EXTRACTION AND REGULARIZATION OF BUILDING CONTOURS FROM LiDAR DATA USING ALPHA-SHAPE ALGORITHM AND PRINCIPAL COMPONENT ANALYSIS

Abstract

The aim of this work is to propose a procedure for extraction and regularization of building roof contours from LiDAR data. The LiDAR data correspond to points sampled by airborne LASER scanning system. The contribution of this project is the development of a methodology that allows the automatic extraction and regularization of straight-line and curved contours in three-dimensional space, since the available methods in the literature have been performed the regularization in 2D space and have not been considered curved segment. To perform the extraction and regularization the idea is to combine principal component analysis and alpha-shape algorithm. Principal component analysis will be applied to calculate the eigenvalues and eigenvectors associated with each LiDAR point and its neighborhood. The eigenvalues will be used to identify the probable edge points by means of a clustering approach, whereas the eigenvectors will be used to identify the type of segment (straight-line or curve) and select the set of points that compose each contour segment. The alpha-shape algorithm will be executed to obtain approximate contour of each building. Regularization will be held by fitting a straight line or a conic curve (parabola, circle or ellipse) on the set of points associated with each segment by Least Square Method. The results generated will be analyzed through qualitative and quantitative analysis. Qualitative analysis will be performed from a visual analysis, while quantitative analysis will be performed using some quality measures, which allow comparing the results with the reference data, such as: root mean square error, completeness and correctness.