A cross-modal guided hyperspectral image classification framework and method under multiple degradation conditions

By employing a cross-modal guided hyperspectral image classification framework, combining HSI and active remote sensing branches, and utilizing cross-modal feature pyramids and the HyperGroupMix module for multi-scale feature fusion and style transfer, the robustness and accuracy issues of hyperspectral remote sensing images under multiple degradation conditions are resolved, achieving stable classification in complex environments.

CN120953709BActive Publication Date: 2026-06-23NANJING UNIV OF INFORMATION SCI & TECH
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
NANJING UNIV OF INFORMATION SCI & TECH
Filing Date
2025-08-22
Publication Date
2026-06-23

AI Technical Summary

Technical Problem

The robustness and classification accuracy of hyperspectral remote sensing image data under various degradation factors are problematic, especially the difficulty in maintaining spectral consistency and spatial continuity under non-independent and compound degradation conditions.

Method used

A hyperspectral image classification framework under multi-degradation conditions guided by cross-modality is adopted, which combines the HSI branch and the active remote sensing branch. Multi-scale feature fusion and cross-sample style transfer are performed through the cross-modal feature pyramid guidance module and the HyperGroupMix module. The complementary information of active remote sensing data is used to enhance the structural awareness and spatial consistency of hyperspectral images.

Benefits of technology

It significantly improves the model's adaptability to different degradation domains and classification reliability, enhances classification accuracy and stability under heterogeneous noise conditions, and alleviates performance degradation under various degradation types.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120953709B_ABST
    Figure CN120953709B_ABST
Patent Text Reader

Abstract

The application discloses a hyperspectral image classification framework and method under multiple degradation conditions guided by cross-modal, which utilizes the complementary information of active remote sensing data to enhance the structural perception and spatial consistency of degraded hyperspectral images. The framework utilizes a cross-modal feature pyramid guiding module to provide multi-level and multi-scale guidance based on active remote sensing data, thereby realizing stable and robust cross-modal feature fusion. Meanwhile, the HyperGroupMix module groups the bands of the hyperspectral image to construct spectral-spatial features, and promotes style variation between samples through feature exchange, thereby significantly improving the adaptability of the model to different degradation domains.
Need to check novelty before this filing date? Find Prior Art