一种类三步搜索策略的快速压缩跟踪算法  

Fast Compressive Tracking Algorithm Using Similar Strategy of Three Step Search

作　　者：钱凯[1] 陈秀宏[1] 孙百伟 

机构地区：[1]江南大学数字媒体学院,江苏无锡214122

出　　处：《小型微型计算机系统》2015年第8期1902-1906,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61373055)资助

摘　　要：为进一步提高压缩跟踪算法的实时性,提出一种改进的快速压缩跟踪算法.在跟踪过程的目标检测阶段采用类似于三步搜索的策略减少计算复杂度.在较大的搜索范围内,以较大的步长用滑动窗进行目标检测.得到目标位置后,在以此位置为中心的较小范围内,以较小步长用滑动窗进行目标检测,得到当前最佳位置.在前一步得到的位置上做步长最小的目标检测,由跟踪算法得到最终的目标位置.对不同视频序列的跟踪结果表明,提出的算法在对跟踪精度几乎没有影响的前提下,提升了算法的效率.相比传统的压缩跟踪算法和快速压缩跟踪算法,提出的算法更能满足实时性要求.In order to further improve the real-time performance of compressive tracking algorithm, an improved fast compressive tracking algorithm was proposed. A strategy of similar three step search was adopted to reduce the computational complexity in the de- tection procedure. First, we detect the object location based on the previous object location by shifting the window with larger step size in a larger search range. After getting object location, we detect the object location based on the previous object location by shifting the window with smaller step size in this location as the center of a smaller range. Finally, we detect the object location based on the previ- ous object location by shifting the window with the smallest step size and get the final object location by tracking algorithm. Results of tests on variant video sequences show that the proposed algorithm increases efficiency meanwhile without loss of tracking accuracy. As compared with the traditional compressive tracking algorithm and fast compressive tracking algorithm, the proposed algorithm can hold a better real-time performance.

关 键 词：压缩跟踪 实时性 三步搜索 跟踪精度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]