Theory And Application Research On Bearing Only Target Tracking

Bearing only target tracking, or target motion analysis (TMA), is a classic problem of estimating the target motion state (such as position, velocity etc.) from noisy electromagnetic or acoustic signal bearing data received passively by the sensor. As an important part of the passive tracking system, it was originally applied to the underwater sonar submarine tracking, with the rapid development of sensors, microprocessors, wireless communication, GPS etc. and the improvment of the modern military demand, it is gradually applied to the airborne radar/ESM tracking, the ground-based acoustic sensor network low altitude surveillance and the multiple drifting sonobuoys early warnings.The problem is highly nonlinear, for the ralative range of the target and the sensor being not observable, it also a state partial observable problem. The initial researchs concern mainly on two aspects: the nonlinear filting and the maneuvering strategy of the sensor ownship. With the expanded application to aircraft tracking of TMA, the maneuvering target tracking, environment clutter and the acoustic signal retardation effect all complicate the estimation problem and the performance of some conventional TMA methods are not satisfying.In this thesis we study the following aspects of the TMA: maneuvering target tracking, acoustic emitter positioning with signal transmit time delay and a special multitarget tracking. First, an intelligent range parameterized Unscented filter (IRPUF) is proposed to estimate the nonmaneuvering target state, the presented method improve the initial perfprmance, the filtering precision and optimize the system resource compared with the traditional method. Second, a combination technique of the interacting multiple model method and the kernel particle filter (IMMKPF) is presented to deal with the maneuvering target tracking, the needed particle number is far less than the traditional particle filter and the estimating precision is comparable to it. Third, two single-platform based recursive-typed algorithms, the parameter online estimation (POE) and the perturbation range division (RD), are proposed to real-time estimate the acoustic signal emitter position, in which the POE was validated by a practical acoustic sensor network surveilance system. Last, a density estimation based robust fusion data association method is presented to passively track multitarget by an ownship with multiple sensors, the proposed approach avoid the combinatorial explosion by the conventional methods and improve the correct association probability.