基于YOLO v5深度网络的隧道衬砌质量评估  

Quality evaluation of tunnel lining based on YOLO v5 deep network

作　　者：杨绪祥 耿元玲 左融 王涵予 Yang Xuxiang;Geng Yuanling;Zuo Rong;Wang Hanyu(Yunnan Research Institute of Highway Science and Technology,Kunming Yunnan 650051,China;Faculty of Civil Aviation and Aeronautics,Kunming University of Science and Technology,Kunming Yunnan 650500,China)

机构地区：[1]云南省公路科学技术研究院,云南昆明650051 [2]昆明理工大学民航与航空学院,云南昆明650500

出　　处：《山西建筑》2024年第17期6-11,共6页Shanxi Architecture

基　　金：云南省交通厅科技创新及示范项目(项目编号:2022-23-1)。

摘　　要：雷达无损检测技术已成为隧道衬砌状态评估特别是在复杂地质条件下非可视区域结构安全性检测方面的重要手段,而当前隧道衬砌检测面临人工分析效率低、精度受限及人力成本高昂等难题。针对此问题引入深度学习技术,通过构建针对性的深度学习框架,实现自动化处理雷达无损检测数据,克服传统检测手段瓶颈,提升检测速度与准确性,减少人为因素影响。采用真实隧道检测数据集进行深度学习模型训练,实验结果表明,深度学习驱动的模型能高效识别各类衬砌缺陷,从微小裂纹到严重结构损伤,同时显著降低误报与漏检率,体现了相较于传统技术的明显优势。该成果为隧道结构安全评估提供了更为精确和高效的工具。Radar non-destructive testing technology has become an important means for evaluating the state of tunnel lining,especially for structural safety testing in non-visible areas under complex geological conditions.However,current tunnel lining testing faces challenges such as low efficiency of manual analysis,limited accuracy,and high labor costs.This article introduces deep learning technology to address this issue and achieves automated processing of radar non-destructive testing data by constructing a targeted deep learning framework.Overcoming the bottleneck of traditional detection methods,improving detection speed and accuracy,and reducing the influence of human factors.This article uses real tunnel detection datasets for deep learning model training.The experimental results show that the deep learning driven model in this article can efficiently identify various types of lining defects,from small cracks to severe structural damage,while significantly reducing false alarms and missed detections,reflecting significant advantages compared to traditional technologies.This achievement provides a more accurate and efficient tool for tunnel safety assessment.

关 键 词：雷达无损检测 深度学习 隧道衬砌 缺陷识别 地质缺陷 

分 类 号：TU997[建筑科学—市政工程] U455.91[建筑科学—桥梁与隧道工程]