Research On Multi-Target Tracking For Passive Sensor Networks And System Design

In the field of Wireless Sensor Networks (WSN), target localization and tracking has been hot topics for years. For the advantages of low-power cost, stealthiness, high-precision measuring, strong anti-jamming, target tracking based on passive sensor networks is in great use. In this paper, multi-target tracking based on bearing-only measurements under passive sensor network is studied, especially focused on problems under complex measuring case with undetected and false alarm.Aim at solving problems on multi-target tracking under complex measuring case, innovative real-time target tracking frame and its workflow is designed. Involving modules in this frame are studied and applied in target tracking system based on acoustic array sensor networks successfully. The main work of the paper is as follows:1) For the problem that target initial state is unknown for target tracking under Bearing-only Sensor Network (BSN), confidence-based multi-target localization algorithm is proposed considering Contaminated Gaussian Measuring (CGM) model with undetected and false alarm, including association process and confidence-evaluation process. In association process, reference-based multi-target association algorithm is designed to deal with the difficulty in measurement-target matching and the tradeoff for association gate setting is studied by establishing optimizing model which decreases the computation greatly while the performance is retained. In confidence-evaluation process, innovative confidence-evaluation function is designed to distinguish target-origin locations from ghosts effectively. Simulations and experiments show the proposed algorithm can solve target initial state estimation problem successfully and provides necessary information for following modules.2) To reduce energy cost in BSN, robust node selection strategy is proposed under measuring model with undetected case. Robust node selection model is constructed according to the considering measuring model. Then greedy algorithm and pseudo-simplex algorithm for robust node selection is given. Also, robust node selection for multi-target case is discussed. Simulations and experiments investigate the tracking performance by taking advantages of the proposed robust node selection strategy and the results show that the proposed strategy can be more robust to the undetected case than the traditional strategies.3) To set up target tracking system based on acoustic array sensor networks, multi-level system structure is proposed under system requirement analysis, also real-time guarantee and low-cost strategy is designed considering the system characteristic. In the meanwhile, system experiments validate the proposed real-time target tracking frame and involving modules showing a fine system performance.