Effect of data density, scan angle, and flying height on the accuracy of building extraction using LiDAR data BHARAT LOHANI* and RAJNEESH SINGH Department of Civil Engineering, Indian Institute of Technology Kanpur, Kanpur, India 208016 (Received ????; in final form ????) 2 A Hough transform based approach for extraction of buildings using LiDAR data is presented. It is argued that LiDAR data should be smoothed and sparsed prior to Hough transform for better result. Algorithms to realize this are presented. Further, an algorithm which fits a vector model to extracted buildings is outlined. Simulated LiDAR data have been used to investigate the effect of three parameters (data density, flying height, and scan angle) on the quality of buildings extracted. A set of accuracy indices is proposed for this purpose. It is shown that the data density is the most significant parameter affecting the accuracy of building identification. Keywords: Building identification; Hough transform; LiDAR; LiDAR simulator 1. Introduction The significance of 3D geo-information is increasing day by day with new applications being found. Accurate 3D geo-information can help to make important decisions like placement of telecommunication towers, planning of disaster management, land tax evaluation, etc. It also helps in various modelling studies like noise simulation, heat and exhaust studies, pollutant movement, etc. The 3D details of buildings are an important item of geo-information and can be obtained from land surveys, through photogramme-try (aerial or satellite stereo images) or from LiDAR (Light Detection And Ranging) data. Till recently, building extraction was being performed mainly by photogrammetry. However, generation of a digital surface model (DSM) and extraction of buildings using photogrammetry is resource and time consuming. In some cases it may even become impossible to obtain DSM due to high object density, occlusions and object complexity in city areas (Guo and Yasuoka 2002). The advent of LiDAR technology has solved the problem of obtaining DSM as it directly provides 3D coordinates of the ground points. The points provided by laser scanner are dense and accurate. To utilize this data there is a need to develop algorithms to efficiently and accurately extract information from a large volume of data. 5 10 15 20 25 30 35 40 45 *Corresponding author. Email: blohani@iitk.ac.in kothandamk 4/9/07 16:44 TGEI_A_220631 (XML) Geocarto International Vol. 00, No. 0, Month 2007, 1–14 Geocarto International ISSN 1010-6049 print/ISSN 1752-0762 online ª 2007 Taylor & Francis http://www.tandf.co.uk/journals DOI: 10.1080/10106040701207100