Research And Application Of Road Extraction From Mobile Laser Scanning Point Clouds

As an important infrastructure,road affects our production and life at all times.Accurate road extraction is of great significance in the fields of intelligent transportation and smart city.In recent years,with the development of autonomous driving,High Definition Maps creation and other fields,road extraction technology based on mobile laser scanning point clouds is playing an increasingly important role.The mobile laser scanning system,integrated with various sensors such as laser scanner,CCD camera,IMU,DMI and GNSS,can quickly obtain a wide range of ground object information.The obtained point clouds data is widely used in road modeling and 3D reconstruction of buildings,high-precision mapping and other fields.The large data volume,complex scenes,uneven density distribution of mobile laser scanning point clouds and the not prominent features of road make it so challenging to automatically extract road from mobile laser scanning point clouds.The main work and innovations of the article are as follows:1.An improved method of road extraction based on scanning lines is presented.Since the three filter windows are set to the same size in the traditional road extraction method based on scanning lines,non-ground points are prone to appear on the road boundary.To solve this problem,this paper reduces the size of window 1 while maintaining the same size of window 2 and window 3 and improves the adaptability of moving windows.Through detailed comparative experimental analysis,it is found that the proposed method can effectively remove the interference of non-ground points at the road boundary and provide more accurate point cloud data for subsequent road boundary and lane marking extraction.2.A spatial straight-line fitting algorithm based on RANSAC method is designed to extract accurate road boundaries.Firstly,curb points are extracted based on elevation difference and slope from extracted road section.Then,the lowest points of curb are retained as the candidate road boundaries.Finally,the RANSAC line fitting algorithm is used to remove the pseudo boundary points.3.A lane marking extraction process combining point clouds intensity image and edge detection algorithm is designed.Firstly,project the point clouds data onto the horizontal plane to generate an intensity feature image and use a median filter algorithm to remove noise;then use the Sobel operator to calculate the gradient of the intensity image;and binarized the intensity gradient image based on the principle of maximum entropy;finally optimize the results based on Connected Component Analysis.Experiments show that this method can effectively extract lane marking.