基于数据微观精分的沥青路面裂缝自动化检测  

Automatic detection of asphalt pavement cracks based on data micro-segmentation

作　　者：傅幼华 罗文婷 李林 倪昌双 杨振 FU Youhua;LUO Wenting;LI Lin;NI Changshuang;YANG Zhen(College of Transportation and Civil Engineering,Fujian Agriculture and Forestry University,Fuzhou 350108,China;College of Transportation Engineering,Nanjing University of Technology,Nanjing 211816,China)

机构地区：[1]福建农林大学交通与土木工程学院,福州350108 [2]南京工业大学交通运输工程学院,南京211816

出　　处：《交通科技与经济》2022年第6期45-52,共8页Technology & Economy in Areas of Communications

基　　金：国家重点研发计划项目(2021YFB3202901);福建省高校产学合作重大项目(2020H6009)。

摘　　要：为提高沥青路面裂缝的自动化检测精度,以深度学习模型数据集为对象,结合传统图像处理,提出一种基于裂缝数据进行微观精分的深度学习裂缝数据集处理方法,使深度学习模型能更有针对性地学习沥青路面裂缝的数据特征。在构造数据集时,通过传统的图像处理方法将裂缝宏观精分为龟裂和线性裂缝。在宏观精分基础上,对线性裂缝和龟裂的像素级特征进行微观精分。YOLOv5模型的实验表明:微观精分的裂缝病害模型mAP@.5相较于宏观精分和未精分数据库分别提高8.9%和14.7%,在随机路段上有较好的检测性能。The purpose of this paper is to improve the accuracy of automatic detection of asphalt pavement cracks. Combined with traditional image processing, this paper takes the data set of the deep learning model as the research object and proposes a deep learning crack data set processing method based on crack data for micro-segmentation. The data characteristics of asphalt pavement cracks can be learned in a more targeted manner. When constructing the dataset, the cracks are macroscopically divided into alligator cracks and linear cracks by traditional image processing methods. On the basis of macro-resolution, micro-resolution is performed on pixel-level features of alligator cracks and linear cracks. The experiments of the YOLOv5 model show that the mAP@.5 of the micro-refined crack disease model is improved by 8.9% and 14.7% respectively compared with the macro-refined and unrefined databases. The micro-refined crack disease model have better detection performance in random road sections.

关 键 词：宏观精分 微观精分 局部自适应阈值 路面裂缝检测 YOLOv5 

分 类 号：U416.2[交通运输工程—道路与铁道工程]