Cascaded Image Fusion Bridged RGB-T Tracking
IEEE Signal Processing Letters · 2026 · 33 · 2640-2644
DOI: 10.1109/lsp.2026.3705376 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Signal Processing Letters, 2026. Verfasst von Shengjia An und 4 weiteren Autor:innen. Die Arbeit nennt 43 Literaturangaben. Gefördert von: Guangxi Key Research and Development Program, National Natural Science Foundation of China, Guangxi Key Laboratory of Automatic Detecting Technology and Instrumentation.
Abstract (Originalsprache)
RGB-T object tracking exploits the complementarity of RGB and thermal infrared modalities to enable robust tracking under complex scenarios. However, existing methods mainly perform cross-modal interactions in deep feature space, and have loose coupling between fusion and tracking tasks. Thus they fail to fully exploit the incremental discriminative information from cross-modal interactions, and the fused features cannot align with the tracking localization demands. To address these issues, we propose CIFBTrack, a cascaded RGB-T tracking framework that bridges the collaborative gap between high-level object tracking and low-level cross-modal fusion tasks. Specifically, our end-to-end cascaded architecture deeply couples the fusion module with the tracking backbone, and uses fusion loss to guide cross-modal information flow towards the backbone. This enables fused features to retain basic bimodal information while highlighting complementary discriminative cues for accurate target localization. In addition, we design a Fusion Prompt Block (FPB) as the core interaction unit, which effectively improves key bimodal feature capture and cross-modal interaction. Extensive experiments on RGBT210, RGBT234 and LasHeR benchmarks demonstrate that CIFBTrack achieves state-of-the-art performance.
Förderung
- Guangxi Key Research and Development Program
- National Natural Science Foundation of China
- Guangxi Key Laboratory of Automatic Detecting Technology and Instrumentation
Zitieren
APA
An, S., Li, Z., Zhang, S., Li, Z., & Zhong, B. (2026). Cascaded Image Fusion Bridged RGB-T Tracking. IEEE Signal Processing Letters, 33, 2640-2644. https://doi.org/10.1109/lsp.2026.3705376
Vancouver
An S, Li Z, Zhang S, Li Z, Zhong B. Cascaded Image Fusion Bridged RGB-T Tracking. IEEE Signal Processing Letters. 2026;33:2640-2644. doi:10.1109/lsp.2026.3705376
BibTeX
@article{An2026,
author = {An, Shengjia and Li, Zhi and Zhang, Shaorong and Li, Zhenjie and Zhong, Bineng},
title = {Cascaded Image Fusion Bridged RGB-T Tracking},
journal = {IEEE Signal Processing Letters},
year = {2026},
volume = {33},
pages = {2640--2644},
doi = {10.1109/lsp.2026.3705376}
}
EUVIMED