基于深度学习的公路路面破损检测识别方法  

Detection and Identification Method of Road Surface Damage Based on Deep Learning

作　　者：乌日娜[1] 白云[1] 韩建峰[1] WU Ri-na;BAI Yun;HAN Jian-feng(Inner Mongolia University of Technology,Hohhot Inner Mongolia 010051,China)

机构地区：[1]内蒙古工业大学,内蒙古呼和浩特010051

出　　处：《计算机仿真》2023年第1期208-212,共5页Computer Simulation

基　　金：内蒙古自治区关键技术攻关计划项目(2019GG271);内蒙古自治区高等学校科学研究重点项目(NJZZ19068)。

摘　　要：路面破损是影响路面性能的主要因素,其中路面裂缝又是破损中最主要的病害形式,所以,准确检测裂缝并进行及时修复是道路养护中的一个重要课题。采用深度学习的方法,提出了一种路面破损自动检测识别的综合模型。运用路面破检测识别网络对破损裂缝进行检测识别,分类后获得裂缝的类别置信度;利用改进的破损分割网络对路面破损裂缝进行精确分割;结合两种网络的优势提出路面破损检测识别综合模型。整体模型通过优化模型参数和改进网络结构,提高了路面破损的分类精确度和分割效果。与传统的单一模型相比,提出的综合模型不仅能提供路面破损的类别信息,还能提取路面主要裂缝的骨架信息,对于路面养护工作有一定的实用价值。The pavement damage is the main factor affecting the pavement performance, and the pavement crack is the most important form of damage, therefore, the accurate detection of cracks and timely repair is an important topic in the road maintenance. Based on the deep learning method, a comprehensive model for automatic detection and identification of pavement damage is proposed. Firstly, the pavement crack detection and identification network was used to detect and identify the damaged cracks, and the category confidence of cracks was obtained after classification;Then, the improved damage segmentation network was used to segment the pavement damage cracks accurately;Finally, combined with the advantages of the two networks, a comprehensive model of pavement damage detection and identification was proposed. By optimizing the model parameters and improving the network structure, the classification accuracy and segmentation effect of pavement damage were improved. Compared with the traditional single model, the proposed comprehensive model can not only provide the classification information of pavement damage, but also provide accurate location information, which has a certain practical value for pavement maintenance work.

关 键 词：城市交通 破损检测 深度学习 路面破损 图像分割 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]