基于深度学习的盾构隧道渗漏水病害混合样本集构建与精细分割  

Deep Learning-Based Shield Tunnel Leakage Mixed Dataset Construction and Fine Segmentation

作　　者：薛亚东[1,2] 贾非 郭春生 郭永发[4] 刘劼 XUE Yadong;JIA Fei;GUO Chunsheng;GUO Yongfa;LIU Jie(Key Laboratory of Geotechnical and Underground Engineering of Education Ministry,Tongji University,Shanghai 200092,China;College of Civil Engineering,Tongji University,Shanghai 200092,China;SGIDI Engineering Consulting(Group)Co.,Ltd.,Shanghai 200092,China;China Railway Eryuan Engineering Group Co.,Ltd.,Kunming 650000,China;China Railway Kunming Bureau Group Co.,Ltd.,Kunming 650000,China)

机构地区：[1]同济大学岩土及地下工程教育部重点实验室,上海200092 [2]同济大学土木工程学院,上海200092 [3]上海勘察设计研究院(集团)有限公司,上海200092 [4]中铁二院昆明勘察设计研究院有限责任公司,云南昆明650000 [5]中国铁路昆明局集团有限公司,云南昆明650000

出　　处：《应用基础与工程科学学报》2023年第4期1032-1042,共11页Journal of Basic Science and Engineering

基　　金：国家自然科学基金项目(52078377);上海市科学技术委员会资助项目(18DZ1205902);云南省科技厅重点科技研发计划项目(202002AC080002)。

摘　　要：渗漏水病害是盾构隧道运营期间最为常见的一种表观病害,对隧道结构安全与周边地层稳定具有不利影响.基于深度学习的图像病害识别方法,构建了包含检测装置与人工巡检两种方式采集图像的混合样本集.以平均准确度为评估指标,训练得到Mask R-CNN深度学习模型的分割准确度达到0.447,优于原样本集(0.386)与扩容样本集(0.403).考虑隧道渗漏水病害形态复杂的特点以及不同病害间较大的特征差异,进一步采用条件卷积动态生成的分割模型参数代替Mask R-CNN模型中静态的模型参数,提高了模型的分割速度与精度.以每秒运算图像数量(Frames Per Second,FPS)为评估指标,模型分割速度由7FPS提升至10FPS,且分割结果与病害真实轮廓更为接近,从而有利于对渗漏水病害的严重程度进行量化分析.The leakage defect is one of the most common surface defects during the operation of shield tunnels,which has adverse effect on the tunnel structure safety and surrounding ground stability.According to the deep learning-based image defect detection methods,this study constructed a mixed dataset,including the images collected by both detection device and manual inspection.Taking the average precision(AP)as evaluation metric,the trained Mask R-CNN achieves better recognition performance(0.447)than the original(0.386)and expanded(0.403)datasets.Considering the complex characteristics of leakage defect and the large difference between different defects,the model parameters dynamically generated by conditional convolution were further adopted to replace the static model parameters in Mask R-CNN,and improve the model segmentation speed and accuracy.With frames per second(FPS)as the evaluation metric,the model segmentation speed is increased from 7 to 10FPS,and the segmentation results are closer to the real defects contour,which is beneficial to realize quantitative analysis of the leakage defects severity.

关 键 词：盾构隧道 渗漏水 深度学习 卷积神经网络 实例分割模型 条件卷积 病害检测 

分 类 号：TU91[建筑科学—建筑理论]