Hyperspectral lithologic mapping method based on MS-1DCNN-DRS network model

By introducing the multi-spectral area spectral information and the MS-1DCNN-DRS network model of the depth residual shrinkage denoising network module in the hyperspectral lithology map, the problem that traditional multispectral remote sensing technology is difficult to identify different lithology in lithology classification and map are solved, and a high-precision and high-efficiency lithology map effect is achieved.

CN119693786APending Publication Date: 2025-03-25CHINA AERO GEOPHYSICAL SURVEY & REMOTE SENSING CENT FOR LAND & RESOURCES
0 Cites 0 Cited by

Patent Information

Application Number
CN202411608347.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-12
Publication Date
2025-03-25

AI Technical Summary

Technical Problem

Traditional multispectral remote sensing technology is difficult to identify and distinguish different lithologies with similar colors and texture characteristics in lithologic classification and mapping, and deep learning methods have problems with noise problems and high computing resources when processing hyperspectral data.

Method used

A hyperspectral lithologic mapping method based on MS-1DCNN-DRS network model is proposed. By introducing multi-spectral area spectral information and deep residual shrinkage denoising network module, the noise problem and spatial attribute learning ability of one-dimensional convolutional neural network are optimized.

Benefits of technology

It realizes high-precision and high-efficiency lithologic mapping, reduces the impact of noise, and is suitable for all types of hyperspectral data, including both satellite and airborne, providing basic information for ore prospecting prediction.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention provides a hyperspectral lithologic mapping method based on an MS-1DCNN-DRS network model. The hyperspectral lithologic mapping method comprises the following steps: step 1, preprocessing hyperspectral data and collecting auxiliary data; step 2, hyperspectral data enhancement; 3, making a sample data set; step 4, an MS-1DCNN-DRS network model is constructed; 5, model training and testing; 6, storing the model; and step 7, lithology mapping. According to the method, the advantages that the one-dimensional convolutional neural network is high in operation efficiency and the geological boundary is matched with the reality are reserved, learning of spatial region spectrum information is added, the problem that the one-dimensional convolutional neural network is large in noisy point number is solved, and by adding the deep residual shrinkage network, the real-time performance of the method is improved. The noise influence of the hyperspectral image is reduced; and repeated information with relatively high correlation between different wavebands is eliminated. High efficiency is kept, high precision is guaranteed, and basic information can be quickly provided for prospecting prediction.
Need to check novelty before this filing date? Find Prior Art