Detection and Classification of Concrete Pavement Cracks Based on Residual Neural Networks and the Bisection Method

Author:

yu zhi1,Wu Qiong1,Tang Changhua1,Shi Qingtao1,Song Wei1,Si Junrui1

Affiliation:

1. College of Humanities and Information Changchun University of Technology

Abstract

Abstract

Concrete pavement cracks can reduce road safety and lead to traffic accidents. Detecting road cracks and implementing appropriate maintenance measures holds significant practical importance. Addressing the low detection accuracy of existing research methods for road crack detection and the limited studies on classifying and rating the severity of road cracks, this study first proposes a COTECANet model based on the ResNet50 architecture. This model effectively detects pavement cracks with a recognition accuracy of 99.8%, surpassing other compared deep learning models. Subsequently, for roads with detected cracks, the maximum inscribed circle radius of the crack contours in the images is computed using the bisection method, thereby obtaining the maximum pixel width of the road cracks. Finally, by proportional conversion, the actual width of the measured pavement cracks is obtained, and the damage severity of the road cracks is classified and rated according to relevant standards. This research can help highway management departments implement corresponding maintenance measures based on the actual conditions of road damage, thereby extending the lifespan of roads and possessing practical application significance.

Publisher

Springer Science and Business Media LLC

Reference26 articles.

1. Crack and Noncrack Classification from Concrete Surface Images Using Machine Learning[J];Kim H;Struct Health Monit,2018

2. Autonomous concrete crack detection using deep fully convolutional neural network[J];Dung CV;Autom Constr,2019

3. Classification and quantification of cracks in concrete structures using deep learning image-based techniques[J];Flah M;Cem Concr Compos,2020

4. Meng XY (2021) Concrete crack detection algorithm based on deep residual neural networks[J]. Scientific Programming, 2021: 1–7

5. CrackU-net: A novel deep convolutional neural network for pixelwise pavement crack detection[J];Huyan J;Struct Control Health Monit,2020

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3