Research On Pavement Crack Detection Based On 3D Image Reconstruction

Due to the massive construction of roads in recent years,road traffic is currently the most widely used,most widely served,and most popular mode of transportation.Maintaining large-scale road assets and strengthening road maintenance and maintenance has become a top priority.Due to the impact of driving pressure,weather and human factors,road disasters occur frequently,and cracks,as the most common road surface disaster,have always been the focus of road maintenance.With the rapid development of computer vision,crack detection based on image processing has become mature.Compared with manual detection,it is faster and more convenient,but there are still some problems.This paper aims to improve the accuracy of crack detection and obtain the authenticity of crack scale,in view of the problems and limitations of the existing pavement crack detection technology,according to the characteristics of the pavement image data collected by the vehicle camera,the key technical issues of the pavement crack detection based on the original data as the image are studied in depth,the main contents are as follows :1.Carry out road surface crack detection on the original image data,continuously collect road surface image data obliquely behind the vehicle by installing a camera on the roof of the car,and use the camera to collect vertically downward road surface crack image data by imitating the shooting angle of a special vehicle for road crack detection,directly use the image processing method to carry out the pavement crack detection experiment on the two types of image data collected.The image is first preprocessed to reduce noise and enhance the image;then edge detection is performed on the preprocessed image to obtain the most basic crack information;and then the morphological filter is used to filter and obtain crack skeleton information.Through the image processing method,a comparative experiment is carried out on the two types of image data to verify the difference between the two types of image data.At the same time,the experimental results of the image data collected by the vehicle camera will be compared with the crack detection results based on the threedimensional reconstruction of the image to verify the advanced nature of the crack detection method based on the three-dimensional reconstruction of the image proposed in this paper.2.A pavement crack detection method based on 3D image reconstruction is proposed.By extracting and matching the feature points of the image data continuously captured by the on-board camera,constructing the image connection graph,generating 3D points through triangulation,using the pnp algorithm to restore the camera pose,and using the Bundle Adjustment optimization method to simultaneously optimize the 3D points and the camera pose,and finally generate a sparse point cloud;then generate 3D points for each pixel of the image one by one through the method of depth map fusion,and finally obtain a dense true-color point cloud with real scale.Since the image is endowed with true color information of the point cloud data,the crack can be preliminarily extracted by using the true color information of the point cloud to obtain the crack candidate points;because the point cloud data contains geometric features compared with the image data,the geometric characteristics of the crack candidate points can be analyzed,compare the crack points and abnormal points,and use the spatial density difference between the two to propose a spatial density filtering method to filter out abnormal points and obtain crack points;then use the point cloud clustering method to cluster the discrete point clouds into one by one The clusters of the fracture point clouds can be obtained to obtain each fracture with a real scale;finally,parameters such as length,width and area can be calculated for each crack to realize qualitative and quantitative analysis of pavement cracks.3.Experiment and analysis.The crack detection method based on two types of image data and the point cloud crack extraction method based on image threedimensional reconstruction are experimentally verified,and the experimental results are analyzed.For two types of image data,the same image processing method is used to conduct experiments,and the experimental results are compared to verify the limitations of the method.For the proposed crack detection method based on 3D image reconstruction,the reconstruction error is analyzed,and compared with the experimental results of the image processing method,the feasibility and advancement of the method proposed in this paper are verified.