集成边缘门控与多尺度空间注意的排水管道缺陷分割模型

doi:10.13474/j.cnki.11-2246.2025.0617

测绘通报 ›› 2025, Vol. 0 ›› Issue (6): 97-102.doi: 10.13474/j.cnki.11-2246.2025.0617

• 学术研究 • 上一篇

集成边缘门控与多尺度空间注意的排水管道缺陷分割模型

陈登峰¹, 赵航辉¹, 刘世鹏², 孟屯良³, 王泽鹏¹

1. 西安建筑科技大学建筑设备科学与工程学院, 陕西西安 710055;
2. 西安建筑科技大学机电工程学院, 陕西西安 710055;
3. 西安建筑科大工程技术有限公司, 陕西西安 710055

收稿日期:2024-11-22 发布日期:2025-07-04
作者简介:陈登峰(1976—),男,博士,副教授,主要从事建筑机器人与智能信息处理方面的研究。E-mail:chdengf@163.com
基金资助:
陕西省自然科学基金面上项目(2024JC-YBMS-286);西安市科技计划项目(2023JH-GXRC-0216;2024JH-KGDW-0112)

An integrated edge-gated and multi-scale spatial attention model for drainage pipeline defect segmentation

CHEN Dengfeng¹, ZHAO Hanghui¹, LIU Shipeng², MENG Tunliang³, WANG Zepeng¹

1. School of Building Services Science and Engineering, Xi'an University of Architecture and Technology, Xi'an 710055, China;
2. School of Mechanical and Electrical Engineering, Xi'an University of Architecture and Technology, Xi'an 710055, China;
3. Xi'an Construction Technology University Engineering Technology Co., Ltd., Xi'an 710055, China

Received:2024-11-22 Published:2025-07-04

摘要/Abstract

摘要： 城镇地下排水管道老化问题导致管道破损、堵塞等现象频发。人工检测方法难以满足不断增长的检测需求,现有智能化检测技术在应对形状不规则、细节丰富的缺陷边界时,其交并比仍存在提升空间。本文提出了一种提升缺陷边界识别性能的排水管道语义分割模型PGGNet,模型中的GEA通过拉普拉斯边缘检测算法整合边缘特征,提高了对缺陷边界的捕捉能力;MGSA-SSM则结合状态空间模型和MGSA的多尺度机制,引导模型从不同尺度捕捉缺陷的全局轮廓与局部细节,从而提升复杂边界的识别能力。试验结果表明,PGGNet在与主流算法对比中表现出色,mPA达94.32%,mIoU达93.08%,可满足排水管道自动化缺陷检测需求。

关键词: 城镇地下排水管道, 缺陷检测, PGGNet, 门控机制, 多尺度引导空间注意模块

Abstract: The aging of urban underground drainage pipelines has led to frequent occurrences of pipe damage, blockages, and other issues. Manual inspection methods struggle to meet the growing demand for detection, while existing intelligent detection technologies still have room for improvement in achieving higher intersection over union(IoU)when addressing irregularly shaped defects and intricate boundaries. This study proposes a semantic segmentation model for drainage pipelines named PGGNet to enhance defect boundary recognition performance. The gated edge attention model(GEA) in PGGNet integrates edge features through the Laplacian edge detection algorithm, significantly improving the ability to capture defect boundaries. Meanwhile, the multi-scale guided spatial attention-state space model(MGSA-SSM)combines a state-space model with MGSA of multi-scale mechanisms to guide the model in capturing both global contours and local details of defects across different scales, thereby enhancing the recognition capability for complex boundaries. Experimental results demonstrate that PGGNet outperforms mainstream algorithms, achieving an mPA of 94.32% and an mIoU of 93.08%, which meets the requirements for automated defect detection in drainage pipelines.

Key words: urban underground drainage pipelines, defect detection, PGGNet, gated mechanism, multi-scale guided spatial attention model

中图分类号:

P258

陈登峰, 赵航辉, 刘世鹏, 孟屯良, 王泽鹏. 集成边缘门控与多尺度空间注意的排水管道缺陷分割模型[J]. 测绘通报, 2025, 0(6): 97-102.

CHEN Dengfeng, ZHAO Hanghui, LIU Shipeng, MENG Tunliang, WANG Zepeng. An integrated edge-gated and multi-scale spatial attention model for drainage pipeline defect segmentation[J]. Bulletin of Surveying and Mapping, 2025, 0(6): 97-102.

参考文献

[1] HAURUM J B,MOESLUND T B.A survey on image-based automation of CCTV and SSET sewer inspections[J].Automation in Construction,2020,111:103061.
[2] 刘春明.城市道路塌陷成因分析与对策[J].城市勘测,2018(S1):184-187.
[3] XU Zuxin,XU Jin,YIN Hailong,et al.Urban river pollution control in developing countries[J].Nature Sustainability,2019,2:158-160.
[4] KUMAR S S,ABRAHAM D M,JAHANSHAHI M R,et al.Automated defect classification in sewer closed circuit television inspections using deep convolutional neural networks[J].Automation in Construction,2018,91:273-283.
[5] 王俊岭,邓玉莲,李英,等.排水管道检测与缺陷识别技术综述[J].科学技术与工程,2020,20(33):13520-13528.
[6] 许铁林,廖立国,周昌林.基于深度学习的复杂城市排水管道缺陷智能识别系统[J].测绘通报,2024(S2):37-41.
[7] 苏常旺,胡少伟,张海丰,等.基于改进YOLOX算法的给水管道内缺陷智能识别与定位[J].测绘通报,2023(12):70-75.
[8] YUSSUF A N,WEERASINGHE N P,CHEN Haosen,et al.Leveraging deep learning techniques for condition assessment of stormwater pipe network[J].Journal of Civil Structural Health Monitoring,2025,15(2):619-633.
[9] 王守志,刘章,王冬,等.基于PP-YOLOE的城市排水管网缺陷检测及应用[J].中国给水排水,2024,40(18):130-136.
[10] 周倩倩,刘汉林,陈维锋,等.基于DeepLabV3+的排水管道缺陷检测与语义分割[J].中国给水排水,2022,38(13):22-27.
[11] PAN Gang,ZHENG Yaoxian,GUO Shuai,et al.Automatic sewer pipe defect semantic segmentation based on improved U-Net[J].Automation in Construction,2020,119:103383.
[12] LI Mingze,LI Mingchao,REN Qiubing,et al.PipeTransUNet:CNN and Transformer fusion network for semantic segmentation and severity quantification of multiple sewer pipe defects[J].Applied Soft Computing,2024,159:111673.
[13] MATTSSON J,HEDSTRÖM A,VIKLANDER M,et al.Fat,oil,and grease accumulation in sewer systems:comprehensive survey of experiences of Scandinavian municipalities[J].Journal of Environmental Engineering,2014,140(3):04014003.
[14] HAURUM J B,MADADI M,ESCALERA S,et al.Multi-scale hybrid vision transformer and Sinkhorn tokenizer for sewer defect classification[J].Automation in Construction,2022,144:104614.
[15] DANG L M,WANG Hanxiang,LI Yanfen,et al.DefectTR:end-to-end defect detection for sewage networks using a transformer[J].Construction and Building Materials,2022,325:126584.
[16] ZHU Lianghui,LIAO Bencheng,ZHANG Qian,et al.Vision mamba:efficient visual representation learning with bidirectional state space model[EB/OL].[2024-10-20].https://arxiv.org/abs/2401.09417v3.
[17] BURT P J,ADELSON E H.The Laplacian pyramid as a compact image code[M]//Readings in Computer Vision.Amsterdam:Elsevier,1987:671-679.
[18] HAURUM J B,MOESLUND T B.Sewer-ML:a multi-label sewer defect classification dataset and benchmark[C]//Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Nashville,TN:IEEE,2021:13456-13467.

集成边缘门控与多尺度空间注意的排水管道缺陷分割模型

An integrated edge-gated and multi-scale spatial attention model for drainage pipeline defect segmentation

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 1

编辑推荐

Metrics

本文评价