加速的相关滤波跟踪-论文解读

加速的相关滤波跟踪-论文解读

小方哥: 微信公众号:AI人工智能模式识别算法工作室

  1. 论文名称

An Accelerated Correlation Filter Tracker

引用

Xu T, Feng Z H, Wu X J, et al. An accelerated correlation filter tracker[J]. Pattern Recognition, 2020, 102: 107172.


2. 论文摘要

随着判别性相关滤波以及鲁棒的深度网络特征的提取,视觉跟踪研究领域出现一些比较先进的算法。尽管两者的组合在很大程度上提高的目标跟踪算法的性能,但是由于深度特征提取的复杂性以及在线更新模型的耗时性,使得这些跟踪的方法在实际的场景中应用的不太广泛。为了解决上述问题,本文通过在ADMM优化方法的序列迭代过程中,增加一个冲量,同时减少DCF模型参数和范数的误差影响来加速ADMM优化方法的优化。根据所提出的优化方法,对DCF相关滤波的方法进行了重新设计,来寻求具有区分性的空间归一化的特征通道。另一个加速是对相关滤波优化过程中进行的自适应初始化方法。通过这两个方法可以加速DCF相关滤波的收敛。在跟踪数据集上测试,证实了该方法的有效性。

Recent visual object tracking methods have witnessed a continuous improve-

ment in the state-of-the-art with the development of efficient discriminative

correlation filters (DCF) and robust deep neural network features. Despite

the outstanding performance achieved by the above combination, existing ad-

vanced trackers suffer from the burden of high computational complexity of the

deep feature extraction and online model learning. We propose an accelerated

ADMM optimisation method obtained by adding a momentum to the optimi-

sation sequence iterates, and by relaxing the impact of the error between DCF

parameters and their norm. The proposed optimisation method is applied to

an innovative formulation of the DCF design, which seeks the most discrimina-

tive spatially regularised feature channels. A further speed up is achieved by

an adaptive initialisation of the filter optimisation process. The significantly

increased convergence of the DCF filter is demonstrated by establishing the op-

timisation process equivalence with a continuous dynamical system for which

the convergence properties can readily be derived. The experimental results

obtained on several well-known benchmarking datasets demonstrate the effi-

ciency and robustness of the proposed ACFT method, with a tracking accuracy

comparable to the start-of-the-art trackers.


3. 目标框架

主要贡献是对ADMM算法的加速以及自适应初始化参数的值。


通过R_ADMM与ADMM的优化

通过R_A-ADMM优化

4. 实验结果

发布于 2021-01-26 17:09