基于新模板匹配的运动目标跟踪算法

摘要

For the problem of moving target under occlusion and environmental variability in processing of moving target tracking, a novel template matching target tracking (NTMTT) algorithm based on affine transformation, state transition and the particle filter is presented. Normalized cross-correlation coefficient is used as membership to choose the optimal affine transformation parameter and the target location is determined according to the state transition. Meanwhile, the particle filter is adopted to learn the state motion of the target, and then the location of the target under occlusion is predicted. Experimental results show that the proposed template matching target tracking algorithm is efficient and effective in moving target tracking and get higher tracking performance in dealing with the illumination change, noise, the background change and occlusion compared with hyper -plane approximation tracking algorithm, mean shift tracking algorithm and covariance tracking algorithm.%针对运动目标跟踪过程中存在的目标被遮挡和环境易变性问题,提出了一种基于仿射变换状态转移与粒子滤波的新模板匹配目标跟踪算法.该算法以归一化互相关系数作为隶属度选出最佳仿射变换参数,依据特定的状态转移方法确定出目标位置,同时利用粒子滤波算法学习目标的运动状态,对遮挡中的目标位置进行预测.实验结果表明,该算法不仅可实现有效的运动目标跟踪,而且在光照适应性、抗噪性、背景适应性以及抗遮挡方面的跟踪效果明显优于超平面逼近跟踪算法、均值平移跟踪算法以及协方差跟踪算法.