一种新的目标跟踪特征融合方法

在粒子滤波目标跟踪框架下，给出一种基于颜色直方图和边缘直方图的新特征融合算法．该算法是一种通过颜色和边缘直方图的特征波和熵实现特征之间粒子数重分配的方法。通过实验比较本文算法与单特征跟踪算法的跟踪误差，结果表明本文方法较传统依靠单一特征进行目标跟踪的粒子滤波方法，在运算时间增幅不明显的情况下，实现了更为准确的目标跟踪，并可稳定的应用于运动背景下的动目标跟踪、背景光照变化下的多目标跟踪以及多目标发生相互遮挡等情况。

New Features Fusion Method for Target Tracking

A new algorithm based on muhi-features fusion is proposed, in which the color histogram and the edge histogram are combined. A method of target tracking via the redistribution of particle number is implemented by feature wave, entropy of the color histogram and the edge histogram under the framework of particle filter. Comparing the tracking errors of the proposed algorithm with single feature algorithm through experiments, the results show that targets can be tracked more accurately with our proposed algorithm than the traditional particle filtering method which is based on one feature. The proposed algorithm has a better performance when the target moves in a moving background, and is also more robust for multi-targets under the occlusion condition and illumination changing.