recursive filtering algorithm for TBD plot-sequences is proposed under the assumption of linear Gaussian measurement model. The proposed algorithm can realize the whole target trajectory tracking and enhance the tracking accuracy.4. Since the DP based TBD detections are not subject to the linear Gaussian measurement model, a nonlinear non-Gaussian measurement model which is more appropriate to describe DP based TBD plot-sequences is proposed. Combination with the proposed model, a particle filtering algorithm using plot-sequences of DP based TBD for target tracking is proposed. Compared with the recursive filtering algorithm under the linear Gaussian assumption, the particle filtering algorithm has a higher estimation accuracy, and its estimation accuracy is less limited by the quantization size.Simulated experiments are used to demonstrate the validity of the proposed algorithms above.