.png){kind=icon}
Return
Robust correlation tracking with closed-loop feedback control
.png){kind=icon}
PRE
AI
DOI：10.1007/s00530-025-01816-3.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Owing to high efficiency and excellent performance, discriminative correlation filter (DCF) based methods have attracted growing attention in the area of visual tracking recently. However, most of them ignore the influence of the tracking output and can be considered open-loop systems, which easily cause error accumulation and even model drift. In this paper, a novel correlation tracking approach is proposed to address these issues by introducing the closed-loop feedback technique into the DCF framework. From a closed-loop perspective, the tracking process is reconstructed into two parts, including forward tracking and backward feedback. In the process of forward tracking, we follow the DCF framework with enhanced feature representation to locate the target object. In the process of backward feedback, we design two effective controllers to automatically regulate forward tracking by employing the information extracted from the tracking output. With the two controllers, the proposed approach is capable of reducing error accumulation and alleviating model drift. Experimental results on five public datasets demonstrate that our tracker (called CLFCT) achieves competitive performance compared with state-of-the-art tracking methods while maintaining a real-time speed of nearly 26 FPS.
Keywords:
Visual tracking
Discriminative correlation filter
Closed-loop feedback
Feature representation
Journal
IF:
3.1
Papers:
2.8K
Citations:
2.7K
Organization
Cited Papers
.png){kind=icon}
ShareSaveGeneric Tracking and Probabilistic Prediction Framework and Its Application in Autonomous Driving
OAAI
OAAIShareSave
ShareSaveShareSaveShareSave
View more