A Maritime Multitarget Tracking Method With Non-Gaussian Measurement Noises Based on Joint Probabilistic Data Association
Jian Chen, Jiacheng He, Gang Wang, Bei Peng · IEEE Transactions on Instrumentation and Measurement · 2025
Multitarget tracking (MTT) plays an important role in maritime missions. However, the typical non-Gaussian noise is introduced into ship sensor measurements due to the complex sea environment, and it degrades the performance of existing multiobject tracking algorithms based on Gaussian assumptions. To address the performance degradation and potential track divergence caused by non-Gaussian noise, a joint probability data association based on noise interaction Kalman filter (JPDA-IKF) is proposed. The novelty of this work lies in the decomposition of measurement noise and the subsequent state fusion based on noise components, which transforms a non-Gaussian linear system into multiple Gaussian linear systems, and it has been extended to nonlinear systems as well. Specifically, the non-Gaussian noise is decomposed into multiple noise components. Then, the JPDA-IKF is employed to handle multiple joint probability data and association (JPDA) filters with different decomposition results. Finally, an interaction method is applied to fuse the filter outputs. Furthermore, we extended the proposed approach to handle nonlinear cases, based on the cubature Kalman filter (CKF). Experimental results using both simulation data and real-world data demonstrate the effectiveness of the proposed approach.