结合自适应时间权重的改进STRCF目标跟踪  

Improved STRCF Target Tracking Algorithm Combining Adaptive Time Weights

作　　者：吴捷[1,2] 马小虎[2] WU Jie;MA Xiao-hu(College of Information Technology,Taizhou Polytechnic College,Taizhou 225300,China;School of Computer Science and Technology,Soochow University,Suzhou 215006,China)

机构地区：[1]泰州职业技术学院信息技术学院,江苏泰州225300 [2]苏州大学计算机科学与技术学院,江苏苏州215006

出　　处：《火力与指挥控制》2021年第12期20-25,共6页Fire Control & Command Control

基　　金：国家自然科学基金(61402310);江苏省自然科学基金(BK20141195);泰州职业技术学院重点科研资助项目(1821819039)。

摘　　要：针对STRCF算法时间权重因子固定,且常因长时遮挡等原因造成跟踪漂移的问题,提出了一种自适应时间权重学习的目标跟踪算法。利用峰值旁瓣比在线动态更新时间权重因子以提升跟踪器的准确性。在跟踪器中嵌入遮挡判断模块,设定条件判断是否发生遮挡,以决定是否停止对目标位置的更新。实验结果表明:该算法的精确度与成功率分别为0.880和0.814,相比STRCF算法分别提升了2.6%和2.0%,且跟踪速度达到28.61帧/s,在CPU上可以达到实时。Aiming at the problem that the time weight factor of STRCF algorithm is fixed and the tracking drift is often caused by long-term shield etc.An adaptive time weight learning target tracking algorithm is proposed.Firstly,the peak side lobe ratio(PSR)is used to dynamically update the time weight factor online to improve the accuracy of the tracker.Secondly,the shield judgment module is embedded in the tracker to set conditions to determine whether shield occurs or not,so as to decide whether to stop updating the target position.The experimental results show that the accuracy and success rate of the algorithm are 0.880 and 0.814 respectively,2.6%and 2.0%are respectively improved compared with STRCF algorithm.The tracking speed reaches 28.61 frames/s,real-time can be realized on CPU.

关 键 词：时间权重因子 长时遮挡 目标跟踪 峰值旁瓣比 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]