Hyperspectral signal prediction method based on hyperspectral remote sensing image and pseudo label guidance

By using a cascaded correction algorithm and a pseudo-label-guided method to correct and train remote sensing spectral data, the problems of low quality and inaccurate prediction of remote sensing spectral data were solved, and high-precision soil composition prediction was achieved.

CN121834477BActive Publication Date: 2026-06-05KUNMING UNIV OF SCI & TECH
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
KUNMING UNIV OF SCI & TECH
Filing Date
2026-03-13
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing technologies for predicting soil composition using remote sensing spectral data suffer from complex noise interference and low quality, making it difficult to guarantee the accuracy of predictions and the generalization ability of the model.

Method used

A method based on hyperspectral remote sensing imagery and pseudo-labels is adopted. The remote sensing spectral data is corrected by a cascade correction algorithm, a two-branch fusion model is constructed, and supervised contrast loss is introduced by classifying pseudo-labels for training to improve the model's performance.

Benefits of technology

It effectively improved the quality of remote sensing spectral data, enhanced the prediction accuracy and robustness of the model, and improved the prediction accuracy of soil composition.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121834477B_ABST
    Figure CN121834477B_ABST
Patent Text Reader

Abstract

The present application relates to a kind of based on hyperspectral remote sensing image and pseudo label guide's hyperspectral signal prediction method, belong to hyperspectral signal prediction technical field.The method includes: acquisition remote sensing spectral data and soil spectral data, and remote sensing spectral data is initially preprocessed;Based on the designed series correction algorithm, the remote sensing spectral data after initial preprocessing is corrected using soil spectral data and is secondarily preprocessed;The remote sensing spectral data after secondary preprocessing is imaged;Dual-branch fusion model containing spectral branch and image branch is constructed;Classified pseudo label is constructed, and supervision contrast loss is introduced to train dual-branch fusion model;The remote sensing spectral data to be predicted is input to the dual-branch fusion model trained, and the hyperspectral signal prediction result is obtained to realize soil content prediction.The present application is aimed at solving the technical problems that existing technology has complex noise and low quality when acquiring remote sensing spectral data, and it is difficult to ensure the accuracy of prediction.
Need to check novelty before this filing date? Find Prior Art