针对基于离散余弦变换的压缩感知哈希算法在光照变化、目标发生形变或者局部遮挡的情况下难以跟踪的问题, 提出了一种基于类间方差和离散余弦变换融合的模板匹配增强哈希算法。该算法是一种运用类间方差阈值分割和离散余弦变换来提取目标不同特征信息, 用快速增强差异法生成哈希序列来降低光照影响, 用抽屉原理缩短汉明距离的比较时间的自动更新模板的目标跟踪算法。本文算法与传统哈希算法、基于DCT的压缩感知哈希算法在视频David, Girl和CarScale中进行了跟踪实验。实验结果表明, 该算法在光照变化、目标形变和局部遮挡的情况下提高了目标的跟踪成功率, 具备良好的鲁棒性, 满足了实时跟踪的要求。

Aiming at the problem that the compression perceptual Hash algorithm based on discrete cosine transform is difficult to track the target in the case of illumination change, target deformation or local occlusion, this paper proposes an enhanced template-matching Hash algorithm based on interclass variance and discrete cosine transform. This algorithm is a target tracking algorithm with automatic template updating, which uses the interclass variance threshold segmentation and discrete cosine transform to extract the different characteristic information of the target, the rapidly intensified method of difference to generate the Hashing sequence for reducing the influence of illumination, and the drawer principle to shorten the time for Hamming distance comparison. Experiments were made by using this algorithm, the traditional Hash algorithm, and the compression perceptual Hash algorithm based on DCT for tracking the three public standard videos of David, Girl, and CarScale.The results show that, the algorithm can improve the success rate of the target tracking under illumination change, target deformation or local occlusion, has good robustness and satisfies the real-time tracking requirements.