Research On Weak Target Track Before Detect Based On Dynamic Programming

With the appearance of stealth targets and high speed targets, the detection and tracking of weak targets is receiving much attention. Track-Before-Detect(TBD) is an effective approach of weak target detection and tracking. TBD can achieve the detection and tracking of weak targets simultaneously when tracking low Signal-to-Noise-Ratio(SNR) targets by directly processing multiple consecutive measurement frames so as to integrate target energies along the possible target trajectories. Dynamic Programming (DP) algorithm is a problem of multi-stage decision, which is effective in detecting and tracking weak targets in all TBD methods, so it has been widely studied owing to their good performances all over the world. This dissertation is mainly concerned with the study of TBD based on DP. The author’s major contributions are outlined as follows:1. The traditional DP-TBD algorithm is introduced. First, the basic principle of DP is studied, and then the procedure of DP-TBD is presented with the example of tracking the point target with constant velocity in Gaussian noise background, followed by a presentation of a new concatenated target discrimination method of the superposition trajectories and the statistics of different level direction histogram, and concludes with an analysis of the algorithm performance.2. Aimed at the shortcomings of the traditional DP-TBD, in this paper three improved DP-TBD methods are proposed. In order to reduce the amount of the computation, two-step thresholds DP-TBD is developed and an additional process, namely merit function transfer, is introduced in the DP recursion to guarantee the full utilization of the existing target information via transferring the target-like merits through the physically admissible target trajectories. The result shows that the computational expense of DP-TBD can be significantly reduced at the cost of tolerable performance loss. In view of weak targets in complex environment, in this paper a complex likelihood DP-TBD approach is presented that includes the phase information of complex data and takes advantage of the given clutter map. The inclusion of phase and the use of clutter map are shown to improve the discrimination of targets from noise and clutter and, in comparison with the envelope likelihood, reduce the computation overhead. Based on the characteristic of radar signal, an new DP-TBD that makes full use of Doppler information is described. By virtue of Doppler information to reduce the search range it not only improves the computational efficiency and decreases the range of energy diffusion but also alleviates the effect of the strong noise and clutter points on the DP-TBD algorithm performance. A comparison of DP-TBD algorithm in optical images with the new method in this paper shows that especially in the complex environment it can achieve superior detection and tracking performance.3. The detection and tracking based on DP-TBD of multiple targets is introduced. To begin with, the theoretical model of multi-target DP-TBD is presented and then this paper deals with the difficulties in the multi-target DP-TBD. By way of the Generalized detection procedure and based on the successive-target-cancellation(STC-VTA) and single-pass STC-VTA (SP-STC-VTA) algorithm, a multi-target DP-TBD algorithm (MT-DP-TBD) is developed to take account into the structure of MT problem. The essence of this approach is how to change the detection and tracking of multiple targets into several single targets detection and tracking. Finally, the simulations show that in comparison with STC-VTA and SP-STC-VTA, MT-DP-TBD can allow to trade better detection and tracking performance with a lower computational burden, and the result from live data also indicates the validity of MT-DP-TBD.