中文摘要：

为了提高复杂背景下红外目标跟踪的准确性和鲁棒性,提出了紧耦合粒子滤波（PF）与均值漂移（mean shift）的红外目标跟踪方法。在PF框架下,利用一组5参数集（中心横坐标、中心纵坐标、宽度、高度以及倾斜角）作为状态变量表征随机的粒子样本;然后使用自适应均值漂移作为一种迭代模式寻找过程,对随机粒子样本进行重新分配,使粒子向目标状态的最大后验核密度估计方向移动,同时利用迭代过程中的Bhattacharyya系数对粒子的权值进行更新;最后利用重新分配后的加权粒子集合实现对红外目标的跟踪。实现结果表明,与一般的PF相比,本文方法能有效减少所需粒子数（N=15）,进而降低跟踪耗时;与现有的PF与均值漂移相结合的方法相比,本文方法在耗费时间仅增加14%的代价上,使跟踪误差大大降低（约为原误差的1/3至1/4）,准确性和鲁棒性得到显著提高;本文方法能够实现在复杂背景下稳健准确地跟踪红外目标。

英文摘要：

A novel infrared target tracking method with tightly coupling particle filer（PF）and mean shift is proposed in order to improve the accuracy and robustness of tracking.Based on the PF,a five-parameter set（2-D central location,width,height,and orientation）is utilized as state vector to describe a random particle sample,and then the adaptive mean shift is introduced as an iterative seeking procedure,in which particles move toward the maximal posterior kernel density estimation of target state.The weights of particles are updated by the Bhattacharyya coefficient as the mean shift iterative operation.Finally,the weighted particles are used to track the infrared target.Experimental results indicate that the number of particles（N=15）needed in the proposed method is much less than that of the classical PF.Compared with the existing methods combining PF and mean shift,the tracking error of the proposed method is greatly reduced（about 1/3-1/4）at the cost of the tracking time increase by only 14%,and the accuracy and robustness are significantly improved.The proposed method can track the infrared target stably and accurately even at complicated background.