基于人工智能的录井岩屑荧光智能检测系统研制  

Design of AI-Based Detection System for Rock Cuttings Fluorescence of Logging

作　　者：陈现军 郭书生 廖高龙 董振国 付群超 CHEN Xianjun;GUO Shusheng;LIAO Gaolong;DONG Zhenguo;FU Qunchao(Hainan Branch,China France Bohai Geoservices Co.,Ltd.,Haikou,Hainan,570312,China;Hainan Branch,CNOOC China Limited,Haikou,Hainan,570312,China)

机构地区：[1]中法渤海地质服务有限公司海南分公司,海南海口570312 [2]中海石油(中国)有限公司海南分公司,海南海口570312

出　　处：《石油钻探技术》2024年第5期130-137,共8页Petroleum Drilling Techniques

基　　金：中海石油(中国)有限公司重大科技项目“海上深层/超深层油气勘探技术”(编号:KJGG2022-0405)部分研究内容。

摘　　要：针对当前荧光录井检测方法存在激活光源单一、定量评估精度较差及检测计算方法复杂等问题,研制了一种基于人工智能的录井岩屑荧光智能检测系统,以便快速检测出含油物质。针对不同岩屑样本的特性,可以根据岩屑类型和面积自由调节灯源波长,并配合工业相机对岩屑样本进行拍摄,采集易于深度学习算法检测的高清图像;使用嵌入于移动端的改进DeepLab v3+算法进行岩屑荧光检测,计算出荧光占比,并在移动设备屏幕上展示出计算结果和检测效果图。不同岩屑样本的测试结果表明,系统对岩屑荧光检测的平均交并比达到72.73%,能够在保证准确性与时效性的同时,实现对岩样中荧光区域的有效量化。基于改进DeepLab v3+算法的岩屑荧光智能监测系统解决了人工探测岩屑荧光过程中存在的不确定因素,能够满足荧光录井技术对岩屑荧光检测的现场应用要求。Existing fluorescent logging detection methods have problems such as single activation light source,limited quantitative evaluation accuracy,and complex detection calculation.To address these issues,an artificial intelligence(AI)-based detection system for rock cuttings fluorescence of logging was developed,so as to detect oily substances rapidly.In order to adapt to the characteristics of different cuttings samples,the lamp source wavelength could be freely adjusted according to the type and area of cuttings,and the cuttings samples could be shot with industrial cameras to collect high-definition images easily detected by deep learning algorithms.An improved DeepLab v3+algorithm embedded in a mobile device was used to detect cuttings fluorescence,calculate the fluorescence ratio,and display the results and detection images on the mobile device screen.Tests of various cuttings samples show that the system achieves an average intersection over union of 72.73%in detecting cuttings fluorescence,ensuring both accuracy and timeliness while quantifying fluorescent areas in the rock samples.The intelligent detection system for cuttings fluorescence based on the improved DeepLab v3+algorithm eliminates uncertainties present in manual fluorescence detection processes of cuttings and meets the practical needs of fluorescent logging technology for cuttings fluorescence detection.

关 键 词：荧光录井 岩屑 深度学习 图像分割 智能检测 

分 类 号：P618.13[天文地球—矿床学]