深水大坝缺陷鲁棒实时检测方法

李扬涛, 包腾飞, 李田雨

李扬涛, 包腾飞, 李田雨. 深水大坝缺陷鲁棒实时检测方法[J]. 武汉大学学报 ( 信息科学版). DOI: 10.13203/j.whugis20220734
引用本文: 李扬涛, 包腾飞, 李田雨. 深水大坝缺陷鲁棒实时检测方法[J]. 武汉大学学报 ( 信息科学版). DOI: 10.13203/j.whugis20220734
Li Yangtao, Bao Tengfei, Li Tianyu. A robust real-time detection method for deepwater dam defects[J]. Geomatics and Information Science of Wuhan University. DOI: 10.13203/j.whugis20220734
Citation: Li Yangtao, Bao Tengfei, Li Tianyu. A robust real-time detection method for deepwater dam defects[J]. Geomatics and Information Science of Wuhan University. DOI: 10.13203/j.whugis20220734

深水大坝缺陷鲁棒实时检测方法

基金项目: 

中央高校基本科研业务费专项资金资助项目(B220203039)。

详细信息
    作者简介:

    李扬涛,博士。主要从事水工建筑物智慧运维和安全诊断方向的研究。邮箱:liyangtao@hhu.edu.cn

    通讯作者:

    包腾飞,博士,教授。邮箱:baotf@hhu.edu.cn

A robust real-time detection method for deepwater dam defects

  • 摘要: 大坝长期服役过程中,在水环境和外部荷载交互耦合作用下,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(Remotely Operated Vehicle,ROV)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,然而如何从这些海量图像视频数据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。基于此,本文结合机器视觉和深度学习理论方法,研究并提出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框架。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,恢复由于剪枝压缩参数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目标框架。以某高坝深水探测为工程实例,引入多种深度学习目标检测算法作为对比手段,验证方法在障碍物遮挡、低可见度、光照不均等复杂深水检测场景的缺陷效果。案例分析表明,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并量化其真实尺寸。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。
    Abstract: 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 (ROVs) 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: Based on this, this paper proposes a real-time multi-class defect automatic identification framework for dam underwater structures. Specifically, the single-stage object detection network YOLOv5s 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: Take the underwater detection of a high dam as an example. The effectiveness of the proposed method was validated in complex underwater scenes like obstacle occlusion, low visibility, and uneven illumination. Conclusions: Experimental results indicated that the proposed method can effectively overcome the interference of complicated underwater imaging environments and accurately identify different types of defects. Moreover, it also achieves the inference speed of processing 100 defect images per second, indicating its real-time detection capability.
  • [1] Cong Xu, Shao Wei Wang, Chong Shi Gu, et al. A Probabilistic Prediction Model for Displacement of Super High Arch Dams Considering the Deformation Spatial Association, 2021:15.(徐丛,王少伟,顾冲时,等.融合空间关联性的特高拱坝位移概率性预测模型[J].武汉大学学报(信息科学版). 2021:15)
    [2] JinBao Sheng, Yan Xiang, DeWei Yang, et al. Key technologies and applications of safety diagnosis and smart management of reservoir dams, 2022, 44(07):1351-1366.(盛金保,向衍,杨德玮,等.水库大坝安全诊断与智慧管理关键技术与应用[J].岩土工程学报. 2022, 44(07):1351-1366).
    [3] (向衍,盛金保,袁辉,等.中国水库大坝降等报废现状与退役评估研究[J].中国科学:技术科学. 2015, 45(12):1304-1310.)2015, 45(12):1304-1310.

    Yan Xiang, JinBao Sheng, Hui Yuan, et al. Research on degrading and decommissioning assessment of reservoir in China,2015, 45(12):1304-1310.

    [4] ZhongRu Wu, ChongShi Gu, HuaiZhiSu, et al. Review and prospect of calculation analysis methods in hydro-structure engineering,2015, 43(05):11.(吴中如,顾冲时,苏怀智,等.水工结构工程分析计算方法回眸与发展[J].河海大学学报(自然科学版). 2015, 43(05):11.)
    [5] LiYang Xiao, Wei Li, Bo Yuan, et al. Pavement Crack Automatic Identification Method Based on Improved Mask R-CNN Model,2022:16.(肖力炀,李伟,袁博,等.一种改进型Mask R-CNN模型的路面裂缝识别方法[J].武汉大学学报(信息科学版). 2022:16.)
    [6] Wen Gao Xie, Yi Xiao Zhang, Ai Rong Liu, et al. METHOD FOR CONCRETE SURFACE CRACKING DETECTION BASED ON ROV AND DIGITAL IMAGE TECHNOLOGY 2022, 39(S1):7.(谢文高,张怡孝,刘爱荣,等.基于水下机器人与数字图像技术的混凝土结构表面裂缝检测方法[J].工程力学. 2022, 39(S1):7.)
    [7] Kang Wei,Fang Yuan,ZhiQiang Dong, et al. Precision analysis and civil application of multi-camera digital image correlation method based on encoded targets2021, 51(02):8.(魏康,员方,董志强,等.基于标志点的多相机数字图像相关方法精度分析及土木工程中的应用[J].东南大学学报(自然科学版). 2021, 51(02):8.)
    [8] YuFei Liu, JianSheng Fan, JianGuoNie, et al. Review and prospect of digital-image-based crack detection of structure surface2021, 54(06):79-98.(刘宇飞,樊健生,聂建国,等.结构表面裂缝数字图像法识别研究综述与前景展望[J].土木工程学报. 2021, 54(06):79-98.)
    [9] YuFei Liu, Yu Qi, BaoLuo Li, et al. Intelligent detection of sensitive service indicators in multiple scenarios of structural maintenance, 2022, 43(10):1-15.(刘宇飞,齐玉,李保罗,等.工程结构运维多场景的敏感服役指标智能识别[J].建筑结构学报. 2022, 43(10):1-15.)
    [10] ChunLei Liu, Tian'en Chen, Cong Wang, et al., A Survey of Few-shot Object Detection, 2022:23.(刘春磊,陈天恩,王聪,等.小样本目标检测研究综述[J].计算机科学与探索. 2022:23.)
    [11]

    Xue Y, Li Y. A Fast Detection Method via Region-Based Fully Convolutional Neural Networks for Shield Tunnel Lining Defects[J]. Computer-Aided Civil and Infrastructure Engineering. 2018, 33(8):638-654.

    [12]

    Zhou Z, Zhang J, Gong C. Automatic detection method of tunnel lining multi-defects via an enhanced You Only Look Once network[J]. Computer-Aided Civil and Infrastructure Engineering. 2022, 37(6):762-780.

    [13] WenQing Shang, HongBo Qi. Identification algorithm of field weeds based on improved Faster R-CNN and transfer learning, 2022, 43(10):7.(尚文卿,齐红波.基于改进Faster R-CNN与迁移学习的农田杂草识别算法[J].中国农机化学报. 2022, 43(10):7.)
    [14] Wei Wang, XinYao Tang, Hua Cui, et al. Accurate Perception of Three-dimensional Vehicle Form in Roadside Monocular Perspective Based on CenterNet, 2022, 35(09):15.(王伟,唐心瑶,崔华,等.基于CenterNet的路侧单目视角车辆3D形态精确感知[J].中国公路学报. 2022, 35(09):15.)
    [15] 杨真真,郑艺欣,邵静,等.基于改进路径聚合和池化YOLOv4的目标检测[J].南京邮电大学学报(自然科学版). 2022, 42(05):7.

    ZhenZhen Yang, YiXin Zheng, Jing Shao, et al. 2022, 42(05):7.

    [16] JiShang Xu, CuiYin Liu, Ming Liu. An object detection method based on improved YOLOv4 path aggregation and pooling, 2022:8.(徐继尚,柳翠寅,刘明.改进YOLOV4-Tiny的SAR图像舰船小目标检测[J].小型微型计算机系统. 2022:8.)
    [17] KeYou Guo, Xue Li, Min Yang. Real-time detection method of traffic information based on lightweight YOLOv4,2022:9.(郭克友,李雪,杨民.基于轻量化YOLOv4的交通信息实时检测方法[J].计算机应用. 2022:9.)
    [18]

    Choi W, Cha Y. SDDNet:Real-Time Crack Segmentation[J]. IEEE Transactions on Industrial Electronics. 2020, 67(9):8016-8025.

计量
  • 文章访问数:  374
  • HTML全文浏览量:  16
  • PDF下载量:  42
  • 被引次数: 0
出版历程
  • 收稿日期:  2023-06-03
  • 网络出版日期:  2023-07-11

目录

    /

    返回文章
    返回