一种改进的相关滤波目标跟踪算法研究  

Research on an Improved Correlation Filter Target Tracking Algorithm

作　　者：甘志英 GAN Zhiying(School of Intelligence and Information Engineering,Tangshan University,Tangshan 063000,China)

机构地区：[1]唐山学院智能与信息工程学院,河北唐山063000

出　　处：《唐山学院学报》2024年第6期27-33,共7页Journal of Tangshan University

基　　金：河北省高等学校科学技术研究项目(QN2022186)。

摘　　要：针对目标跟踪中的目标运动模糊、光照变化、尺度变化等多种因素影响所造成的跟踪目标丢失的问题,采用复合相关滤波器实现目标跟踪.利用位置相关滤波器估计目标位置;利用尺度相关滤波器估计目标尺度;在线更新两种滤波器以适应跟踪中的目标位置、表征变化.通过OTB2015数据集对算法进行验证,该算法准确度达到74.5%,成功率达到63%,跟踪速度达到165fps.实验表明,该算法可对目标运动模糊、光照变化、尺度变化序列中的目标进行有效跟踪,在多个指标上优于其他多个比较算法.In order to solve the problem of target loss in the tracking caused by various factors such as target motion blur,illumination change and scale change,a composite correlation filter is used to track the target,the position correlation filter being used to estimate the target position and the scale correlation filter being used to estimate the target scale;and both filters are updated online to adapt to the changes in target position and representation during tracking.The algorithm is validated with the OTB2015 dataset,with an accuracy rate of 74.5%,a success rate of 63%and a tracking speed of 165 fps.Experimental results show that the algorithm can effectively track targets in sequences with motion blur,illumination change and scale change,outperforming several comparative algorithms in many indicators.

关 键 词：滤波器 目标跟踪 算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]