基于机器学习的找矿预测模型在湖南岳溪锑矿田的应用  

Application of Machine Learning Based Mineral Prospectivity Mapping in the Yuexi Antimony Orefield,Hunan Province

作　　者：秦耀祖 吴伟成 谢丽凤 欧鹏辉 黄小岚 QIN Yao-zu;WU Wei-cheng;XIE Li-feng;OU Peng-hui;HUANG Xiao-lan(Key Laboratory for Digital Land and Resources of Jiangxi Province, East China University of Technology, Nanchang 330013, China;School of Earth Science, East China University of Technology, Nanchang 330013,China)

机构地区：[1]东华理工大学江西省数字国土重点实验室,江西南昌330013 [2]东华理工大学地球科学学院,江西南昌330013

出　　处：《东华理工大学学报（自然科学版）》2021年第1期28-40,共13页Journal of East China University of Technology(Natural Science)

基　　金：东华理工大学博士科研启动基金项目(DHBK2019040);东华理工大学江西省数字国土重点实验室开放研究基金项目(DLLJ201901)。

摘　　要：分析各类地质找矿因子与矿体之间的相关性,建立合理高效的找矿预测模型,是优选找矿勘查靶区的关键。位于雪峰弧形构造带中段之金锑钨多金属成矿带上的岳溪矿田,找矿前景良好。以岳溪断裂(F1)南部缓冲区、从Landsat8-OLI影像数据中提取的铁染及羟基蚀变信息及从地球化学勘探获取的Sb等10种元素含量作为地质找矿因子,分别构建了基于随机森林和支持向量机的分类预测模型。通过分析认为,随机森林模型具有更为优异的预测功能,故选取该模型的预测结果进行成矿潜力制图。研究结果表明F1以南的缓冲区和地表的As、Sb含量,对矿体定位有较高的贡献;划分的成矿高潜力区,占研究区的7.69%,包含着98.60%的含矿单元,可优选为后续找矿勘查工作的靶区。Analyzing the spatial correlation of various geological prospecting factors with orebodies,and constructing efficient mineral prediction model,are very critical for identifying optimal targets for future mineral exploration.The Yuexi orefield located in the Au-Sb-W polymetallic ore belt of the mid-Xuefengshan tectonic arc is worthy of farther prospecting.In this study,the southern buffer zone of the Yuexi Fault(F1),the ferric-alteration and hydroxyl-alteration information that were extracted from Landsat8-OLI imagery,and ten elements content that extracted from geochemical exploration,were used as prospecting factors to construct Random Forest(RF)-and Support Vector Machine(SVM)based classification and prediction models,respectively.The prediction probability of RF based model was used to map mineral prospectivity thanks to its excellent prediction performance.The results show that the south buffer zone of the F1,as well as the As and Sb content in the surface,are closely related to orebody location;the created high prospectivity areas,which occupy 7.69%of the total study areas with containing 98.60%orebodies,can be determined as the target areas for future mineral exploration.

关 键 词：地质找矿因子 Landsat8-OLI影像 随机森林 支持向量机 雪峰弧形构造带 岳溪矿田 

分 类 号：P612[天文地球—矿床学] P628.1[天文地球—地质学]