基于深度学习技术的无人机玻璃幕墙缺陷检测方法  

Research on Application of Drones in Glass Curtain Wall Defect Detection Based on Deep Learning Technology

作　　者：凌均健 徐东华[1] Ling Junjian;Xu Donghua(Guangzhou Maritime University,Guangzhou 510725,Guangdong)

机构地区：[1]广州航海学院,广东广州510725

出　　处：《武汉工程职业技术学院学报》2024年第3期26-31,共6页Journal of Wuhan Engineering Institute

基　　金：2023年广东省普通高校重点领域专项课题“建筑玻璃幕墙损坏机理及智能无人检测系统的研究应用”(项目编号:C2301002681)。

摘　　要：针对玻璃幕墙存在的各种缺陷问题,提出了一种基于深度学习技术的无人机玻璃幕墙缺陷检测方法。通过无人机搭载RGB相机采集玻璃幕墙缺陷图像并导入计算机中进行处理,利用改进的YOLOv5轻量型表面缺陷检测算法对玻璃幕墙缺陷图片进行分类识别,提高网络收敛速度和泛化能力;最后通过人工标注数据集对所构建的无人机检测系统进行测试验证,结果表明该系统能够完成玻璃幕墙缺陷图片的自动检测与识别。Aiming at the various defects existing in glass curtain walls,this paper proposes a UAV defect detection method for glass curtain walls based on deep learning technology.The defective images of the glass curtain wall are collected by the RGB camera carried by the UAV and imported into the computer for processing,and the defective pictures of the glass curtain wall are classified and recognized by the improved YOLOv5 lightweight surface defect detection algorithm,which improves the network convergence speed and generalization ability;finally,the constructed UAV detection system is tested and verified by manually labeling the dataset,and the results show that the system is able to finish the automatic detection and recognition of glass curtain wall defect pictures.This method solves the deficiencies of the current manual detection and traditional visual methods,improves the efficiency of the detection process and the safety of the inspectors,and has certain application value.

关 键 词：无人机 深度学习技术 玻璃幕墙 缺陷检测 YOLOv5 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TQ171.6[自动化与计算机技术—计算机科学与技术]