煤系地层三维地质模型精细化表达研究  

Research on expression of three-dimensional geological model of coal measure strata

作　　者：宋越 高振记[2,3] SONG Yue;GAO Zhenji(National Engineering Research Center for Geographic Information System,China University of Geosciences(Wuhan),Wuhan 430074,China;Development&Research Center,China Geological Survey,Beijing 100037,China;Technology Innovation Center of Geological Information of Ministry of Natural Resources,Beijing 100037,China)

机构地区：[1]中国地质大学(武汉)国家地理信息系统工程技术研究中心,湖北武汉430074 [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工程技术创新中心,北京100037

出　　处：《中国矿业》2020年第9期147-151,159,共6页China Mining Magazine

基　　金：中国地质调查局“国家地质大数据汇聚与管理”项目资助(编号:DD20190381);中国地质调查局“雄安新区深部三维地质结构探测”项目资助(编号:DD20189134)。

摘　　要：三维地质体建模精细化已成为现阶段建模技术日渐重视的研究方向之一。模型贴图作为地质体模型精细化表达的重要技术手段,长期处于建模环节中被忽略的部分,仅能初步实现粗糙的贴图效果,且存在贴图变形扭曲、边界缝隙等问题丞待解决。本文以某煤矿地层模型为研究对象,建立基础纹理样本库并提出了一种基于机器学习的纹理样本库扩充方法,采用Delaunay三角剖分技术对三维地质体进行重构,提高了网格模型质量,并选用切线空间法线贴图与边界无缝处理技术进行三维地质体贴图研究。研究结果表明,通过三角剖分以及法线技术贴图的三维地质体比现有三维地质体贴图效果更加精细美观,大大提高了三维地质体模型的真实感,解决了边界不连续的问题。In the technology of three-dimensional geological body,texture mapping has always been one of the parts ignored in the process of texture mapping.Although three-dimensional geological body texture mapping could be primarily realized,there still exists problems such as distortion and boundary gap.To improve the effect of texture mapping this study takes a coal mine strata model as the research object based on computer vision by building and completing the sample database of texture mapping,and by using Delaunay triangulation to improve the quality of grid model for three-dimensional reconstruction,and by choosing normal mapping and seamless border processing technology to conduct three-dimensional geological body texture mapping.The results show that the three-dimensional geological bodies by using DT and normal mapping technology are finer and prettier than the present ones.Thus,the sense of reality of three-dimensional geological body models have been greatly increased and the problem of discontinuous boundary has been solved.

关 键 词：机器学习 纹理样本库 三维地质体 切线空间法线贴图 图像融合 

分 类 号：P536[天文地球—古生物学与地层学]