非线性时间一致性的相关滤波目标跟踪  

Correlation filtering based target tracking with nonlinear temporal consistency

作　　者：姜文涛[1] 李宛宣 张晟翀[2] JIANG Wentao;LI Wanxuan;ZHANG Shengchong(School of Software,Liaoning Technical University,Huludao Liaoning 125105,China;Science and Technology on Electro-Optical Information Security Control Laboratory,Tianjin 300308,China)

机构地区：[1]辽宁工程技术大学软件学院,辽宁葫芦岛125105 [2]光电信息控制和安全技术重点实验室,天津300308

出　　处：《计算机应用》2024年第8期2558-2570,共13页journal of Computer Applications

基　　金：国防预研基金资助项目(172068);辽宁省自然科学基金资助项目(20170540426);辽宁省教育厅重点基金资助项目(LJYL049)。

摘　　要：针对现有目标跟踪算法主要采用线性约束机制LADCF(Learning Adaptive Discriminative Correlation Filters)跟踪模型容易漂移的问题,提出非线性时间一致性的相关滤波目标跟踪算法。首先,结合史蒂文斯定律,提出贴近人类视觉感知特性的非线性时间一致项,使模型相对平滑地跟踪目标,从而保证跟踪连续性,避免跟踪模型漂移;其次,采用交替方向乘子法(ADMM)求解最优函数值,保证算法的跟踪实时性;最后,利用史蒂文斯定律非线性更新滤波器,使滤波器更新因子可以根据目标的变化增强和抑制滤波器,以适应目标变化,防止滤波器退化。在4个标准数据集上与主流相关滤波和深度学习算法对比实验,相较于基线算法LADCF,所提算法的跟踪精确度和成功率在OTB100数据集上分别提升了2.4和3.8个百分点;在UAV123上分别提升了1.5和2.5个百分点。实验结果表明,所提算法能有效避免跟踪模型漂移,降低滤波器退化概率,跟踪精确度和成功率较高,面对遮挡、光照变化等复杂场景时具有较强的鲁棒性。Concerning the problem that existing target tracking algorithms mainly use the linear constraint mechanism LADCF(Learning Adaptive Discriminative Correlation Filters),which easily causes model drift,a correlation filtering based target tracking algorithm with nonlinear temporal consistency was proposed.First,a nonlinear temporal consistency term was proposed based on Stevens’Law,which aligned closely with the characteristics of human visual perception.The nonlinear temporal consistency term allowed the model to track the target relatively smoothly,thus ensuring tracking continuity and preventing model drift.Next,the Alternating Direction Method of Multipliers(ADMM)was employed to compute the optimal function value,ensuring real-time tracking of the algorithm.Lastly,Stevens’Law was used for nonlinear filter updating,enabling the filter update factor to enhance and suppress the filter according to the change of the target,thereby adapting to target changes and preventing filter degradation.Comparison experiments with mainstream correlation filtering and deep learning algorithms were performed on four standard datasets.Compared with the baseline algorithm LADCF,the tracking precision and success rate of the proposed algorithm were improved by 2.4 and 3.8 percentage points on OTB100 dataset,and 1.5 and 2.5 percentage points on UAV123 dataset.The experimental results show that the proposed algorithm effectively avoids tracking model drift,reduces the likelihood of filter degradation,has higher tracking precision and success rate,and stronger robustness in complicated situations such as occlusion and illumination changes.

关 键 词：目标跟踪 相关滤波 史蒂文斯定律 非线性时间一致性 非线性滤波器更新 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]