Accuracy of 3D models derived from aerial laser scanning and aerial ortho-imagery

Author： Park Hongjoo Salah Mahmoud Lim Samsung

ISSN： 1752-2706

Source： Survey Review, Vol.43, Iss.320, 2011-04, pp. : 109-122

Disclaimer: Any content in publications that violate the sovereignty, the constitution or regulations of the PRC is not accepted or approved by CNPIEC.

Previous Menu Next

Abstract

Three dimensional (3D) mapping has been used widely in the spatial industry as a powerful technique for rendering artificial objects and their surrounding topography. However, an accurate and effective 3D modelling of complex features e.g. feature rich buildings and trees, is still challenging. The aim of this study is to develop an efficient framework to obtain a high-accuracy 3D model of urban buildings using Airborne Laser Scanning (ALS) data and aerial ortho-imagery. First, building outlines and a digital surface model are extracted from ALS data. Aerial ortho-imagery is then integrated to improve the accuracy of building outlines. Digital photos of building facades are patched to the 3D model for texture mapping. The accuracy analysis is conducted by assessing the heights and outlines of extracted features. As far as the authors know, this is the first accuracy evaluation of constructed 3D models. The digital surface model (DSM) shows vertical errors of less than 12 cm. Building heights are less accurate than the DSM, with errors of less than 22 cm. This difference is explained in the paper. In order to examine the building model more closely, the buildings are classified into three categories: simple rectangular objects, complex polygons, and curved outlines. The horizontal accuracy of the three categories ranges from 42 to 64 cm in Easting (1σ) and 16 to 48 cm in Northing (1σ). The results show that the horizontal coordinates of simple rectangular buildings are more accurate than those of complex polygons or circular-shape buildings. Mean errors and root mean square errors for each category are presented in the paper.