基于YOLOv3的农村公路路面病害自动识别研究  

Research on Automatic Identification of Rural Highway Pavement Diseases Based on YOLOv3

作　　者：陈昕[1] 王曦 黄晶晶 杨硕 陈佳雯 CHEN Xin;WANG Xi;HUANG Jing-jing;YANG Shuo;CHEN Jia-wen(School of Automobile and Traffic Engineering,Liaoning University of Technology,Jinzhou 121001,China;Great Wall Motor Co.,LTD.,Baoding 071000,China;Industrial Corporation,Liaoning University of Technology,Jinzhou 121001,China)

机构地区：[1]辽宁工业大学汽车与交通工程学院,辽宁锦州121001 [2]长城汽车股份有限公司,河北保定071000 [3]辽宁工业大学实业总公司,辽宁锦州121001

出　　处：《辽宁工业大学学报（自然科学版）》2023年第5期293-295,共3页Journal of Liaoning University of Technology(Natural Science Edition)

基　　金：辽宁省先进装备制造业基地建设工程中心项目(LNZC2023-0041-1);辽宁工业大学教学改革研究项目(xjg2022014)。

摘　　要：对农村公路路面病害进行自动识别研究,为农村公路路面病害识别工作节省时间、人力和财力,提高识别效率。以农村公路路面病害普查图片作为数据集,提出农村公路路面病害自动识别架构,运用改进后的YOLOv3算法,采用darknet-53网络框架,对农村公路路面病害图片进行标注,之后依次进行样本处理、特征值提取、锚框选择、损失计算及病害定位,训练得到损失值逐渐降低的农村公路路面病害特征权重值,运用训练得到的路面病害特征权重值对测试集中的图片进行识别。识别结果表明,YOLOv3算法病害自动识别能够把图片中的病害识别出来,可用于农村公路路面病害识别工作中。Based on the improved Yolov3 algorithm in the deep learning convolutional neural network,the automatic identification of rural highway pavement diseases is researched.It can save a lot of time,manpower and financial resources and improve the identification efficiency for rural pavement disease identification.Based on the survey pictures of rural highway pavement diseases as the data set,an automatic recognition framework for rural highway pavement diseases was proposed.The improved YOLOv3 algorithm and darknet-53 network framework were used to label the rural highway pavement diseases pictures,and then sample processing,feature value extraction,anchor frame selection,loss calculation and disease location were carried out in sequence.The characteristic weight value of rural pavement disease with gradually decreasing loss value is obtained by training.The characteristic weight value of road disease obtained by training is used to identify the pictures in the test set.The recognition results show that the YOLOv3 algorithm can identify the diseases in the pictures,and can be used in rural road pavement disease recognition.

关 键 词：农村公路 路面病害 YOLOv3算法 自动识别 

分 类 号：U16[交通运输工程]