Extraction And Regularization Of Road Signs Based On UAV Point Cloud Data

With the rapid development of cities,more and more families own private cars,which bring convenience to people but also cause road problems such as traffic jams and accidents.Therefore,it is crucial to accurately model roads and extract road sign outlines quickly and efficiently.Currently,UAVs are making a splash in the field of point cloud data collection and 3D modelling by virtue of their fast and efficient data collection and automated data processing capabilities.This paper takes the campus of Anhui University of Technology as the study area,and uses the Genie 4 RTK UAV to plan the route,set up image control points and collect UAV images for the whole campus.The main research contents and findings include.(1)Introduction of the principle and technology of UAV photogrammetry,as well as the deployment of UAV image control points and image acquisition methods,the use of ContextCapture software for aerial triangulation of UAV images to obtain a dense point cloud of the target area,as the point cloud data generated by the UAV contains a large number of noisy points affecting the subsequent extraction,Gaussian filtering is carried out to linearly smooth the point cloud data processing.(2)As the point cloud data contains ground points and non-ground points,the RANSAC algorithm is used to separate the ground points first,and then the Otsu method-maximum variance between classes(OTSU algorithm)is used to extract the road marker point cloud from the ground points.The paper introduces four clustering methods:k-means clustering,DBSCAN clustering,KNN clustering and FCM clustering,and conducts comparative experiments to obtain a visualisation and accuracy analysis:the k-means quality factor is 0.9213,which is more accurate and can solve the curvature problem better and more conveniently..(3)The thesis introduces three algorithms:convex packet algorithm,Delaunay triangular mesh method,Alpha-Shapes algorithm for contour extraction,and after comparative experiments it is obtained that Alpha-Shapes algorithm extracts road sign contours with the best effect,and for the problems of bending and unclear folding of the extracted contour lines,four methods:corner point detection method,DouglasPeucker algorithm,the iterative endpoint fitting method and the improved polygon fitting method are compared and tested for regularization.The experiments show that the root mean square error of the improved polygon fitting method can reach 0.0523m,and the extracted boundary is highly accurate and fast to fit.Figure[34]Table[6]Reference[70]...