基于深度学习SSD算法的高密度电法智能解译方法技术研究  

High Density Electrical Apparent Resistivity Anomaly Intelligent Interpretation Method Based on SSD Target Detection Algorithm

作　　者：师学明[1] 黄崇钰 王瑞 李斌才 郑洪[2] Shi Xueming;Huang Chongyu;Wang Rui;Li Bincai;Zheng Hong(School of Geophysics and Geomatics,China University of Geosciences,Wuhan Hubei 430074,China;China Railway Siyuan Survey and Design Group Co.,Ltd.,Wuhan Hubei 430063,China)

机构地区：[1]中国地质大学地球物理与空间信息学院,湖北武汉430074 [2]中铁第四勘察设计院集团有限公司,湖北武汉430063

出　　处：《工程地球物理学报》2024年第1期1-11,共11页Chinese Journal of Engineering Geophysics

基　　金：国家重点研发计划项目(编号:2021YFB2600402)。

摘　　要：高密度电法在探测灰岩区地下溶洞病害体方面得到广泛应用,但高密度电法反演结果依赖于初始模型,存在多解性,地质解译容易受专业人员主观因素影响。为此,本文从具有唯一性的视电阻率数据出发,研究了基于深度学习的SSD(Single Shot Multi-box Detector)目标检测算法的视电阻率异常智能解译方法技术。针对岩溶地质病害,设计了不同填充类型、形状、规模、数量的溶洞电性异常模型,利用Res2dmod软件进行视电阻率正演计算,构建了包含1400个样本的高密度电法视电阻率智能解译学习样本库(样本和标签)。基于TensorFlow框架,建立了基于深度学习SSD算法的高密度电法视电阻率异常智能解译方法技术,使用学习样本库训练网络权值,训练结束后对高密电法温纳装置视电阻率异常进行智能解译,单个视电阻率剖面异常智能解译耗时不到1 s,各类目标(填充型溶洞、未填充型溶洞)平均准确率为90.68%。研究结果表明:基于SSD算法的高密度电法视电阻率异常智能解译技术可显著提高高密度电法视电阻率解译效率,避免专业人员主观因素影响。The high-density electrical method has been widely used in detecting underground karst cave disease bodies in limestone areas,but the inversion results of the high-density electrical method rely heavily on the initial model and have multiple solutions.Geological interpretation is easily affected by subjective factors of professionals.Therefore,this article studies the intelligent interpretation method technology of apparent resistivity anomalies based on the deep learning SSD(single shot multi-box detector)object detection algorithm,starting from unique apparent resistivity data.In response to karst geological disasters,electrical anomaly models of karst caves with different filling types,shapes,scales and quantities were designed.Res2dmod software was used for forward calculation of apparent resistivity,and a high-density electrical method-based intelligent interpretation and learning sample library(samples and labels)containing 1400 samples was constructed.Based on the TensorFlow framework and deep learning SSD algorithm,a high-density electrical resistivity anomaly intelligent interpretation technology was established,where the SSD algorithm weights were trained using a learning sample library,and intelligent interpretation numerical experiments were conducted on the apparent resistivity anomaly of the high-density electrical resistivity Wenner device.The learning sample library was trained and tested,with less than 1 second intelligent interpretation time for a single apparent resistivity profile anomaly,and the average accuracy of various targets(filled and unfilled karst caves)at 90.68%,and the error in interpreting the scale and location of karst caves at centimeter level.The results show that the intelligent interpretation technology of high-density electrical resistivity anomalies based on SSD algorithm significantly improves the efficiency of high-density electrical resistivity interpretation,while avoiding the interference of subjective factors by professionals on the interpretation results.The categor

关 键 词：高密度电法 温纳装置 视电阻率 SSD目标检测算法 智能解译 

分 类 号：P631.3[天文地球—地质矿产勘探]