Research On Maneuvering Target Tracking Algorithm On Passive Acoustic Network

Acoustic network detection is a passive sensor detection technology, which canrealize target locating and tracking using the detected acoustic signal of low-altitudetarget. It can assist the radar detection system for early warning. Some factors such asobservable, bearings-only, time-delay need be considered while the maneuveringtarget is tracked with acoustic network. Meanwhile, it’s very important to studyefficient algorithm for maneuvering target tracking, especially ground attacking, airlanding or avoid monitoring and so on. With the support of general armamentsdepartment weapons and equipment pre-research foundation project, a deep researchon maneuvering target detection with Acoustic network is studied in this dissertation.The main contents are as follows：First, based on the characteristics of time-delay and bearing-only for acousticnetwork low-altitude detection, the mechanism of time-delay and registration isanalyzed in this thesis. Tacking into acoustic, the influence of dense clutter fortracking accuracy, an improved minimum direction of arrival algorithm is proposed.Using a local search method, the proposed algorithm has performance of shortcalculation, real-time and high accuracy.Second, for the problem of low-altitude maneuvering target tracking, aninteracting multiple model algorithm based on unscented kalman filter (IMM-UKF)algorithm is proposed. The nonlinear of measurement equation is solved using theunscented transformation method, due to its simple structure and less computation.Meanwhile, the model weight is updated using the likelihood function, and statevalues are estimated by weighted model estimations. The simulation results show thatproposed method has a better tacking performance by adjusting the matchingprobability of models.Third, in order to further improve the tracking performance, a variable structuremultiple model (VSMM) algorithm based on digraph switch is proposed. Thetime-delay bearing data is associated with the network of multi acoustic sensors and adigraph switch method is designed to adjust the model set. The experimental resultsshow that the proposed method can match the motion of targets accurately, meanwhilecan reduce the calculation complexity and improve the tracking accuracy. Forth, all above proposed algorithms are tested with true data in a fusion platform,which is developed by our group. A pre-processing module is used to solve the clutterinterference, and a data fusion module to track the target. The experimental resultsshow that the proposed methods have real-time and reliable performance.