铀矿地质2025，Vol.41Issue(4)：692-706,15.DOI:10.3969/j.issn.1000-0658.2025.41.048
高光谱数据深度学习岩性分类及软件研制
Hyperspectral Data Base Deep Learning Lithology Classification and Software Development:Lithological Classification for Uranium Mineralization
摘要
Abstract
Geological mapping is one of the key tasks in the remote sensing for uranium mineralization exploration.Its main work is to classify different types of rocks according to the features of color and texture of remote sensing images.In the lithological classification of traditional manual visual interpretation of remote sensing images,the phenomena of"same material with different spectra"or"different materials with the same spectrum"of rocks are often ignored,resulting in relatively large errors in geological mapping.Therefore,it is of great significance to explore a hyperspectral remote sensing geological mapping method.This article uses the deep learning method of LSTM-2DCNN(Long short-term memory networks-2 Dimensional convolutional neural networks)to study a spectral-spatial joint hyperspectral lithological classification method and successfully has developed a hyperspectral lithological classification software.This method and software can realize functions such as training of the lithological classification pattern of hyperspectral data,predictive classification,and map output of lithological classification result.The software can perform high-precision lithological classification on with hyperspectral data and have the advantages of high classification accuracy,fast operation speed,and a good human-machine operation interface.This method and software not only effectively solve the problem of geological mapping in unmanned areas in the field of uranium mineralization,but are also suitable for the fields such as geological exploration,environmental monitoring,and land planning..关键词
高光谱数据/岩性分类软件/二维卷积神经网络/三维卷积神经网络/长短时记忆网络Key words
hyperspectral data/lithology classification software/two-dimensional convolutional neural networks/three-dimensional convolutional neural networks/long short-term memory networks分类
天文与地球科学引用本文复制引用
韩晓青,朱黎江,袁晓波,伊丕源,刘鹏飞,杨云汉..高光谱数据深度学习岩性分类及软件研制[J].铀矿地质,2025,41(4):692-706,15.基金项目
铀资源探采与核遥感全国重点实验室基金项目(编号:6142A012401)资助. （编号:6142A012401）
