基于目标检测和图像分割的岩芯RQD自动生成算法  

Automatic computing algorithm for rock core RQD based on object detection and image segmentation

作　　者：张国华 姜晋云 吕国磊 胡俊杰 熊峰 廖一凡 郑洪[4] 李漪 ZHANG Guohua;JIANG Jinyun;LYU Guolei;HU Junjie;XIONG Feng;LIAO Yifan;ZHENG Hong;LI Yi(Faculty of Engineering,China University of Geosciences(Wuhan),Wuhan 430074,China;Baika(Shanghai)Management Consulting Co.,Ltd.,Shanghai 200120,China;Digital Management Department,Hubei Communications Planning and Design Institute Co.,Ltd.,Wuhan 430051,China;Department of Digital Intelligence,China Railway Siyuan Survey and Design Group Co.,Ltd.,Wuhan 430063,China;PowerChina Hubei Electric Engineering Co.,Ltd.,Wuhan 430040,China)

机构地区：[1]中国地质大学(武汉)工程学院,湖北武汉430074 [2]百咖(上海)管理咨询有限公司,上海200120 [3]湖北省交通规划设计院股份有限公司数字化管理部,湖北武汉430051 [4]中铁第四勘察设计院集团有限公司数智化事业部,湖北武汉430063 [5]湖北省电力规划设计研究院有限公司,湖北武汉430040

出　　处：《安全与环境工程》2025年第1期100-106,共7页Safety and Environmental Engineering

基　　金：国家重点研发计划项目(2021YFB2600402);国家自然科学基金项目(52209148)。

摘　　要：为了对岩体质量进行分类和评估,岩石质量指标(RQD)在地矿工程中被广泛应用。以往需要手动测量岩芯片段的长度并手动计算RQD,这一过程既费时又费力。随着计算机软、硬件的进步,基于计算机视觉的图像处理使自动获取RQD逐步成为现实。针对现有方法的局限性,提出了一种基于目标检测和图像分割的RQD自动生成算法,首先根据颜色、纹理等特征的相似性,采用SAM模型对图像中的岩芯进行检测,然后采用YOLOv8算法训练模型,通过提取岩芯片段的缝隙特征进行不同岩芯片段的分割。为了测试该算法的性能,选取了来自宜涪铁路五峰段的10段岩芯样本开展案例研究。结果表明:采用该算法获取的RQD的准确率与传统的手动方式相当,平均相对误差不超过5%;该算法的效率明显优于传统的手动方式,能够节约超过60%的时间成本。To classify and evaluate rock mass quality,the rock quality designation(RQD)is extensively utilized in geotechnical engineering.Traditionally,obtaining core RQD requires manual measurement of rock core segment lengths and manual calculation of RQD,a process that is both time-consuming and labor-intensive.With advancements in computer hardware and software,image processing based on computer vision has gradually made the automatic acquisition of RQD a reality.Addressing the limitations of existing methods,this paper proposes an RQD automatic generation algorithm based on object detection and image segmentation.This algorithm employs the SAM(segment anything model)to detect rock cores in images based on the similarity of features such as color and texture,followed by the training of a model using the YOLOv8 algorithm to extract fissure features of rock core segments for segmentation of different rock core segments.To test the performance of this algorithm,we conducted a case study using 10 rock core samples from the Wufeng section of the Yichang-Fuling Railway.The results indicate that the accuracy of RQD obtained by this algorithm is comparable to that by traditional manual methods,with an average relative error not exceeding 5%;in terms of efficiency,this algorithm significantly outperforms traditional manual methods,saving over 60%of time costs.

关 键 词：岩石质量指标(RQD) 钻孔岩芯 图像处理 深度学习 目标检测 图像分割 

分 类 号：X93[环境科学与工程—安全科学]