Research On Pavement Extraction Based On Vehicle 3D Laser Scanning Data Classification

With the development and progress of information technology, how to achieve the rapid acquisition of spatial data has become the focus of the study of geographic information system. Compared with the traditional measurement method, the vehicle mounted laser scanning technology can quickly acquire the 3D data of large area and high precision. As a new method, the new data acquisition method has been applied in the geographic information industry. Classification and extraction of laser scanning data is the premise and key of 3D modeling, and the data can be used according to the actual needs of users. Vehicle borne laser scanning system in the urban data acquisition to obtain the pavement, buildings, vegetation, wire rod and other different types of point cloud data, the traditional classification method of vehicle mostly research how to extract buildings, windows, poles and other data, and according to the working principle of the vehicle borne laser scanning system, in the data collection process, the road runs through the on scene of a scan of the vehicle and a low elevation, if will first classification of pavement data out will reduce follow-up study on the amount of data and other terrain classification accuracy can be improved.Aiming at the above problems, this paper introduces the structure and principle of vehicle system and data acquisition method, acquisition process, classification and summarizes other scholars based on the laser scanning data, on the basis of previous studies, this paper adopt the method based on surface characteristics and asymptotic grid combination, classification and extraction the road.Firstly, according to the characteristics of the road surface, the elevation value of the point cloud data is small. Set the threshold for high elevation elevation point cloud data. This will reduce the amount of data classification. Then based on gradual method of Grid Classification and extraction of road, built according to the measured area size, and acquisition of data points of coordinate grid. By setting the threshold and point cloud data of elevation values, within the grid elevation difference comparison road classification and extraction. Classification and extraction of Road on the basis of the above methods, grid spacing setting will affect the final classification result, if the grid is set too large, in the process of classification may will appear the misjudgment of ground objects, and removing the ground points, and smaller grid may leads to the misjudgment of the feature points for pastry. Therefore, this paper uses the method of repeated adjustment of grid spacing, to improve the classification performance, classification and extraction of the pavement.Through experiments on the existing data can be obtained in scanning the scene, due to the terrain in constant change, the size of the grid should be adjusted according to the actual situation, the grid adjusted big or small, and classification effect are not is ideal. In the actual situation, should be adjusted many times, so that the grid size is moderate, ensure the effect of classification.