耦合颜色和纹理特征的矿物图像数据深度学习模型与智能识别方法  

A Deep Learning and Intelligent Recognition Method of Image Data for Rock Mineral and its Implementation

作　　者：李明超[1] 刘承照 张野[1] 朱月琴[2,3] LI Mingchao;LIU Chengzhao;ZHANG Ye;ZHU Yueqin(State Key Laboratory of Hydraulic Engineering Simulation and Safety,Tianjin University,Tianjin 300354,China;Development Research Center of China Geological Survey,Beijing 100037,China;Key Laboratory of Geological Information Technology,Ministry of Natural Resources,Beijing 100037,China)

机构地区：[1]天津大学水利工程仿真与安全国家重点实验室,天津300354 [2]中国地质调查局发展研究中心,北京100037 [3]自然资源部地质信息技术重点实验室,北京100037

出　　处：《大地构造与成矿学》2020年第2期203-211,共9页Geotectonica et Metallogenia

基　　金：天津市杰出青年科学基金项目(17JCJQJC44000);国家优秀青年科学基金项目(51622904);国家重点研发计划项目(2016YFC0600510)联合资助。

摘　　要：在野外地质勘探过程中,矿物识别是一项重要的任务。但是在复杂的地质勘探条件下准确识别矿物是一项极具挑战性的任务。本文首先结合矿物知识,对图像中矿物的颜色和亮度进行判别,从而强化矿物图像的纹理特征,基于强化后的纹理特征,利用深度学习模型和迁移学习方法建立矿物深度学习模型;同时利用K-means算法提取矿物颜色特征,建立颜色特征模型;最终基于矿物识别深度学习模型和颜色特征模型,建立矿物图像的耦合识别模型,提出了一套完整的矿物智能识别分析方法。在深度学习模型建立过程中,共对19类矿物,总计6203张图像进行了训练,同时使用456张图像对耦合模型进行测试,其top-1准确率和top-3准确率分别达到72.2%和91.2%。基于所训练的模型,在Android系统开发了矿物图像智能识别应用系统,主要由矿物识别、矿物知识卡和矿物资料存储三个模块组成,提供矿物实时识别、矿物基础知识及实时储存等功能。作为野外地质勘察的辅助工具,此应用可以减少重复劳动,提升野外地质勘察的数字化和信息化水平。Mineral identification is an important task of geological survey,which can be a big challenge in complex geological conditions.In this research,the color and brightness are distinguished using mineral knowledge firstly,which is used to bring out the texture features more obviously.Next,the deep learning model and transfer learning method are applied to build a mineral identification model with enhanced features.Meanwhile,a color model for mineral identification using K-means is also established.Finally,a comprehensive model combining deep learning model and color model is built and a framework of identifying minerals is proposed.In the model establishing process,6203 images with 19 labels are trained and the top-1 and top-3 accuracy reach 72.2%and 91.2%,respectively.Furthermore,a mineral identification application is developed on the Android system.The application is made up of three modules including mineral knowledge,mineral identification and data storage,which provide real-time recognition,mineral knowledge and data storage.

关 键 词：矿物识别 图像分类 深度学习 迁移学习 智能识别 

分 类 号：P628.5[天文地球—地质矿产勘探]