1 School of Civil and Hydraulic Engineering, Huazhong University of Science and Technology, Wuhan 430074, China
2 National Innovation Center for Digital Construction Technology, Huazhong University of Science and Technology, Wuhan 430074, China
3 Wuhan Metro Group Co., Ltd., Wuhan 430070, China
Show less
文章历史+
Received
Published
Revised
2025-04-24
2026-04-28
2025-11-19
Online Date
2026-05-29
PDF (26263K)
摘要
为提升施工现场积水全天候自动排查水平,基于SegNet网络与VGG19卷积神经网络改进下采样网络,通过采用重要性感知损失函数(importance-aware loss function, IAL)作为交叉熵损失函数(cross-entropy loss function),提出了一种用于施工现场积水检测的VGG-SegNet语义分割模型,实现了对积水的像素级分割。该模型的总体类别交并比(overall intersection over union, OIoU)达到98%,F1分数达到96%,均显著优于SegNet模型。同时提出了目标像素尺寸法,通过构建透视投影映射关系,运用像素的位置面积权重计算现场积水面积。施工现场测试显示,积水的像素分割准确率平均值为97.47%,面积计算误差率(绝对)平均值为4.24%,验证了方法的有效性。VGG-SegNet模型在施工现场积水图像识别方面表现良好,在准确、快速智能化检测施工现场积水隐患具有较好的应用前景。
Abstract
To improve the all-weather automatic detection of waterlogging at construction sites, a VGG-SegNet semantic segmentation model based on the SegNet network was proposed. The downsampling network was improved by using the VGG19 convolutional neural network, and the importance-aware loss function (IAL) was adopted as the cross-entropy loss function to achieve pixel-level segmentation of waterlogging. The model’s overall intersection over union (OIoU) reaches 98%, and the F1 score reaches 96%, both significantly better than the SegNet model. A target pixel size method was proposed for calculating the surface area of waterlogged areas. It construed a perspective projection mapping relationship and uses pixel position and area weights. Construction site tests validated the proposed method, showing an average pixel segmentation accuracy of 97.47% for waterlogged areas and an average absolute error rate of 4.24% for area calculation. The VGG-SegNet model performs well in recognizing waterlogging images at construction sites. Therefore, it has great prospects for promotion and application in the accurate, rapid and intelligent detection of waterlogging at construction sites.
XU Shu-wang, YAO De-ning, YU Qun-zhou, LUO Han-bin, ZHU Zhang-qin, MA Zhi-yu, FENG Wen-jie.
Pixel-level Segmentation Method for Waterlogging on Construction Sites Based on VGG-SegNet[J].
Science Technology and Engineering, 2026, 26(12): 5220-5231 DOI:10.12404/j.issn.1671-1815.2502956
针对以上问题,提出施工现场积水像素级分割方法。基于SegNet网络[21],利用VGG19卷积神经网络[22]改进下采样网络,采用重要性感知损失函数(importance-aware loss function, IAL)[23]作为交叉熵损失函数,运用像素的位置面积权重计算积水面积,通过施工现场测试对所提方法进行验证,可以实现施工现场积水图像的精准分割和面积精确计算。VGG-SegNet模型在施工现场积水图像识别方面表现良好,可应用于施工现场积水准确、快速的智能化检测。
ZhangFeilian, XuWeijing, WangMengqin, et al. Risk assessment and trend research of railway tunnel construction safety[J]. Journal of Safety and Environment, 2024, 24(7): 2473-2482.
WangXuyang, HanYuzhen, HuaFucai, et al. Analysis of the influence relationship of shield tunnel defects based on ISM-MICMAC[J]. China Civil Engineering Journal, 2024, 57(S2): 60-66.
ZhouZhong, YanLongbin, ZhangJunjie, et al. Review and prospect of intelligentidentification of apparent diseases in highway tunnels based on deep learning[J]. China Civil Engineering Journal. 2022, 55(S2): 38-48.
LiMinghong, LiuHong. Detection technology of water accumulation zones in rail transit tunnels based on no target image[J]. Railway Engineering, 2022, 62(10): 166-169.
GongYangfeng, LinXiuwei, HuangJieling, et al. Study on distribution of urban waterlogging based on DEM and UAV orthophoto images[J]. Journal of Harbin University of Commerce(Natural Sciences Edition), 2023, 39(2): 169-177.
XueFengchang, LüXin, ChenXiaojuan. Urban flood monitoring method based on improved DeeplabV3+[J]. Science of Surveying and Mapping, 2023, 48(10): 216-224.
[13]
GengP, JiaM Y, RenX X. Tunnel lining water leakage image segmentation based on improved BlendMask[J]. Structural Health Monitoring, 2023, 22(2): 865-878.
[14]
ZhangC H, ChenX S, LiuP, et al. Automated detection and segmentation of tunnel defects and objects using YOLOv8-CM[J]. Tunnelling and Underground Space Technology, 2024, 150:105857.
XueYadong, JiaFei, GuoChunsheng, et al. Deep learning-based shield tunnel leakage mixed dataset construction and fine segmentation[J]. Journal of Basic Science and Engineering, 2023, 31(4): 1032-1042.
[17]
WangB X, HeN N, XuF, et al. Visual detection method of tunnel water leakage diseases based on feature enhancement learning[J]. Tunnelling and Underground Space Technology, 2024, 153: 106009.
ZhangZhijian, WuGuangsheng, LiJieyi, et al. A method for urban road waterlogging image recognition based on deep learning[J]. Journal of Tropical Meteorology, 2024, 40(6): 918-930.
JinXiaofei, GuoShuai, HuangQiong, et al. Amethod for rapid identification of waterlogging in flood-prone area based on YOLOv7[J]. China Water & Wastewater, 2024, 40(9): 123-128.
WangNanyang, ShenJianghai. Image segmentation method of rock core hole based on MSHAM-UNet[J]. Science Technology and Engineering, 2024, 24(24): 10362-10369.
WangHai, CaiBaixiang, CaiYingfeng, et al. Detection of water-covered and wet areas on road pavement based on semantic segmentation network[J]. Automotive Engineering, 2021, 43(4): 485-491.
[28]
XueY D, CaiX Y, ShadabfarM, et al. Deep learning-based automatic recognition of water leakage area in shield tunnel lining[J]. Tunnelling and Underground Space Technology, 2020, 104: 103524.
ZhangYulin, JiangChangbo, LongYuannan, et al. Dynamic monitoring of water area in Dongting Lake during flood season based on Sentinel-1A SAR data[J]. Remote Sensing Technology and Application, 2024, 39(3): 741-752.
[31]
ParkS, SongA H. Shoreline change analysis with deep learning semantic segmentation using remote sensing and GIS data[J]. KSCE Journal of Civil Engineering, 2024, 28(2): 928-938.
BaiGanggang, HouJingming, HanHao, et al. Intelligent monitoring method for road inundation based on deep learning[J]. Water Resources Protection, 2021, 37(5): 75-80.
LuanQinghua, QinZhiyu, WangDong, et al. Review on monitoring technology of urban road waterlogging after rainstorm and its application[J]. Water Resources Protection, 2022, 38(1): 106-116, 140.
[36]
KacholeS, HuangX Q, NaeiniF B, et al. Bimodal SegNet:fused instance segmentation using events and RGB frames[J]. Pattern Recognition, 2024, 149: 110215.
[37]
BeissenovaG, MadiyarovaA, AliyevaA, et al. Using pretrained VGG19 model and image segmentation for rice leaf disease classification[J]. International Jouranl of Advanced Computer Science and Applications, 2024, 15(8): 743-752.
[38]
XiangK T, WangK W, YangK L. Importance aware semantic segmentation with efficient pyramidal context network for navigational assistant systems[C]// IEEE Intelligent Transportation Systems Conference(ITSC). Auckland: IEEE, 2019: 3412-3418.