파티클 필터 구조의 HPDA를 이용한 수동형 및 능동형 소나 시스템에서의 다수표적 자동추적 기법에 관한 연구

Abstract

In a naval environment, target signal can often be weaker than spurious clutter signal. In such cases, data association plays an important role in target tracking. The trend of modern target tracking is to integrate signal and data processing units and to accomplish processing in real time so that simple but accurate data association methods are required. Data association approaches including the probabilistic data association(PDA), joint PDA(JPDA), and multiple hypothesis tracking(MHT) consider data association events thus they can be complex and time consuming especially in a heavily cluttered multi-target environment. They are mainly confined to use with conventional filter structures such as the Kalman filter and the extended Kalman filter(EKF), with particle filters covered in a small number of publications. Standard measurement selection gating under the assump- tion of a Gaussian approximation is not consistent for nonlinear particle filtering. In our paper, the validation gate is extended to include entire surveillance region such that the probability that the target measurement falls in the validation gate is equal to 1. Soft gating with the expected likelihood is proposed in for particle filtering with PDA. Particle filtering for multi-target tracking is suggested with the Gibbs sampler that is used for association probability calculation. In order to make JPDA applications feasible for particle filtering, gating with Gaussian approximation of the predictive likelihood. The PHD filter and the cardinalized PHD(CPHD) filter with amplitude feature likelihood, are able to provide the generic multi-target nonlinear tracking scheme both for known and unknown target signal-to-noise ratio. In contrast to common technique for multi-target tracking like JPDA, the PHD filter requires no explicit association between tracks and target measurements. While many publications use particle filtering for nonlinear target tracking, not many consider real-time automatic target detection and multi-target tracking with particle filtering. I propose a new method of data association called highest probability data association(HPDA) combined with particle filtering and applied to real-time recursive nonlinear tracking in heavy clutter. The proposed HPDA method is a unification of probabilistic nearest neighbor(PNN) and probabilistic strongest neighbor(PSN) approaches. Unlike other approaches, the HPDA algorithm uses the amplitude information for only ordering the measurements in intensity. All current scan measurements are ordered by signal strength and used in calculating the association probability simply considering the rank order. The measurement with the highest measurement -to-track probability is selected for track update, and used for probabilistic weight update of particle filtering. The HPDA algorithm can be used in automatic target detection for track confirmation and estimation of the number of the targets and can be easily extended to multi-target tracking with nonlinear particle filtering in which the PNN and PSN have theoretical problems. This results in a computation- ally efficient particle filter solution. It can be used in automatic target detection to confirm tracks similar to the IPDA(Integrated PDA). It can also be used to avoid track coalescence phenomenon that prevails when several tracks move very close together as is often seen in target tracking with the JIPDA(Joint IPDA) filter. The performance of the proposed algorithm is evaluated by a series of Monte Carlo simulation runs for multi-target tracking in clutter. Simulation studies demonstrate that the HPDA algorithm is robust in maintaining tracks and performing automatic initiation of tracks in a hostile environment with various scenarios.