文献类型：期刊文献

英文题名：CCBNet: collaborative calibration and bridging network for visible-infrared person re-identification

作者：Gao, Wen[1,2];Dai, Lei[1,2];Chen, Zhihua[1,2]

机构：[1]East China Univ Sci & Technol, Dept Comp Sci & Engn, Shanghai 200237, Peoples R China;[2]East China Univ Sci & Technol, State Key Lab Ind Control Technol, Shanghai 200237, Peoples R China

年份：2026

卷号：42

期号：8

外文期刊名：VISUAL COMPUTER

收录：;EI(收录号：20262120768686);WOS:【SCI-EXPANDED(收录号:WOS:001771673900003)】；

基金：This work was supported by the National Natural Science Foundation of China (Grant Nos. 62272164, 62572188, and 62306113) and the State Key Laboratory of Industrial Control Technology, China (Grant No. ICT2026A25).

语种：英文

外文关键词：Visible-infrared person re-identification; Global-Local Channel Augmentation; Cross-modal bridging; Distribution consistency loss

摘要：Visible-infrared person re-identification (VI-ReID) aims to match pedestrian identities across heterogeneous modalities with substantial cross-modal discrepancy. Most existing approaches attempt to reduce the gap via direct feature alignment, which may distort modality-specific semantics and lead to inconsistent representations. To address these limitations, we propose a Collaborative Calibration and Bridging Network (CCBNet) for VI-ReID. The proposed framework is composed of three complementary components. First, we introduce a Global-Local Channel Augmentation (GLCA) strategy to mitigate modality-specific color bias by applying channel exchange at both global and local scales, improving cross-modal robustness while preserving semantic structure. Second, we design an Adaptive Modal Alignment Bridging (AMAB) module that constructs an intermediate representation between the visible and infrared modalities through lightweight cross-modal interactions, enabling a smoother semantic transition across modalities. Third, we propose a Modality Bridge Distribution Consistency (MBDC) loss, which facilitates refined optimization of the intermediate-modality features, enhancing their accuracy and robustness while promoting coherent alignment across visible, infrared, and intermediate representations. Extensive experiments conducted on the SYSU-MM01 and RegDB datasets demonstrate that the proposed CCBNet consistently outperforms state-of-the-art VI-ReID methods. The related code is available at https://github.com/sereinlll/CCBNet.