Occlusion-robust object tracking based on the confidence of online selected hierarchical features

• Published in: IET image processing Vol. 12; no. 11; pp. 2023 - 2029
• Main Authors: Liu, Mingjie, Jin, Cheng-Bin, Yang, Bin, Cui, Xuenan, Kim, Hakil
• Format: Journal Article
• Language: English
• Published: The Institution of Engineering and Technology 01.11.2018
• Subjects: appearance model update strategy; CF‐based trackers; CNN; computation redundancy; convolutional layers; convolutional neural networks; correlation filters; deep learning‐based trackers; feature extraction; feature map selection method; learning (artificial intelligence); model corruption; neural nets; object detection; object tracking; occlusion‐robust object tracking; online selected hierarchical features; OTB‐2015 datasets; Research Article; tracking accuracy; tracking performance; useless information; visual object tracking; CNN; deep learning-based trackers; useless information; tracking accuracy; CF-based trackers; object detection; occlusion-robust object tracking; model corruption; online selected hierarchical features; feature extraction; correlation filters; visual object tracking; feature map selection method; object tracking; appearance model update strategy; tracking performance; computation redundancy; convolutional neural networks; learning (artificial intelligence); OTB-2015 datasets; convolutional layers; neural nets
• ISSN: 1751-9659; 1751-9667
• DOI: 10.1049/iet-ipr.2018.5454

In recent years, convolutional neural networks (CNNs) have been widely used for visual object tracking, especially in combination with correlation filters (CFs). However, the increasing complex CNN models introduce more useless information, which may decrease the tracking performance. This study proposes an online feature map selection method to remove noisy and irrelevant feature maps from different convolutional layers of CNN, which can reduce computation redundancy and improve tracking accuracy. Furthermore, a novel appearance model update strategy, which exploits the feedback from the peak value of response maps, is developed to avoid model corruption. Finally, an extensive evaluation of the proposed method was conducted over OTB-2013 and OTB-2015 datasets, and compared with different kinds of trackers, including deep learning-based trackers and CF-based trackers. The results demonstrate that the proposed method achieves a highly satisfactory performance.