摘要
建筑外墙空鼓与脱落的识别对于确保城市老旧建筑物周围公共安全至关重要。传统的人工原位检测方法需要耗费大量人力物力且存在一定的安全风险,此外识别结果也会受到专业人员的工作经验和工作状态等主观因素的影响。近年来,采用无人机进行图像采集并通过人工智能模型对建筑外墙缺陷进行识别的方法逐渐流行开来。然而,目前对于缺陷检测的研究仅针对单一模态的可见光图像或者红外图像,往往只能对某一缺陷进行检测,且没有考虑缺陷之间的转换关系。针对这一问题,通过融合建筑外墙的可见光和红外图像,结合两种模态的图像信息,并通过不同深度的UNet和Res-UNet模型对融合后图像进行建筑外墙缺陷识别,比较了不同深度模型的识别效果。试验结果表明,深度为4的Res-UNet模型对建筑外墙的空鼓和脱落具有很好的识别效果。
The detection of hollowing and missing of building exterior walls is crucial to ensure the public safety around aging buildings in cities.The traditional artificial in-situ detection methods are time-and labor-consuming with safety risks.In addition,the detection results will also be affected by subjective factors such as professional experience and working status.The method of image acquisition by UAV and detection of building exterior wall defects by artificial intelligence model has become popular.However,the current research on defect detection only focuses on visible images or infrared images of a single modality,and only detect a certain defect without considering the mutual conversion between defects.To address this issue,this research combined the visible and infrared images of the building exterior wall,considered the image information from two modalities,and compared the UNet and Res-UNet models of different depths to identify the building exterior wall defects in the fused images.The experimental results showed that the Res-UNet model with a depth of 4 performed excellent on the hollowing and missing of the building exterior wall.
作者
王玮
米庆仁
肖云
杨新聪
WANG Wei;MI Qingren;XIAO Yun;YANG Xincong(Guangzhou Metro Group Co.,Ltd.,Guangzhou 510220,China;Harbin Institute of Technology(Shenzhen),Shenzhen 518055,China;CRCC South China Construction Co.,Ltd,Guangzhou 510220,China)
出处
《工业建筑》
2024年第5期51-59,共9页
Industrial Construction
基金
国家重点研发计划(2022YFC3801203)
国家自然科学基金青年科学基金项目(52108286)
深圳市高等院校稳定支持计划项目(GXWD20220818002513001)。
关键词
多模态融合
可见光和红外图像融合
外墙缺陷识别
深度学习
multimodal fusion
visible and infrared image fusion
exterior wall defect detection
deep learning
