Crack and Corrosion Detection with Few-Shot Data in Thermal Power Plants Based on Deep Learning

ZHANG Hui; LI Tonghao; FENG Fei; HU Mingyue; GUO Yulun; LUO Zhenbao; FU Zhitao

doi:10.3969/j.issn.1008-0198.2026.01.017

您当前的位置：

首页 >

文章列表页 >

Crack and Corrosion Detection with Few-Shot Data in Thermal Power Plants Based on Deep Learning

更新时间：2026-03-11

- Crack and Corrosion Detection with Few-Shot Data in Thermal Power Plants Based on Deep Learning
- Vol. 46, Issue 1, Pages: 125-133(2026)
- 作者机构：
  
  1. 国能长源汉川发电有限公司,湖北汉川,431614
  2. 西南技术物理研究所,四川,成都,610041
  3. 昆明理工大学国土资源工程学院,云南,昆明,650093
- 作者简介：
- 基金信息：
- DOI：10.3969/j.issn.1008-0198.2026.01.017
  CLC： TK28
- Online First：10 March 2026，
  
  Published：10 March 2026
- 稿件说明：
移动端阅览
张晖, 李瞳昊, 冯飞, 胡明玥, 郭羽纶, 罗镇宝, 付志涛. 基于深度学习的火电厂小样本裂缝腐蚀检测[J]. 湖南电力, 2026, 46(1): 125-133.

ZHANG Hui, LI Tonghao, FENG Fei, et al. Crack and Corrosion Detection with Few-Shot Data in Thermal Power Plants Based on Deep Learning[J]. 2026, 46(1): 125-133.
张晖, 李瞳昊, 冯飞, 胡明玥, 郭羽纶, 罗镇宝, 付志涛. 基于深度学习的火电厂小样本裂缝腐蚀检测[J]. 湖南电力, 2026, 46(1): 125-133. DOI： 10.3969/j.issn.1008-0198.2026.01.017.

ZHANG Hui, LI Tonghao, FENG Fei, et al. Crack and Corrosion Detection with Few-Shot Data in Thermal Power Plants Based on Deep Learning[J]. 2026, 46(1): 125-133. DOI： 10.3969/j.issn.1008-0198.2026.01.017.

摘要

针对火电厂内部烟囱、墙壁、金属设备等关键部位在复杂光照条件下存在裂缝图像标注数据稀缺、特征提取困难的问题

提出一种基于小样本学习的双分支原型度量网络的裂缝检测模型PowerCrackNet

该网络利用Retinex理论整合图像中与光照无关的反射率信息

引导网络学习光照不变性。同时在特征提取阶段通过度量高维空间中查询和支持特征之间的相似度

生成包含裂缝位置和结构信息的先验掩码。然后在特征交互阶段采用卷积注意力和空洞空间卷积池化金字塔机制

融合来自不同尺度的特征信息和先验信息

以改善小样本学习中的样本间空间不一致问题。最后在原型度量阶段利用互补注意力机制

增强支持原型向量对裂缝类别的表达能力。该方法在公开的真实低照度裂缝数据集LCSD上同现有其他方法进行了量化对比实验与可视化结果分析

结果表明在样本数量受限（1-shot）的情况下

本文方法对低照度裂缝图像具有更强的裂缝分割能力。同时

在实际场景的验证进一步表明

该方法可为火电厂关键部位裂缝的早期预警提供高效可靠的技术支持

具有较强的工程应用价值。

Abstract

For the problem that there is scarce image annotation data and difficult feature extraction in the cracks of key parts such as chimneys

walls and metal equipment in thermal power plants under complex lighting conditions

a crack detection model named PowerCrackNet is proposed

which is based on a dual-branch prototype metric network with few-shot learning. The network integrates the reflection rate information irrelevant to lighting in the image using the Retinex theory

guiding the network to learn the invariance of lighting. During the feature extraction stage

a prior mask containing crack location and structural information is generated by measuring the similarity between query and support features in a high-dimensional space. Then

in the feature interaction stage

the convolution attention and atrous spatial convolution pooling pyramid mechanism are employed to integrate the feature information and prior information from different scales

thereby improving the spatial inconsistencies among samples in few-shot learning. Finally

in the prototype metric stage

a complementary attention mechanism is used to enhance the expressiveness of support prototype vectors for crack categories. Quantitative comparisons and visualization analysis on the public low-illumination crack dataset LCSD demonstrate that the proposed method achieves stronger crack segmentation performance under low-light conditions with limited samples(1-shot)

outperforming the current other method. Further validation in real-world scenarios confirms that the proposed approach can provide efficient and reliable technical support for the early warning of cracks in key parts of thermal power plants

highlighting its significant practical value.

关键词

Keywords

references

张大厚. 火力发电厂烟囱防腐存在的问题及建议[J]. 高科技与产业化,2009,15(1):88-92.

王志,朱传刚,蔡连成,等. 特高压架空输电线路混凝土基础裂缝注浆加固措施研究[J]. 湖北电力,2023,47(5):69-76.

余建飞,吴军,周学杰,等. 湖北电网金属材料土壤腐蚀等级分布图绘制[J]. 湖北电力,2024,48(2):140-145.

KIM H,SIM S H,SPENCER B F.Automated concrete crack evaluation using stereo vision with two different focal lengths[J]. Automation in Construction,2022,135:104136.

LIU Y F,NIE X,FAN J S,et al.Image-based crack assessment of bridge piers using unmanned aerial vehicles and three-dimensional scene reconstruction[J]. Computer-Aided Civil and Infrastructure Engineering,2020,35(5):511-529.

罗畅,杨学毅. 不同燃烧方式下燃煤锅炉水冷壁高温腐蚀问题探讨[J]. 湖北电力,2021,45(5):120-126.

冯红武,梅云平,田泽,等. 某电厂Super304H钢锅炉高温再热器爆管分析[J]. 湖北电力,2021,45(6):106-111.

冯治. 火电厂锅炉受热面爆管问题分析及对策[J]. 陕西电力,2006,34(5):31-34.

SHABAN A,BANSAL S,LIU Z,et al. One-shot learning for semantic segmentation[C]//Proceedings ofthe British Machine Vision Conference2017. London,UK.British Machine Vision Association,2017:167.1-167.13.

KULKARNI S,SINGH S,BALAKRISHNAN D,et al.CrackSeg9k:a collection and benchmark for crack segmentation datasets and frameworks[C]//Computer Vision-ECCV 2022 Workshops. Cham:Springer,2023:179-195.

KEERTHANA B,DEVI J A,MADHURI S,et al.Building crack detection using CNN and class activation mapping[C]//2024 International Conference on Artificial Intelligence and Quantum Computation-Based Sensor Application(ICAIQSA). Nagpur,India. IEEE,2025:1-6.

DONG N,XING E P.Few-shot semantic segmentation with prototype learning[C]//Proceedings of the British Machine Vision Conference. 2018,3:4.

陈善娟,于云龙,李英明. 小样本语义分割研究现状与分析[J]. 计算机学报,2024,47(10):2417-2451.

YAO Z,XU J,HOU S,et al.CrackNex:a few-shot low-light crack segmentation model based on retinex theory for UAV inspections[C]//2024 IEEE International Conference on Robotics and Automation(ICRA). Yokohama,Japan. IEEE,2024:11155-11162.

CAO P,ZHANG C,HUANG J,et al.Cross-light contrastive learning of pavement crack segmentation in low-light conditions for autonomous road inspection vehicles[J]. IEEE Transactions on Intelligent Vehicles,2024:1-11.

YANG L,ZHUO W,QI L,et al.Mining latent classes for few-shot segmentation[C]//2021 IEEE/CVF International Conference on Computer Vision(ICCV). Montreal,QC,Canada. IEEE,2022:8701-8710.

FAN Q,PEI W,TAI Y W,et al.Self-support few-shot semantic segmentation[C]//Computer Vision-ECCV 2022. Cham:Springer,2022:701-719.

渠延模,顾盛,刘从喜,等. 某烟囱裂缝及腐蚀检测鉴定与处理[J]. 工业建筑,2017(增刊3):708-711.

张艳飞,韩钢城,谢利明,等. 火电厂锅炉再热器管道开裂原因分析[J]. 内蒙古电力技术,2021,39(3):38-42.

王志永,牛晓光,李树军,等. 火电厂钢结构常见缺陷及可能性因素分析[J]. 河北电力技术,2009,28(3):47-48.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Multi-Factor Short-Term Power Load Forecasting Based on Adversarial Training

Infrared Image Segmentation Method of Lightning Arrester Casing Based on Deep Learning and Local Perception Attention Mechanism

Insulator Pollution Monitoring Based on Acoustic Signal and One-Dimensional Convolutional Neural Network

Detection Method of Metal Fitting Rust Defects for Overhead Transmission Lines Based on UAV Patrol of Deep Learning

Point Cloud Semantic Segmentation of Distribution Net‍work Tower Based on Self-Attention Perception

Related Author

SHANG Longkang

HE Lubing

LI Xiaoping

YE Jinxiang

CHEN Haibo

WANG Shengqi

WU Lin

ZHANG Lei

Related Institution

Xuchang Xuji Software Technology Co., Ltd.

College of Electrical Engineering, Xuchang Electrical Vocational College

State Grid Zhejiang Electric Power Company Limited Ultra High Voltage Branch

State Grid Hubei Electric Power Company Limited Technology Training Center

China Three Gorges University

AI问答

⁰