Abstract: In order to reduce the influence of illumination variation and target occlusion on moving object tracking, this paper proposes a multi-region target tracking algorithm based on improved local sensitive histogram. First, the localized sensitive histogram is improved and its fast algorithm is designed. Then, the improved local sensitive histogram is used as the target modeling method in the multi-region tracking algorithm to improve the accuracy of the target modeling and reduce the computational complexity of feature extraction. Finally, the method of parameter estimation based on statistical ranking and least squares estimation is adopted to compute the similarity between the whole object and template, which is needed to fuse the similarity of each region block in the multi-region tracking algorithm. Experimental results show that the algorithm can effectively cope with the interference of illumination changes, object deformation and occlusion, achieving real-time robust object tracking.

Key words: object tracking, illumination variation, target occlusion, locality sensitive histogram, multi-region tracking

摘要： 为克服光照变化和目标遮挡对运动目标跟踪的影响，提出了一种基于改进的局部敏感直方图的多区域目标跟踪算法。改进了局部敏感直方图并设计了快速算法；将改进的局部敏感直方图作为多区域跟踪算法中的目标建模方式，提高目标建模的准确性且降低提取目标特征的计算复杂度；针对多区域跟踪算法中融合各个区域块的特征相似值的需要，采用基于统计排序和最小二乘估计的参数估计方法计算整个目标块与模板的相似度。实验结果表明该算法能有效应对目标跟踪过程中光照变化、目标形变与遮挡的干扰，实现实时鲁棒的目标跟踪。

关键词: 目标跟踪, 光照变化, 目标遮挡, 局部敏感直方图, 多区域跟踪