当背景与目标颜色相近时, 传统基于颜色直方图的粒子滤波跟踪算法容易被干扰发生错误跟踪。彩色二进制局部不变特征(CBLID)具有旋转、尺度缩放和光照不变性等特点, 将CBLID特征应用到粒子滤波跟踪算法中, 以提高跟踪算法的抗干扰能力。考虑到纹理少的目标的CBLID特征点较少, 仅仅依据CBLID特征进行跟踪不稳定, 提出一种结合CBLID特征与颜色直方图特征的粒子滤波跟踪算法。该算法依据颜色直方图的Bhattacharyya距离以及CBLID特征点匹配数目确定粒子的权值, 并根据当前帧跟踪结果及CBLID目标模板中特征点匹配数来决定是否对目标颜色模板进行更新。实验结果表明, 当目标出现遮挡或者背景颜色相近时, 该方法能够有效地提高跟踪精度, 同时它比基于SIFT特征的粒子滤波算法速度快。

In traditional particle filter algorithm, the weight of each particle is based on color feature only and updated by calculating the distance of Bhattacharyya, which may easily lead to error tracking when the object and the background have similar color or the object is occluded. Colored binary local invariant descriptor (CBLID) feature is highly discriminative, it can improve the performance of the algorithm effectively when the target is occluded or they have similar color. But using CBLID feature only is insufficient to describe small targets. A new method is proposed to handle with this situation, in which the target model is built by CBLID feature and color feature together. The color target model will be updated through the numbers of matching points between current frame and the CBLID target model. Experimental results show that the proposed method can effectively improve the tracking precision when the object is occluded or under a background with similar color. Besides, it is faster than using scale invariant feature transformation (SIFT) in particle filter algorithm.