一种跨模态行人重识别方法及系统
By constructing a cross-modal semantic hypergraph and generating dynamic semantic regions, extracting multi-granular local features and introducing environment-aware modulation, the problems of modal gap and illumination variation in cross-modal pedestrian re-identification are solved, and highly accurate cross-modal pedestrian matching is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUZHOU UNIVERSITY
- Filing Date
- 2026-03-09
- Publication Date
- 2026-07-17
AI Technical Summary
Existing cross-modal pedestrian re-identification methods struggle to achieve accurate matching when faced with challenges such as the modal gap between visible light and infrared images, illumination variations, complex background interference, and the variability of pedestrian postures. In particular, they lack cross-scene generalization capabilities in all-weather monitoring scenarios.
A cross-modal pedestrian re-identification method is adopted. By constructing a cross-modal semantic hypergraph, combining hypergraph convolution and dynamic semantic region generation, multi-granular local features are extracted, and dynamic weighted metric and environment-aware modulation are introduced to achieve adaptive cross-modal pedestrian matching.
It breaks through the limitations of traditional rigid alignment, achieves accurate multi-scale representation of non-rigid appearance, enhances the model's adaptability to different imaging conditions, and improves the accuracy and stability of cross-modal similarity measurement.
Smart Images
Figure CN121811455B_ABST
Abstract
Citation Information
Patent Citations
Cross-modal pedestrian re-identification method based on multi-granularity feature utilization
CN114998928A
Video pedestrian re-identification method based on attention space-time diagram network
CN116797966A