中文摘要：

针对核相关滤波器跟踪算法在目标尺度变化、快速运动及光照变化情况下跟踪性能降低的问题,提出一种基于前瞻性更新及快速异判技术的核相关滤波器跟踪算法。算法对目标历史状态以逐渐遗忘的方式加以更新,同时引入状态差分来提前应对环境变化,并且利用哈希编码匹配来控制分类器更新：首先对先前正确的目标进行哈希编码,新来一帧分类得到的最终目标同样进行哈希编码来计算相似度;然后依据相似度决定是否更新分类器或者重检测目标。实验结果表明,该算法不仅对尺度变化、快速运动有很强的鲁棒性.对其他属性如光照变化、遮挡等也有较强的鲁棒性。同时跟踪仍然保存很高的速度,平均的处理速度可达100帧/s,能实现快速精准的目标跟踪。

英文摘要：

To solve the problems of scale variation, fast motion and illumination variation in the visual tracking, a Kernel Correlation Tracking algorithm based On forward looking updating and quick abnormality judging techniques is proposed. In this algorithm, history state information is updated gradually, and the differential signal of the target is adopted to early response to environmental changes. Meanwhile, the hash code matching is used to control the classifier updating： the previously correct targets have been hash encoded to calculate similarity of the hash code of the classification goal obtained by a new frame; and then the similarity is used to decide whether to update the classifier or whether to re-detect target. Experimental results indicate that the proposed algorithm not only can obtain improvement in scale change, fast motion, but also has strong robustness for other attributes, such as illumination variation and occlusion. Moreover, it still maintains high tracking efficiency with a speed of a hundred frames per second.