一种改进的红外多目标跟踪算法

An Improved Algorithm for Multiple Infrared Targets Tracking

为了实现对多个特性相似的红外目标进行精确跟踪,在均值场蒙特卡罗(MFMC)算法基础上,提出了一种改进的基于粒子滤波的红外多目标跟踪算法.对MFMC算法的粒子采样部分融合了Mean-Shift,在粒子权值计算之前用Mean-Shift将其迭代至候选目标相似度最大位置,增加了粒子的有效性;在MFMC算法信息评估机制中加入了目标大小因素,提高了机制的自适应性.实验结果表明,该改进的跟踪算法比原算法具有更好的鲁棒性和精确性,实时性得到了很大提高,能够有效解决多个红外目标的跟踪问题.

Based on the mean field Monte Carlo（MFMC） algorithm, an improved one for multiple infrared targets tracking on basis of particle filter was proposed. In this new algorithm, Mean-Shift is integrated in the sampling of particles in MFMC, so before the weighting of particles, Mean-Shift will move them to the place where reaches maximum similarity and then the effectiveness of particles is greatly improved. This algorithm also revises the mechanism for message assessment in MFMC with the consideration of the sizes of targets, and it improves the self-adaptability of the mechanism. The experimental results show that this new algorithm is more robust, accurate and efficient than MFMC. It can effectively solve the issue of multiple infrared targets tracking.