基于相关滤波跟踪算法的测试与分析

Test and Analysis of Tracking Algorithm Based on Correlation Filter

基于相关滤波的跟踪算法具有跟踪速度快以及在线学习并同步更新的特点,对其进行测试与分析。详细介绍几种经典的相关滤波跟踪模型,根据诸多算法在OTB-2013(object tracking benchmark)上的测试运行结果,对比分析各种算法的总体跟踪性能和在跟踪难点问题上的跟踪鲁棒性能,横向分析相关滤波算法在应对各种跟踪难点问题时所做的一系列改进,并对未来的发展给出几点看法。结果表明,该分析可对相关研究提供参考借鉴。

The tracking algorithm based on correlation filter has the characteristics of fast tracking speed,online learning and synchronous updating,and it is tested and analyzed.Several classical correlation filter tracking models are introduced in detail,according to testing results of many algorithms in the OTB-2013(object tracking benchmark),overall tracking performance and tracking robustness of various algorithms are compared and analyzed.A series of improvements of correlation filter algorithms in dealing with various tracking difficulties are analyzed laterally,and some views on the development of the future are given.The results show that this analysis can provide reference for related research.