首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于深度学习的镜下矿石矿物的智能识别实验研究
引用本文:徐述腾,周永章.基于深度学习的镜下矿石矿物的智能识别实验研究[J].岩石学报,2018,34(11):3244-3252.
作者姓名:徐述腾  周永章
作者单位:广东省地质过程与矿产资源探查重点实验室, 广州 510275;中山大学地球环境与地球资源研究中心, 广州 510275;中山大学地球科学与工程学院, 广州 510275,广东省地质过程与矿产资源探查重点实验室, 广州 510275;中山大学地球环境与地球资源研究中心, 广州 510275;中山大学地球科学与工程学院, 广州 510275
基金项目:本文受国家重点研发计划重点专项(2016YFC0600506)、国家自然科学基金项目(41273040)、中国地质调查局项目(12120113067600)和广东省地质过程与矿产资源探查重点实验室基金联合资助.
摘    要:矿石矿物鉴定的智能化是智能地质学和智能矿床学的基础技术之一。计算机视觉技术和深度学习理论使矿石矿物鉴定的智能化成为可能。本研究基于深度学习系统Tensor Flow,以吉林夹皮沟金矿和河北石湖金矿的黄铁矿、黄铜矿、方铅矿、闪锌矿等硫化物矿物为例,设计有针对性的Unet卷积神经网络模型,有效自动提取矿相显微镜下矿石矿物的深层特征信息,实现镜下矿石矿物智能识别与分类。实验显示,模型在训练过程中,随着训练次数的增加,模型精度在不断增大,损失函数不断减小;经过3000个批处理之后,模型精度和损失函数基本趋于稳定。训练出的模型对测试集中的显微镜镜下矿石矿物照片的识别成功率均高于90%,说明实验所建立的模型,具有很好的图像特征提取能力,能完成镜下矿石矿物智能识别的任务。

关 键 词:卷积神经网络算法  深度学习  矿物自动识别  地质大数据  智能地质学  机器学习
收稿时间:2018/5/30 0:00:00
修稿时间:2018/8/20 0:00:00

Artificial intelligence identification of ore minerals under microscope based on deep learning algorithm
XU ShuTeng and ZHOU YongZhang.Artificial intelligence identification of ore minerals under microscope based on deep learning algorithm[J].Acta Petrologica Sinica,2018,34(11):3244-3252.
Authors:XU ShuTeng and ZHOU YongZhang
Institution:Guangdong Provincial Key Laboratory of Mineral Resources and Geological Processes, Guangzhou 510275, China;Center for Earth Environment & Resources, Sun Yat-sen University, Guangzhou 510275, China;School of Earth Sciences and Engineering, Sun Yat-sen University, Guangzhou 510275, China and Guangdong Provincial Key Laboratory of Mineral Resources and Geological Processes, Guangzhou 510275, China;Center for Earth Environment & Resources, Sun Yat-sen University, Guangzhou 510275, China;School of Earth Sciences and Engineering, Sun Yat-sen University, Guangzhou 510275, China
Abstract:Intelligent identification of ore minerals is one of the basic technologies for intelligent geology and intelligent ore deposit geology. Application of deep learning in computer vision opens a way to intelligent identification of ore minerals. Exemplified by pyrite, chalcopyrite, galena, sphalerite and other sulfide minerals from the Jiapigou gold deposit in Jilin Province and the Shihu gold deposit in Hebei Province, a Unet Convolution Neural Network (CNN) model under the TensorFlow framework is designed to automatically extract the deeply hidden essential features of the ore minerals for realizing their automatic identification and classification. It is shown that the model accuracy increases and loss function decreases with training iteration growing during training process. After 3000 iterations, the model accuracy and loss function become stable. Identification accuracy of the trained model on the test dataset is higher than 90%, demonstrating that the established model is good enough to extract the image features, thus it is capable to intelligently identify the ore minerals under microscope.
Keywords:Convolution Neural Network  Deep learning  Mineral automatic identification  Geological big data  Intelligent geology  Machine learning
本文献已被 CNKI 等数据库收录!
点击此处可从《岩石学报》浏览原始摘要信息
点击此处可从《岩石学报》下载免费的PDF全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号