覆盖区智能地质填图的探索与实践——以森林沼泽区为例  

Technical innovation and practice of intelligent geological mapping in the coverage area:A case study in the forest-swamp area

作　　者：陈虹[1,2] 杨晓 田世攀[4] 胡健民 邱士东[5] 王东明 CHEN Hong;YANG Xiao;TIAN Shipan;HU Jianmin;QIU Shidong;WANG Dongming(Institute of Geomechanics,Chinese Academy of Geological Sciences,Beijing 100081,China;Key Laboratory of Paleomagnetism and Tectonic Reconstruction,Ministry of Natural Resources,Beijing 100081,China;Institute of Geology and Geophysics,Chinese Academy of Sciences,Beijing 100029,China;Heilongjiang Geological Science Institute,Harbin 150080,Heilongjiang,China;China Geological Survey,Beijing 100037,China)

机构地区：[1]中国地质科学院地质力学研究所,北京100081 [2]自然资源部古地磁与古构造重建重点实验室,北京100081 [3]中国科学院地质与地球物理研究所,北京100029 [4]黑龙江省地质科学研究所,黑龙江哈尔滨150080 [5]中国地质调查局,北京100037

出　　处：《地质通报》2022年第2期218-241,共24页Geological Bulletin of China

基　　金：中国地质调查局项目《特殊地质地貌区填图试点》(编号:DD20160060)和《地表过程与系统演变地质调查》(编号:DD20221644)

摘　　要：截至目前,中国已经完成的15万区域地质填图工作主要分布于基岩裸露地区,很少涉及覆盖区。为了满足和适应新时代国家经济建设对地质调查工作的需求,未来中国地质填图工作必须向覆盖区推进。所以,利用机器学习与数据挖掘技术,按照地质填图的要求对海量多源异构地质数据融合与综合分析,是实现覆盖区智能化地质填图的关键环节。以浅覆盖森林沼泽区为例,充分利用航空磁测、土壤地球化学等结构化数据和遥感影像、地表地质等非结构化数据,开展聚类分析与人机交互深度学习2种算法模型的对比试验。结果表明,单一数据的聚类分析无法进行有效的地质单元划分,而利用多源数据进行人机交互深度学习和训练所获得的预测模型结果图件经检验与实际地质单元基本一致。本次试验,充分利用了机器学习功能和特殊算法,实现了计算机代替地质人员进行地质填图的探索,为森林沼泽区地质填图工作中设计地质图、工作部署和成果总结提供了示范案例,为覆盖区智能地质填图提供了借鉴。Presently,the 150000 geological mapping completed in China is mainly distributed in the bedrock-outcropped area,rarely involving the coverage area which accounts for more than one-third of the national land area.To meet and adapt to the needs of national economic construction for a geological survey in the new era,the geological mapping must be extended to the covered area in the future.Making full use of big data,cloud computing,artificial intelligence and other technologies,the fusion and comprehensive analysis of massive multi-source heterogeneous geological data according to the requirements of geological mapping by computer data extraction is the key to realize the intelligent geological mapping in coverage areas.The experiment was carried out in the shallow covered forest-swamp region.The structured data such as aeromagnetic surveys,soil geochemistry,and non-structural data such as remote sensing image and surface geological surveys were fully utilized to carry out the comparative experiment on two algorithm models of cluster analysis and human-computer interaction deep learning.The results show that the clustering analysis based on single data cannot effectively divide the geological units,and the prediction model obtained by human-computer interactive deep learning and training with multi source data is basically consistent with the actual geological units.The machine learning function and special algorithm were used in this experiment and realized the geological mapping exploration by computer instead of geological personnel.It provides a demonstration case for primary geological map,work layout and achievement integration of the geological mapping in forest-swamp area,and offers a reference for the intelligent geological mapping in the coverage area.

关 键 词：智能地质调查 覆盖区 大数据 深度学习 聚类分析 森林沼泽区 地质调查工程 

分 类 号：P623[天文地球—地质矿产勘探]