Research On Road Extraction Algorithm Based On Oblique Photography Point Cloud Data

As a classic linear feature,road has always been one of the hot research objects in the field of information feature extraction.In traditional photogrammetry and remote sensing technology,road information is extracted based on orthophoto or remote sensing images and classical digital graphics theory.Due to factors such as spatial resolution,spectral resolution and imaging time,the properties of the roads in the image are seriously disturbed,which makes the results and accuracy of road extraction difficult to achieve the expected results.Tilt photogrammetry technology can quickly and efficiently acquire point cloud data of large area survey area,and extract road information based on point cloud data,which is becoming one of the hot research contents.In this paper,the point cloud data generated by oblique photogrammetry is taken as the research object.Based on the analysis of the theoretical methods and application status of road extraction based on point cloud data,the key technologies involved in road extraction around point cloud data are studied.Firstly,based on the principle of tilt photography,the point cloud noise is divided into two categories: coarse noise and fine noise.The two types of noise are separately eliminated by different methods,and experimental analysis shows that in various denoising methods.Voxel filter denoising has good denoising effect on both types of noise.Based on the experimental comparison of the existing mainstream point cloud filtering algorithm,it is shown that the progressive encryption filtering based on TIN has a good filtering effect for most scenarios.For the initial stage,the maximum size and other parameters of the building need to be set.Insufficient,an improved algorithm for determining the initial ground point using a virtual grid is proposed,on which all ground points are extracted.The RGB attribute of the extracted ground point is re-presented in the HSL color space and the L component is segmented.By selecting the road seed point,the initial road point is obtained by using the region growth extraction.After the initial road point cloud is segmented and projected,the road boundaries are clustered and repaired by operations such as boundary extraction and smoothing,and finally the complete road surface is obtained.Experiments show that based on the ground point obtained by voxel filtering denoising and the improved filtering algorithm,the road extraction algorithm proposed in this paper can effectively extract the road surface from the oblique photography point cloud,and the completeness of road extraction is 86.07.%,the correct rate is 84.79%,and the quality index is 80.86%,which can meet the processing requirements based on oblique photography point cloud road extraction.