Urban Basic Information Extraction From Airborne Laser Scanning Point Cloud

With the development of Airborne Laser Scanning(ALS)technology,point cloud data has gradually become one of the main data sources for urban information extraction.Vehicle is an important means of urban ground transportation.Vehicle extraction is the basis of urban vehicle management,traffic control,digital city construction and other applications.At present,the vehicle extraction method based on ALS point cloud can not effectively solve the problem of over-segmentation and undersegmentation of vehicle point set,which needs to be improved.Therefore,the research on how to improve the accuracy of vehicle extraction can be carried out.As the basic feature of the city,building is also the research hotspot of urban information extraction.At present,the research of building outline extraction mainly focuses on the straight line building.When there are curves in the building,it can not be effectively processed.In general,the outline of urban buildings may contain orthogonal lines,oblique lines and curves.So this thesis proposes an outline extraction algorithm suitable for general urban buildings,which plays an important role in urban basic information extraction.At present,the vehicle extraction algorithm and building contour extraction algorithm from ALS point cloud are not perfect.Aiming at the weakness of existing algorithms,the research contents and conclusions are as follows:A vehicle extraction algorithm based on ALS point cloud is proposed.The concept of potential vehicle occupied area is introduced to avoid the problem of oversegmentation.The improved re-segmentation method based on gap is used to solve the problem of under-segmentation of vehicle point set.Because the geometric features used in the existing algorithms are not the decisive basis for vehicle recognition,the new algorithm uses the vehicle shape as the judgment standard,which can greatly improve the accuracy of vehicle recognition to more than 90%.In vehicle recognition algorithm based on vehicle shape,dynamic time warping similarity is used to measure whether the shape curve of point set is similar to the standard vehicle shape curve.Therefore,vehicle recognition can determine whether the object is a vehicle and which category the vehicle belongs to.Four typical urban scene point cloud data are used to test the performance of the new algorithm,and compared with object-based Point cloud analysis(OBPCA)algorithm and decision tree(DT)algorithm.The experimental results show that the accuracy,recall and F1-score of the new algorithm are 96.7%,91.1% and 93.8% respectively,which are higher than the existing algorithms.On the basis of improved minimum bounding rectangle(IMBR)algorithm,a general building outline extraction(GBOE)algorithm is proposed by introducing boundary point set division.GBOE algorithm first divides the boundary points according to the change of the distance between the boundary points and the primary outline,and obtains three types of boundary point sets.Then,different outline line extraction methods are used for different types of boundary points.Finally,all outline lines are combined to get regular building outline.GBOE algorithm inherits the advantage of IMBR algorithm: it can ensure the orthogonality of right angle outline.At the same time,GBOE algorithm expands the application scene of IMBR algorithm,and can correctly extract the outlines of oblique line building and curve building.Both GBOE algorithm and iterative CD-Spline algorithm can extract the outlines of right angle building,oblique line building and curve building.But GBOE algorithm can identify the type of boundary point accurately,so that the correctness of the extracted building outline line type is higher.Moreover,the GBOE algorithm can extract building outline that meet the requirements of different scales by changing the size of the distance threshold.