Hyperspectral image spatial domain reconstruction method and system based on coupling measure and generalized ranking

By constructing a coupled metric space and hierarchical generalized ordering, spatial reconstruction of hyperspectral images is performed, which solves the problems of increased intra-class variance and decreased classification accuracy caused by viewpoint changes. This achieves robust viewpoint-independent data representation and improves classification accuracy and system generalization ability.

CN122435448APending Publication Date: 2026-07-21HAINAN NORMAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
HAINAN NORMAL UNIV
Filing Date
2026-04-24
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

Existing hyperspectral remote sensing image classification methods suffer from increased intra-class variance and decreased classification accuracy when faced with changes in sensor imaging perspective. Existing methods also suffer from high computational complexity or fail to effectively address perspective sensitivity.

Method used

By constructing a coupling metric space and combining Euclidean and cosine distances to form a weighted geometric average coupling distance, a hierarchical generalized sorting strategy is adopted to perform spatial reconstruction on hyperspectral image data blocks, generating viewpoint-independent data blocks.

Benefits of technology

It achieves robustness to changes in perspective, improves classification accuracy and the system's generalization ability in open scenarios, reduces computational burden, and provides a stable data foundation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122435448A_ABST
    Figure CN122435448A_ABST
Patent Text Reader

Abstract

The application relates to the technical field of computer vision and remote sensing image processing, in particular to a hyperspectral image space reconstruction method and system based on coupling measurement and generalized sorting, which comprises the following steps: acquiring a hyperspectral image data block to be processed; constructing a coupling measurement space, calculating the coupling distance between a center pixel and each neighborhood pixel; performing spectral generalized similar sorting, dividing the peripheral neighborhood into multiple concentric layers with the center pixel as the origin, independently ascendingly sorting the pixels in each layer according to the coupling distance between the pixels and the center pixel, and assigning a layer-in-order sorting identifier to each pixel in the layer; performing spectral vector space recombination, filling the pixel values of the corresponding positions in the data block into the corresponding layer positions of the recombined data block according to the layer-in-order sorting identifier, and generating a space reconstruction data block; the application can generate a perspective-independent and robust hyperspectral data representation, reduce the intra-class feature difference caused by the change of imaging perspective, and provide a stable and robust data basis.
Need to check novelty before this filing date? Find Prior Art