Design Of Road Defect Detection And Identification System Based On Machine Vision

With the rapid development of China’s transportation industry,the length of highways continues to increase,and the requirements for road safety operation are getting higher.As a result,there is a growing demand for road inspection and maintenance.Currently,most of the investigation and evaluation of road defects rely on manual inspection and semi-automatic detection,which are inefficient,dangerous,and susceptible to human subjectivity.This article designs a machine vision-based road defect detection and recognition system that can obtain the category,feature parameters,and location information of road defects.These pieces of information can provide data support for scientific road maintenance.Firstly,road cracks and potholes are the two main types of road diseases.Currently,there is no applicable public image dataset for studying road defect detection.Therefore,based on standard principles of image acquisition,sources,and pre-processing,an image dataset is constructed.We conducted research on road defect classification and recognition algorithms,and selected the classic VGG16,Res Net50,and Mobile Net V3 network models.Through experimental simulation and comparative analysis of the three network models,it was found that the Mobile Net V3 network model has the advantages of high classification accuracy and small parameter quantity,which meets the demand for road defect classification and recognition.Secondly,we studied parameter extraction for road potholes.For images with obvious features only containing pothole defects,we used grayscale conversion,OSTU adaptive binarization,morphology processing,and edge contour extraction to obtain the parameter information of the quantified pothole target area.We further achieved the level division of pothole damage,and verified the effectiveness of this method through pothole defect detection.Thirdly,we studied parameter extraction for road crack defects.For images containing only crack defects,we proposed an improved YOLOv5 s object detection algorithm,named YOLOv5s-DA.The introduction of SCConv modules in the backbone region of the original network model structure helps to improve the feature information extraction ability of the main network,while the introduction of SE-Net modules in the neck region of the original network model structure helps to improve the high-level feature information extraction ability of the original model.Experimental results show that the improved object detection algorithm has the advantages of high detection accuracy and good detection effect.Using the improved YOLOv5s-DA object detection algorithm,we obtained the peripheral boundary box parameter information of the quantified crack target area and further achieved the level division of crack damage.We verified the effectiveness of this method through crack defect detection.Finally,we designed a vehicle-mounted road defect detection system.The system can realize road image acquisition,classification and recognition,parameter extraction,4G communication,location positioning,and map display functions.The system detection and recognition results,as well as location maps,are displayed visually through the Qt interface.Testing of the vehicle-mounted road defect system showed that the system can detect and recognize the categories of potholes and crack defects,and complete the parameter extraction and level division of defect types.