当前位置: X-MOL 学术J. Real-Time Image Proc. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Channel-independent spatially regularized discriminative correlation filter for visual object tracking
Journal of Real-Time Image Processing ( IF 3 ) Pub Date : 2020-04-22 , DOI: 10.1007/s11554-020-00967-y
A. Varfolomieiev

The study proposes the improvements for visual object trackers based on discriminative correlation filters. These improvements consist in the development of the channel-independent spatially regularized method for filter calculation, which is based on the alternating direction method of multipliers as well as in the use of additional features that are the result of the backprojection of normalized weighted object histogram. The VOT Challenge 2018 benchmark has confirmed that the proposed approaches allow to increase the tracking robustness. Particularly, by the value of expected average overlap (EAO = 0.1828), the tracker that uses these approaches (CISRDCF) can reach the level of more computationally complex competitors that utilize convolutional neural features. At the same time, the software-optimized version of the CISRDCF tracker, which implements the suggested improvements has moderate computational complexity and can operate in the real-time both on the PC and on the mid-range ARM-based processors, making the CISRDCF tracker promising for embedded applications.



中文翻译:

与通道无关的空间正则化判别相关滤波器,用于视觉对象跟踪

该研究提出了基于判别相关滤波器的视觉对象跟踪器的改进。这些改进包括开发用于滤波器计算的与通道无关的空间正则化方法,该方法基于乘法器的交替方向方法,以及使用归一化加权对象直方图的反投影结果所带来的其他功能。2018年VOT挑战赛基准测试已经确认,所提出的方法可以提高跟踪的鲁棒性。特别是,通过预期的平均重叠值(EAO = 0.1828),使用这些方法的跟踪器(CISRDCF)可以达到利用卷积神经特征进行计算复杂的竞争者的水平。同时,CISRDCF跟踪器的软件经过优化,

更新日期:2020-04-23
down
wechat
bug