Sixian Chan, Jian Tao, Xiaolong Zhou, Binghui Wu, Hongqiang Wang and Shengyong Chen
Visual tracking technology enables industrial robots interacting with human beings intelligently. However, due to the complexity of the tracking problem, the accuracy of visual…
Abstract
Purpose
Visual tracking technology enables industrial robots interacting with human beings intelligently. However, due to the complexity of the tracking problem, the accuracy of visual target tracking still has great space for improvement. This paper aims to propose an accurate visual target tracking method based on standard hedging and feature fusion.
Design/methodology/approach
For this study, the authors first learn the discriminative information between targets and similar objects in the histogram of oriented gradients by feature optimization method, and then use standard hedging algorithms to dynamically balance the weights between different feature optimization components. Moreover, they penalize the filter coefficients by incorporating spatial regularization coefficient and extend the Kernelized Correlation Filter for robust tracking. Finally, a model update mechanism to improve the effectiveness of the tracking is proposed.
Findings
Extensive experimental results demonstrate the superior performance of the proposed method comparing to the state-of-the-art tracking methods.
Originality/value
Improvements to existing visual target tracking algorithms are achieved through feature fusion and standard hedging algorithms to further improve the tracking accuracy of robots on targets in reality.
Details
Keywords
Xiaolong Zhou, Pinghao Wang, Sixian Chan, Kai Fang and Jianwen Fang
Visual object tracking plays a significant role in intelligent robot systems. This study aims to focus on unlocking the tracking performance potential of the deep network and…
Abstract
Purpose
Visual object tracking plays a significant role in intelligent robot systems. This study aims to focus on unlocking the tracking performance potential of the deep network and presenting a dynamic template update strategy for the Siamese trackers.
Design/methodology/approach
This paper presents a novel and efficient Siamese architecture for visual object tracking which introduces densely connected convolutional layers and a dynamic template update strategy into Siamese tracker.
Findings
The most advanced performance can be achieved by introducing densely connected convolutional neural networks that have not yet been applied to the tracking task into SiamRPN. By using the proposed architecture, the experimental results demonstrate that the performance of the proposed tracker is 5.8% (area under curve), 5.4% expected average overlap (EAO) and 3.5% (EAO) higher than the baseline on the OTB100, VOT2016 and VOT2018 data sets and achieves an excellent EAO score of 0.292 on the VOT2019 data set.
Originality/value
This study explores a deeper backbone network with each convolutional network layer densely connected. In response to tracking errors caused by templates that are not updated, this study proposes a dynamic template update strategy.