An Edge Detection And Building Extraction Method Based On Normal Vector Of TIN

Li DAR(Light Detection And Ranging, Li DAR) is one of the most revolutionary achievements in a certain field of photogrammetry and remote sensing in recent decades. It can quickly get accurate high resolution digital terrain model and three-dimensional coordinates of ground objects,while it has the advantages of high precision, high density, high efficiency and low cost. In recent years, with the high-speed development of urban construction, how to acquire precisely fast distribution of urban terrain, buildings, and various edge features is especially important. Extracting buildings in several fields such as city modeling and 3D simulation is very extensive. In this paper, according to the external characteristics of different terrain and ground objects, the edge detection data and building extraction of Li DAR were studied. The main contents include the following two parts:1. In view of large gradient computation problem of the traditional edge detection method, a new method of edge detection of the Li DAR data cloud based on the normal vector of Triangulated Irregular Network is proposed. Firstly, the method constructes the TIN, and calculats the normal vector of each triangle and the angle between each triangle and the horizontal plane. Secondly, by comparing with the angle threshold, extract the edge triangle, and consider the highest point of the edge triangle as the edge points. The angle threshold value is obtained by histogram statistics. The experimental results show that the proposed algorithm can better extract edge points of Li DAR data cloud. Finally, the blank data of the river area is interpolated, which gets the good edge of rivers.2. In view of the problem of traditional method of detecting building incompletely, a new method of derivation of building region information from Li DAR based on building contour is proposed. Firstly, structure TIN for data cloud after filtering, and determine whether triangles in TIN is coplanar, according to angles between normal vectors of triangles around each point, which can detect initial points of buildings. Secondly, the building outline is extracted from border triangle nets and stitched with short edge, according to separate the TIN into building triangle nets and border triangle nets with length threshold. Finally, the building points within the contour is extracted by determining whether each point is in the building outline. The experimental results show that the proposed algorithm can better extract buildings from Li DAR data points, and accurately get the 3D contour line of buildings. Eventually, the extraction accuracy rate of buildings can reach above 95%, while the integrity rate can reach above 90%, which achieved the extraction accuracy requirement of building extraction.