An RGB-T target tracking method based on low-rank gating and cross-modal fusion
By employing a low-rank gating and cross-modal fusion-based RGB-T target tracking method, the problems of modal complementarity and high computational cost in complex scenarios are solved, achieving efficient and robust multimodal visual tracking on resource-constrained platforms.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HENAN UNIV OF SCI & TECH
- Filing Date
- 2026-03-11
- Publication Date
- 2026-07-17
AI Technical Summary
Existing multimodal visual tracking technologies struggle to achieve effective complementarity and dynamic trade-offs between modalities in complex scenarios, leading to decreased robustness and accuracy. Furthermore, high-dimensional feature modeling requires significant computation, making it unsuitable for resource-constrained platforms.
An RGB-T target tracking method with low-rank gating and cross-modal fusion is adopted. Features are enhanced by grouped hierarchical pooling attention modules, lightweight recalibration is performed by combining low-rank gating adaptation modules, long and short term template libraries are constructed for adaptive fusion, and bidirectional cross attention and self-attention modeling are performed in the cross-modal attention fusion module.
While maintaining efficient computing, it improves tracking robustness and accuracy in complex scenarios, making it suitable for applications on resource-constrained platforms such as drones.
Smart Images
Figure CN122415674A_ABST