首页|期刊导航|路基工程|基于注意力机制的道路雷达图像裂缝特征识别方法

基于注意力机制的道路雷达图像裂缝特征识别方法OA

Method for Crack Feature Identification in Road-Surface 3D GPR Images Based on Attention Mechanism

中文摘要英文摘要

针对传统三维探地雷达图像人工解译效率低下、专业门槛高且易受杂波干扰等局限,提出基于改进YOLOv8模型的道路裂缝智能检测方法.以广东省某高速公路为试验对象,基于三维探地雷达数据进行道路裂缝的识别.融入 ECA(Efficient Channel Attention)注意力机制对 YOLOv8模型进行优化,通过避免通道降维操作实现高效且无损的跨通道信息交互,结合主干网络强大的特征提取能力,显著增强多尺度裂缝特征的感知能力.结果表明:所提出的模型在精确率和召回率上分别达 96.5%和 94.4%,该方法能有效抑制雷达图像中的散斑噪声干扰,实现道路裂缝的高精度、高效率识别.

Considering the limitations of manual interpretation of traditional 3D Ground Penetrating Radar(3D GPR)images,including low efficiency,high professional threshold,and susceptibility to clutter interference,this paper proposes an intelligent road crack detection method based on the improved YOLOv8 model.Taking an expressway in Guangdong Province as the research object,road crack identification is performed using 3D GPR data.The Efficient Channel Attention(ECA)mechanism is integrated to optimize the YOLOv8 model,which achieves efficient and lossless cross-channel information interaction without channel dimension reduction.Combined with the powerful feature extraction capability of the backbone network,the model's perception of multi-scale crack features is significantly enhanced.The experimental results show that the precision and recall of the proposed model reach 96.5%and 94.4%,respectively.This method can effectively suppress speckle noise interference in GPR images and realize high-precision and high-efficiency road crack identification.

冯宝平;李彦斌;王凯;臧博;李伟雄

中交二公局东萌工程有限公司,西安 710119中交二公局东萌工程有限公司,西安 710119中交二公局东萌工程有限公司,西安 710119广州肖宁道路工程技术研究事务所有限公司,广州 510000广州肖宁道路工程技术研究事务所有限公司,广州 510000

交通工程

道路内部损伤道路裂缝识别雷达图像目标检测信息交互科学养护决策

road internal damageroad crack identificationradar imageobject detectioninformation interactionscientific maintenance decision-making

《路基工程》 2026 (5)

55-60,6

10.13379/j.issn.1003-8825.202603042

评论