The Research On Filtering And Extraction Of Road Marking From Vehicle-borne Laser Scanning Point Clouds Data

The Vehicle-Borne Laser Scanning is a new technology way of mobile mapping technology.It can quickly obtain the three-dimensional coordinate information of the object on the surface of the earth. The system has the characteristics of fast speed, high precision, low cost, fine detail and less affected by the weather. The feature on both sides of the road distributed complicated and regular. Vehicle-borne LiDAR system can provide more comprehensive information of the features, so it has been widely used in the construction of "digital city"..The process of extracting the ground points from vehicle-borne LiDAR point clouds data is called filtering. The ground points reflect the change of topography and contain the information of road marking lines. Meanwhile, that’s called thematic information extraction or classification of the point clouds data that marking lines information extracted from the ground point or the non-ground point is distinguished further to obtain the artificial or natural features such as buildings, vegetation etc. These information in urban area plays a pivotal role in the current construction of "digital city" and "smart city". Therefore, the emphases of this paper’s research are mainly on two aspects:the filtering and the extracting of urban road marketing form the point clouds data.The filtering algorithms on the vehicle-borne LiDAR point clouds data, is based on filtering of airborne LiDAR and improved. Although the filtering development of airborne LiDAR has been very mature and some have been very successful and applied to some commercial software, there are many difference between airborne LiDAR and vehicle-borne LiDAR, such as the way of access, the density of point cloud and the distance of point cloud.Therefore the filtering algorithm of airborne LiDAR cannot be directly used, it has to be improved according to the actual characteristics of the vehicle-borne LiDAR.Because of its stability and applicability, the incremental triangular irregular network encryption algorithm is often used in the filtering processing of airborne LiDAR. According to the actual situation of the urban vehicle-borne LiDAR point clouds data, the filtering of traditional incremental triangular irregular network encryption algorithm has some problems. In terms of choosing seed points, because of the occlusion area which is appeared easily in moving of the vehicle-borne LiDAR. There is no ground nadir in the grid corresponding to occlusion area. The seed piece will fall into the air, so the points in the air are divided into ground points. For the points into the perforated strainer etc below the ground water, apparently became the lowest point in the grid, to participate in the initial construction of irregular triangle net, It will product accumulated error in the later encryption process. In the case of sections between the gap, the seed point selection is not reliable, because the seeds point selected from road clearance also may not be ground points. In addition, considering the quantity of LiDAR point cloud data, the algorithm of time cost reduction is also in need of improvement.Based on the above considerations, this paper proposed a method of improved incremental triangular irregular network encryption. In terms of choosing seed points, in order to improved the reliability of the initial seed point selection; using the neighborhood convolution to optimized the seed point and to the constructed the virtual seed point.. In order to verified the effectiveness and feasibility of the algorithm in this paper, the author selected different sample data from SSW system to test. The experimental results showed that, compared to the classic incremental irregular triangulation encryption algorithm, the proposed method can solve the actual problem and reduced the Ⅱ type of error. Because the vehicle-borne LiDAR system was mainly used in urban areas, the filtering algorithm of the object of this study in this paper was just for the urban area and flat area. In general, the algorithm can be applied to most cases of LiDAR data filtering and it had good practical value.After filtering, the point clouds was divided into ground points and non ground points, According to the actual demand, the classification and extraction of point cloud can be further carried out. The information of road marking line, as the urban basic features, it played an important role in the field of automatic driving, precise navigation, urban planning and so on. There has been a lot of road lane extraction methods based on images information. In this paper, the high precision LiDAR point clouds data as the data source, from the view of efficient and accurate, the information of the marking line was accessed. A hierarchical template method based on the intensity information and Alpha_Shapes algorithm based on blocked point clouds was used to extract the identification point cloud. Firstly, the original point clouds data was filtered, which was divided into ground points and non ground points. The ground point was segmented to obtain the pavement data. Secondly, based on the scan line index, considering the geometric characteristics of the road marking line, a rectangular template R was built. After template differentiation, R became into several Ro. R and Ro were calculated to extract the road marking line point cloud. Finally, Alpha_Shapes boundary search algorithm was chosen to reconstructed the boundary based on the extracted identification point cloud. Taking into account the amount of LiDAR point cloud data and the distance of point cloud was small, so the traditional Alpha_Shapes algorithm was not efficient. Therefore, the identification point clouds were blocked in advance, using the neighborhood estimation method to eliminate the point clouds in the non boundary region, thus reduced the amount of algorithm judgment. Finally, Alpha_Shapes model was adopted to deal with the boundary points of the boundary points to get the road marking line entities.In order to verified the effectiveness and feasibility of the method for extracting road marking information proposed in this paper, the author selected four different sample data from SSW, and evaluated the results of the two aspects by the means of visual discrimination and quantitative analysis. The experimental results showed that the proposed method can extract the information of most road markings. In addition to the common long lines, virtual line information, the direction arrow, the road surface instruction text, the vehicle guide belt and so on can also be extracted. From the view of visual discrimination and quantitative analysis, the boundary search algorithm in this paper, compared to the convex hull of search method and external rectangle searching method, had obvious advantages. It can solve the defect of the reconstruction of concave polygon, which was not finished by the convex hull of search method and external rectangle searching method. But it also missed some information which has not enough intensity or has a complex shadow region. Overall, the proposed algorithm can be used on the extraction of road marking information in urban areas and had a certain practicality and feasibility.