| With the rapid development of the information age, how fast and efficient access to the city feature information has become a key issue to build a digital city. As an advanced means of measurement mobile laser scanning can quickly get the three-dimensional spatial coordin ates and the reflection intensity information of city objects; it has been widely applied in tran sportation, military and other fields. The mainly study of mobile laser scanning point cloud concentrated in the point cloud filtering, classification and building reconstruction, but there is a few study of extraction of the road information. Road as the lifeblood of the national eco nomy, plays a crucial importance in the economic development of the country, therefore, the study of mobile laser scanning technology in road measurement has an important practical si gnificance. In this paper, author takes a study of extraction of the road information with obje ct-oriented method on the basis of fully understand the characteristics of the city road, the main research and innovations are as follows:(1) Improved TIN point clouds filtering method. Progressive TIN densification is one of the classic methods for filtering LiDAR point clouds; it divided point clouds into two parts through the establishment of triangulation after excluding gross error and selects the four feet elevation points and the lowest point in each grid point and as seed points, calculate the angle and distance between pending points and triangular patches, then compare distance and angle with threshold, the pending points would be classified as a ground point If the conditions are met. After classify the pending points, Improved TIN point clouds filtering method classify e very triangular patches based on the percentage of ground point in each triangular patches.(2) Road boundary information extraction. There always have curbs above the ground at the edge of city roads, and we can separate sidewalk from road by curbs. In point clouds, pa vement elevation is different form curbs’, and then this paper proposes a method for extractin g road boundary based features image. The first step of the method is calculate the average ele vation of each grid, then select lane track as the starting area for the growth of the region, take the elevation difference between two adjacent grid should be small as grow condition. After g et the road border, there is also need to remove noise (Using elevation variance feature imag e), reserved trunk information (proposed and implemented the "Fork" and "Branch" algorithm s) and disconnect the border (proposed and implemented the marking the start endpoint conne ction algorithm). (3)Lane marking line extraction. The road reflection intensity is different from lane marking lines’, lane marking line is brighter than road in point clouds data, we can extract lane marking line from the road surface if set the intensity threshold, after remove nois e process, lane marking line can be better extraction. Lane marking lines include solid and bro ken lines. This paper proposed and implemented dashed detection and extraction algorithm after fully understanding the characteristics of broken lines. For the broken of solid lines caused by rainfall and other reasons, this paper connection breaks with the marking the start e ndpoint connection algorithm, and get a good results. |
PDF Full Text Request