基于相关滤波的视觉目标跟踪算法新进展

相关滤波跟踪算法凭借其优良的综合性能, 已成为视觉目标跟踪领域中理论研究和实践应用的热点。尽管目前已有大量研究, 但仍缺乏从跟踪框架层面对现有相关滤波跟踪算法的系统分析。因此, 文中从视频目标跟踪算法基本框架出发, 深入剖析了相关滤波跟踪算法的特性以及各工作阶段存在的基本问题。以此为依据, 归纳总结了近十年来其主要技术新进展及相应算法特点, 并对20种典型相关滤波跟踪算法进行了测试与分析。最后, 给出了相关滤波跟踪算法亟待解决的重点问题以及未来可研究方向。

Advances in visual object tracking algorithm based on correlation filter

With excellent comprehensive performance, correlation filter-based tracking algorithms have become a hotspot of the theoretical research and practical application in the field of visual object tracking. Despite many studies, there is still a lack of systematic analyses on the existing correlation filter-based tracking algorithms from the level of tracking framework. Therefore, starting from the basic framework of object tracking algorithms, the characteristics of correlation filter-based tracking algorithms are deeply analyzed, and their basic problems in each working stage are presented in this paper. On this basis, the main technological progress of correlation filter-based tracking algorithms and characteristics of corresponding algorithms in recent ten years are summarized, and 20 typical correlation filter-based tracking algorithms are evaluated and analyzed. Finally, the outstanding issues to be urgently solved and the future research directions of correlation filter-based tracking algorithms are discussed.