基于深度学习的页岩黄铁矿扫描电镜图像分割及环境指示意义:以四川盆地泸州Ⅰ区为例  

Deep Learning SEM Image Segmentation of Shale Pyrite and Environmental Indications:A Study of Luzhou Block,Sichuan Basin

作　　者：邓乃尔 徐浩[2] 周文[2] 唐小川 陈雨露 刘永旸 刘绍军 张益 蒋柯 刘瑞崟 宋威国 DENG Naier;XU Hao;ZHOU Wen;TANG Xiaochuan;CHEN Yulu;LIU Yongyang;LIU Shaojun;ZHANG Yi;JIANG Ke;LIU Ruiyin;SONG Weiguo(College of Energy(College of Modern Shale Gas Industry),Chengdu University of Technology,Chengdu 610059,China;State Key Laboratory of Oil and Gas Reservoir Geology and Exploitation,Chengdu University of Technology,Chengdu 610059,China;College of Computer Science and Cyber Security(Demonstrative Software College),Chengdu University of Technology,Chengdu 610059,China;Research Institute of Shale Gas,PetroChina Southwest Oil&Gasfield Company,Chengdu 610051,China;College of Electronics and Information Engineering,Sichuan University,Chengdu 610065,China;Guizhou Geological Survery,Guiyang 550081,China)

机构地区：[1]成都理工大学能源学院(页岩气现代产业学院),四川成都610059 [2]成都理工大学油气藏地质及开发工程全国重点实验室,四川成都610059 [3]成都理工大学计算机与网络安全学院(示范性软件学院),四川成都610059 [4]中国石油西南油气田公司页岩气研究院,四川成都610051 [5]四川大学电子信息学院,四川成都610065 [6]贵州省地质调查院,贵州贵阳550081

出　　处：《地球科学进展》2024年第5期476-488,共13页Advances in Earth Science

基　　金：国家自然科学基金项目(编号:42202189);四川省自然科学基金项目(编号:24NSFSC4997)资助。

摘　　要：黄铁矿作为页岩体系中最具代表性的重矿物之一,对其进行微观特征识别对于页岩沉积环境研究具有重要意义。以四川盆地泸州I区五峰组一龙一,亚段为例,通过岩心矿物实验、扫描电镜观测、网络模型优化和特征参数统计,构建了适用于黄铁矿扫描电镜图像分割的网络模型,实现了基于草莓状黄铁矿参数对研究区沉积环境的判断。结果表明:①优化后的UNet-Im模型对草莓状黄铁矿扫描电镜图像的分割精度可达0.863,证明了改进措施的优越性;②对比黄铁矿含量,龙一^(1-3)_(1)小层黄铁矿含量最高,为2.95%,随后降低至龙一^(4)_(1)小层的2.03%以及五峰组的0.83%;③基于草莓状黄铁矿特征参数,推断出黄铁矿沉积环境为深水硫化环境、深水强还原环境、深水强一弱还原环境以及深水还原一次氧化环境。实现了黄铁矿扫描电镜图像的精准化分割,对于提升行业勘探开发智能化具有借鉴意义。Pyrite,a significant heavy mineral in shale,aids in the comprehension of shale depositional environments.Referencing the Wufeng-Long1 subsection Formation of the Luzhou Block in Sichuan Basin,a network model for pyrite SEM image segmentation was established via core mineral experiments,SEM observations,network model refinement,and feature parameter analysis.The model assesses the sedimentary environment of the study block using pyrite framboid parameters.①Our findings indicated that enhancement of the UNet-Im model for pyrite framboid SEM images resulted in a segmentation precision of 0.863,demonstrating the effectiveness of the enhancement measures.②Pyrite content varied from 2.95%in the Long minor layer to 0.83%in the Wufeng Formation,with the Long minor layer at 2.03%.③Pyrite depositional environments are deduced as deep-water sulfide environments,strong reducing environments,strong-weak reduction environments,and reductive-suboxidative environments based on pyrite framboid characteristics.This study accurately segmented pyrite SEM images to enhance the exploration and development of intelligence in this industry.

关 键 词：黄铁矿 深度学习 沉积环境 泸州Ⅰ区 

分 类 号：P572[天文地球—矿物学]