Research On The Key Technology Of Extraction Of Roof Feature Line From Airborne LiDAR Data

Airborne LiDAR technology integrated with laser scanning, differential GPS and inertial measurement technology is a new aerial remote sensing technology, the use of the technology can achieve mass3D spatial data acquisition rapidly and precisely, which makes airborne LiDAR data become the latest and most competitive source data used to three-dimensional building reconstruction in the process of digital city construction.3D building reconstruction using laser point cloud has been a hotspot in the research of the application of laser point cloud data. During the building reconstruction based on data-driven method,the extraction of roof feature line from laser point cloud is an essential step. Roof feature line can be classified into roof contour line and non roof contour line,and non roof contour line includes main ridge line,oblique ridge line and valley line.In this paper, with various feature lines on the roof contained in all kinds of monomer roofs induced key technologies and detailed algorithms of extraction of a variety of feature lines on the roof (including main ridge line on gable roof, oblique ridge line,valley line and contour line)has been further studied based on point cloud data from airborne LiDAR data.The main research work is as follows:1) Classification of roof feature line.Fist,this paper conducts induction and summary on various buildings in building architecture at home and abroad,and then the roof feature line is divided into main ridge line,oblique ridge line,valley line and contour line according to the difference of space geometry and function of various pitched roof. At the same time, the main ridge line and valley line are further classified because they have various geometry shape respectively;2) This paper presents a strategy to extract a gable roof ridge line-crawling search strategy. Gable roof ridge line usually represents the main direction of the house, has an important role in the reconstruction of houses. Climbing method is the field of artificial intelligence in a way to get some solutions to the problem. Based on the idea of climbing method, this paper proposes a strategy to extract a gable roof ridge line-crawling search strategy, and implements a prototype algorithm and an improved algorithm. Among them, prototype algorithm implements the extraction of main ridge line on regular gable roof,and the improved algorithm implements the extraction of multiple gable roof ridge lines existed in a complex roof structure;3) The method of detection of feature points on feature line has been studied based on the curve plane fitting using the neighbor points.Firstly, a method of calculating approximate curvature of each point to participate in the quadratic fitting of curve plane has been proposed; secondly the common constraints and the distinguish constraints of feature points on various types of feature lines have been studied, and the common constraints include curvature statistics,roof plane condition and the dist of point to plane,and the distinguish constraints include elevation statistics and the angle between Z axis and normal vector determined by the current laser point located at fitted curve plane;4) This paper presents a grouping method of discrete points located on feature lines based on the point-to-line distance and direction density constraints. Hough transform is the most classic line detection methods, but for more discrete laser points, traditional Hough transform is not applicable. Based on Gestalt figure organizing principle and referring to the discretization method of line,a grouping method of discrete points located on feature lines is proposed;5) This paper presents a method of extracting roof contour line from roof laser point cloud. This paper determines the edge point through the distribution analysis of laser points located in buffer built by a segment which is determined by current laser point and neighbour pointOn one hand, it avoids the effect of concave polygon shape of the roof contour edge,on the other hand,it ensures the density of edge points.On this basis,contour lines are extracted by grouping and fitting as well as regularization and expansion.