SiamRCSC: Robust siamese network with channel and spatial constraints for visual object tracking
Yu Zheng,Yong Liu,Xun Che
DOI: https://doi.org/10.1007/s00530-024-01524-4
IF: 3.9
2024-10-25
Multimedia Systems
Abstract:Locating and classifying the target object is performed by the siamese-based tracking framework by evaluating the similarity on the feature maps from the template and search branches. While the promising tracking performances have been achieved by the state-of-the-art (SOTA) trackers, the robustness and accuracy of these trackers significantly decline in complex scenes, such as deformation in appearance and interference in background. In order to suppress these defects, sophisticated template updating mechanisms as well as refinement modules have been proposed by many recent works, but these methods do not consider higher-order correlations between features, which play an important role in obtaining accurate localization and classification information. For this, we propose a high-order spatial constraint (HOSC) module, considering high-order correlations between features through recursively executing interactions in feature maps. Additionally, to learn superior feature representations and more discriminative features, a more efficient and effective channel attention with mutual compensation (CAMC) module is proposed in this work, where channel attention from the template branch is utilized to enhance the channel constraint of the search branch for improving the learning of discriminative features and it would be advantageous for the template branch to encode more contextual information from the search image. Finally, extensive experiments were conducted on datasets (OTB100, VOT2018, LaSOT and GOT-10K), and the proposed method achieves competitive performance compared to SOTA trackers (CNN-based).
computer science, information systems, theory & methods