基于PN学习的雷达目标鲁棒跟踪算法

Robust tracking algorithm for radar target based on PN-learning

传统雷达目标跟踪算法在强杂波环境下跟踪时会产生大量虚警估计的情况,单独跟踪或检测算法都不能对干扰杂波进行有效滤除。针对这个问题,在传统概率假设密度滤波器(PHD)算法的基础上,提出一种联合检测—跟踪—学习的目标鲁棒跟踪算法,即PN—PHD,引入属性检测器,将检测跟踪结果一起送入PN学习器,通过PN学习迭代更新检测器,并修正PHD算法的跟踪估计,以此实现在强杂波环境目标鲁棒跟踪的要求。仿真实验结果表明:PN—PHD滤波算法与传统跟踪算法相比,在强杂波环境下有效地提高了目标跟踪准确性和跟踪精度,同时也弥补了PHD算法在提供目标航迹信息方面的不足。

Traditional radar target tracking algorithms usually cause more false alarms under complex environment. It is useless for single tracking or detecting algorithm to filter clutters effectively. Aiming at this problem,a new learning pattern combined attribution-detector and PN-learning framework is proposed to improve performance of joint detection and tracking framework which is based on PHD algorithm. The PN module is used to extract high-confidence targets from the results of attribution detector. Then these filtered results is used to retrain detector and modify outputs of PHD. Simulation tests show that the new method PN—PHD not only effectively improve target tracking capability in strong clutter environment,but also provide tracking information which original PHD method cannot offer.