时间驱动的异常学习相关滤波器的目标跟踪

Aberrance Learning via Time-driven Correlation Filter for Object Tracking

为了解决传统的相关滤波跟踪算法在复杂环境中容易跟踪失败的问题,本文提出时间驱动的异常学习相关滤波器,旨在提高模型在复杂环境下的适应性,实现安全有效的目标跟踪。通过引入结合异常学习的时间正则项,该模型不仅可以结合滤波器响应相似度和时间域特征搜索到目标,达到抑制异常的效果,还可以提高外观模型在时域中的鲁棒性,缓解时间滤波器退化。另外,本文采用交替方向乘子法(Alternating Direction Method of Multipliers,ADMM)算法实现模型的优化过程,大大减少模型的计算复杂度。大量的实验结果证实了所提出的跟踪算法性能的优越性。

In order to solve the problem that the traditional correlation filter tracking algorithm is easy to fail in complex environments,this paper proposes time-driven correlation filter with aberrance learning(ALTCF)to improve the adaptability of the model in complex environments and achieve safe and effective object tracking.By introducing temporal regularization term with aberrance learning,the model in this paper can not only search for objects by combining the similarity of filter response maps and temporal features to achieve the effect of suppressing aberrance,but also improve the robustness of appearance model and alleviate temporal filter degradation.In addition,this paper uses the alternating direction method of multipliers(ADMM)algorithm to achieve the optimization process of the model,which greatly reduces the computational complexity of the model.A large number of experiments confirm the superiority of ALTCF tracking performance.