东南大学学报（英文版）2023，Vol.39Issue(4)：340-349,10.DOI:10.3969/j.issn.1003-7985.2023.04.003
农村公路水泥路面裂缝智能检测
Intelligent detection of cracks on cement pavements of rural highways
摘要
Abstract
Traditional artificial image processing methods suffer from problems in the detection of damage in rural highway pavements,such as low efficiency,nonobjective results,and the inability to process a large amount of data in time.To solve these problems,an intelligent method is proposed for the detection of cracks on rural cement pavements.The proposed method is integrated with ResNet50 for pavement classification and an improved YOLOv5 crack detection algorithm,considering the distribution characteristics of rural highway sections.Different training strategies and different network depth were compared to construct an efficient pavement classification model based on ResNet50 with the aim of automatically identifying cements and asphalt pavements in rural highways.A dataset that contains 18 028 pieces of crack detection data for cement pavements of rural highways was created.A comparative experimental study of single-and two-stage object detection algorithm was performed,and the optimal detection algorithm with both detection accuracy and efficiency was obtained.Furthermore,the adaptive spatial feature fusion strategy and the optimized regression loss function are integrated into the optimization algorithm to effectively solve the problem of multi-scale crack leakage detection in the image,and further improve the overall detection accuracy.The integrated method was applied to the field measurement of real cement pavements of rural highways.The results demonstrate that the accuracy of pavement type classification is 98.4％and that of crack detection is 93.0％,indicating that the proposed method can provide accurate and efficient solutions for the detection of cement pavements of rural highways.关键词
农村公路/水泥路面/裂缝/深度学习/图像分类/目标检测Key words
rural highway/cement pavement/crack/deep learning/image classification/object detection分类
交通工程引用本文复制引用
王萌,张小月,刘诚,徐慧通,杨燕泽..农村公路水泥路面裂缝智能检测[J].东南大学学报（英文版）,2023,39(4):340-349,10.基金项目
Beijing Nova Program(No.20220484103),Beijing Municipal Natural Science Foundation(No.8222027),the Fundamental Research Funds for the Central Universities(No.2022YJS071). （No.20220484103）
