深水大坝缺陷鲁棒实时检测方法  

A Robust Real-Time Detection Method for Deepwater Dam Defects

作　　者：李扬涛 包腾飞 李田雨[1,2,3] LI Yangtao;BAO Tengfei;LI Tianyu(School of Civil Engineering,Nanjing Forestry University,Nanjing 210042,China;College of Water Resources and Hydropower Engineering,Hohai University,Nanjing 210098,China;State Key Laboratory of Water Disaster Prevention,Hohai University,Nanjing 210098,China)

机构地区：[1]南京林业大学土木工程学院,江苏南京210042 [2]河海大学水利水电学院,江苏南京210098 [3]河海大学水灾害抵御国家重点实验室,江苏南京210098

出　　处：《武汉大学学报(信息科学版)》2025年第4期684-698,共15页Geomatics and Information Science of Wuhan University

基　　金：西藏自治区科技计划(XZ202501ZY0009)。

摘　　要：大坝长期服役过程中,在水环境和外部荷载的交互耦合作用下,其深水结构部位易出现各类缺陷病害,影响工程服役安全稳定和功能发挥。水下机器人搭载可见光相机,可以非接触形式实现结构损伤的高分辨率空间信息采集,如何从海量图像视频数据中提取结构损伤密切相关信息成为当前亟待解决的关键问题。结合机器视觉和深度学习理论方法,提出了一种兼顾检测精度和推理效率的大坝深水多类别缺陷实时目标检测方法。该框架以单阶段目标检测网络YOLOv5-s为基模型,构建大坝多类别缺陷识别器;利用模型稀疏化和剪枝策略,改变模型批处理层权重分布并去除模型冗余参数;进一步地,综合运用模型迁移和知识蒸馏理论,恢复由于剪枝压缩带来的精度劣化问题,据此构建出强背景干扰下大坝深水多类别缺陷实时检测方法。以某高坝深水探测工程为实例,引入多种深度学习目标检测算法作为对比,验证所提方法在障碍物遮挡、低可见度、光照不均等复杂深水检测场景的效果。案例分析结果表明,该方法可有效克服多种水下不利成像环境干扰,并准确辨识、区分不同类型缺陷,量化其真实尺寸。此外,剪枝后轻量化模型每秒可推理超过100张缺陷图像,具备较强的实时推理能力。Objectives:Under the coupling action of environment and loads,dam underwater structures suffer from defects,affecting the safety,stability,and functional performance of the project.Underwater robots like remotely operated vehicles(ROV)equipped with visible light cameras can realize the high-resolution spatial information in a non-contact form for underwater damage.However,it is still a challenging task that needs to be solved urgently to efficiently extract effective information from massive image and video data.Methods:First,we propose a real-time multi-class defect automatic identification framework for dam underwater structures.Specifically,the single-stage object detection network YOLOv5-s is utilized as the base model to develop the damage detector.Then,the model sparsity and pruning strategies are combined to change the batch layer weight distribution and remove model redundant parameters.Next,transfer learning and knowledge distillation are combined to recover the accuracy degradation caused by model pruning and compression.Results:As an example,the underwater detection of a high dam is considered.The effectiveness of the proposed method is validated in complex underwater scenes like obstacle occlusion,low visibility,and uneven illumination.Conclusions:The experimental results indicate that the proposed method can effectively overcome the interference of complicated underwater imaging environments and accurately identify different types of defects.Moreover,the proposed method achieves an inference speed of processing 100 defect images per second,demonstrating its real-time detection capability.

关 键 词：深水检测 隐蔽病害识别 损伤识别 目标检测 人工智能 

分 类 号：P258[天文地球—测绘科学与技术]