Research On Three-dimensional Reconstruction And Analysis Of Pavement Asphalt Potholes Based On Machine Vision

With the continuous development of highway construction and traffic transportation,the main center of gravity of highway construction in our country has gradually shifted to road detection and maintenance.Because pit groove disease is the common type of asphalt pavement,at the same time is the main disease type that can affect traffic safety and the driving experience.In addition,the detection and maintenance of pit diseases involves volume problems,so compared with other types of diseases such as cracks,the detection and maintenance of pit diseases will be more complicated and difficult.Based on this,this thesis conducted the following research on the three-dimensional detection method of pothole disease based on machine vision:(1)The multi-angle sequence image data of the crater disease collected by the UAV was reconstructed with the Structure from Motion method to obtain the three-dimensional crater disease model,which could be visually displayed in 360°.(2)Preprocess the point cloud data of the pothole disease model,convert the coordinates of the point cloud through principal component analysis,and conduct the normalization processing of the point cloud model.Conduct the noise reduction of the point cloud through the point cloud filtering,eliminate a large number of hash points and isolated points,carry out the plane fitting of the point cloud data,and identify the pavement and the disease respectively.Plane segmentation is carried out to separate the pit damage part from the pavement part for the next step.(3)Calculate and analyze the damage factors of the pit disease after acquiring the pit disease part,and obtain the damage area of the pit disease through the projection area after triangulation of the pit disease point cloud.On the basis of triangulation,the pit damage was divided into several geometric columns by the idea of integration,and the total volume of the pit damage was obtained by calculating the volume of each geometric column.The maximum damage depth of pothole is obtained by point cloud traversal calculation based on the fitted pavement plane equation.(4)Divide the damage degree of pothole diseases according to the calculated data of each damage factor of pothole diseases combined with the Technical Assessment Standard of Highway Condition,and give the damage layer division table considering the volume factor of pothole diseases.The experimental results have shown that the machine vision-based 3D crater disease detection method proposed in this thesis can display the crater disease intuitively,with a high degree of reduction,and at the same time can accurately obtain the three-dimensional parameter information of the crater disease.The calculation accuracy of damage factors of pit diseases is above 90%,which has practical reference value.In practical application,compared with the traditional detection method,it can provide better technical support for the detection and maintenance of pit diseases.