Download files
Abstract
Urban areas are rapidly changing over time in most of the countries. Buildings are the main entities of urban areas, and building boundaries are one of the key factors for urban mapping and city modelling. Hence, building-boundary extraction from various data sources has been widely studied. Approaches to the building-boundary extraction can be categorised as either raster-based methods or point-based methods. Many research efforts have been made to raster-based methods, while few point-based methods exist. The main aim of this study is to detect and extract building boundaries from airborne lidar data using a vector-based method. Three steps are applied directly to the raw lidar points rather than the rasterised lidar data. Firstly, an adaptive morphological filter is developed and applied to separate ground and non-ground points. Secondly, a fusion of methods including Normalized Difference Vegetation Index (NDVI), hierarchical clustering and thresholding is developed in order to remove unwanted points such as trees and cars. Finally, building-boundary polygons are extracted and delineated, based on alpha-shape and Douglas-Peucker algorithms. The test results show that the adaptive morphological filter can accurately classify ground and non-ground points. The extracted building-boundary polygons are proven to be reliable statistically and visually. It is concluded that the proposed point-based method is able to achieve accurate results, even though the low-accuracy horizontal coordinates of the lidar data have limited the algorithm design. The proposed algorithm can also be used for more accurate building-boundary extraction if supplemental data resources such as aerial images are provided.