Semi-Supervised Road Surface Crack Detection Method Based on Wide Residual Networks
LIU Xunxing
Abstract：The"Village-to-Village"project has improved the travel conditions for farmers.However,more than a decade later,the road surface is seriously damaged and requires extensive maintenance.Road cracks are one of the most common types of road damage.To enhance maintenance efficiency,accurate classification of road cracks is essential.This paper proposes a semi-supervised deep learning model MixMatch,based on wide residual networks,for road crack recognition methods to achieve precise classification.The wide residual network is enhanced,and the MixMatch algorithm is designed to reduce the workload of data labeling.The proposed method utilizes only 10％of the labeled data to achieve an accuracy of 91.25％in road crack recognition,with F1-score values of 92.37％,96.20％and 90.24％for horizontal cracks,vertical cracks,and spider cracks,respectively.When compared to four traditional model(VGG16,ResNet50,AlexNet,MobileNetV3),the accuracy is increased by 8.08％,0.6％,8.89％,6.97％respectively.Experimental results demonstrate that the proposed method not only provides category information for road cracks and reduces the workload of sample labeling but can also be directly utilized for road condition evaluation.
Keywords：semi-supervisedMixMatchresidual networkcrack detection
Publication Date：2024-03-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：8( 92-99 )