Passive Multi-object Tracking And Location Technology

Passive detection has received extensive attention and research for its concealment advantage,where the detection result are generally expressed as the azimuth measurements with noise and outliers.Tracking and locating can remove the outliers,divide the measurements batches,and estimate the number and location of the objects.Underwater unmanned vehicle(UUV)has become a hot research area in underwater equipment for its advantages in strategic importance and low cost.Therefore,more attention has paid on the tracking and locating technology of passive detection with UUV.In this paper,the technique of multi-object tracking and location in passive detection is analyzed,which is divided into two parts: tracking and locating.According to the number of sensor,it can be divided into single sensor measurement and multi-sensor measurement.The multi-hypothesis tracking and Kalman filter are combined in the framework of particle filter,and the prior distribution,likelihood distribution and posteriori distribution of data association process are analyzed.Monte Carlo data association method that suitable for the condition with konwn number of targets is obtained.The variation of object azimuth and its applicable motion model are also studied.By introducing the mechanisms of trajectory initiation and termination in data association,a single sensor passive multi-object tracking method suitable for unknown number of objects is obtained.Then the tracking method is extended to the case where the number of measured values is zero or more in the same period to match the actual detection.According to the existing estimation,RTS smoother is used to smooth the tracking results to improve the tracking accuracy.After the passive tracking of single sensor,azimuth only positioning solution is needed to estimate object position.On the basis of analyzing the observability of single sensor,a progressive unbiased azimuth location algorithm for single sensor is studied,and the tracking and localization are combined.A passive multi-object tracking and locating algorithm based on single sensor measurement is presented.Multi-sensor tracking requires the fusion of object batches among different sensors,so that the method of tracking first and then locating is difficult to be used in multi-sensor case.One feasible method is to replace the standard Kalman filter in the tracking algorithm with the extended Kalman filter to realize the nonlinear measurement equation and extend the probability model of data association.Thus,a multi-sensor passive multi-object tracking algorithm is obtained to estimate and track the position of the object directly from the multi-sensor azimuth measurement.In order to evaluate the tracking results,a multi-object tracking evaluation criterion is introduced to realize the quantitative evaluation of the tracking and positioning results.