Zheng Wang 
  
BUILDING EXTRACTION AND RECONSTRUCTION FROM LIDAR DATA 
Zheng Wang 
EarthData International 
Gaithersburg, Maryland 
USA 
zwang @earthdata.com 
  
Tony Schenk 
Department of Civil Engineering 
The Ohio State University 
Columbus, Ohio 
USA 
aschenk @magnus.acs.ohio-state.edu 
Commission III, Working Group 3 
KEY WORDS: Building Extraction, Building Reconstruction, Object Recognition, LIDAR. 
ABSTRACT 
This paper presents an approach for building extraction and reconstruction from high quality terrain surface, such as 
LIDAR data generated surface. The approach takes terrain surface data as input and goes through edge detection, edge 
classification, building points extraction, TIN model generation, and building reconstruction to extract and reconstruct 
buildings and building related information. For building detection, it detects edges from the surface data and classifies 
edges to distinguish building edges from other edges based on their geometry and shapes, including orthogonality, 
parallelism, circularity and symmetry. The classified building edges are then used as boundaries to extract building 
points and TIN models are generated with the extracted points. Each building has its own TIN model and its surfaces are 
derived from the TIN model. The test results demonstrate that the approach is capable to extract and reconstruct 3-D 
building models from high quality terrain surface. The paper presents the concept, algorithms and processes/procedures 
of the approach and discusses its advantages and limitations. The paper also shows experimental results and discusses 
the performance/effects of the geometric and shape constraints utilized in the experiments. At the end, the paper 
concludes with some thoughts for future development. 
1. INTRODUCTION 
Over the last decade, many research and development efforts have been put into extracting and reconstructing building 
from images (mono and stereo) and DEMs (Haala and Hahn 1995, Schenk 1995, Henricsson et al. 1996, Weidner 1996, 
Fritsch and Amen 1998). However, the contribution from DEMs, especially DEMs generated through image matching, 
has been very limited (Moons, 1997), because of the poor quality of such kind DEMs. Building reconstruction is a 3-D 
modeling process, therefore 3-D information should be used whenever available. The lacking of independent and reliable 
elevation information has made the task more difficult. Several years ago, LIDAR (LIght Detection And Ranging) data 
started to become available for researchers. It didn’t take long for them to realize the great value in LIDAR data and use 
LIDAR data in building extraction and reconstruction (Ackermann 1996, Haala et al. 1998, Wang 1998). The high quality 
of LIDAR data guaranties the reliability of the provided 3-D information. The high quality of LIDAR data is reflected in 
several aspects: 1. high accuracy. A typical LIDAR system can provide data withl5 cm vertical accuracy and less than 
50 cm horizontal accuracy, 2. High consistency of the accuracy, i.e., the accuracy is same everywhere, and 3. High 
consistency in coverage, i.e., points are evenly distributed in the covered area. This paper presents an approach for 
building extraction and reconstruction using LIDAR data generated surface. The approach was developed by Zheng 
Wang partially for his Ph.D. dissertation. 
The approach utilizes 3-D corners and their relative orientations/associations to reconstruct or model a building. At 
current stage, approach only extracts and reconstructs buildings that have right corners and parallel egdes. The 
approach first detects edges or contours from a surface data (in elevation image format) and then uses shape and 
geometrical properties, including symmetry, circularity, orthogonality, and parallelism to classify edges/contours into 
building and non-building edges/contours, respectively. Since most buildings have simple/regular shapes, which other 
natural features (such as trees) do not have, shape has been selected as one of the properties to separate buildings from 
other objects. After the edge classification, the approach makes a TIN model for each building with a group of 3-D 
points. Each group of 3-D points is extracted by using the corresponding building edge as boundary, which means all 
the points in the group belong to the same building. Then, the TIN model of each building is used to generate surfaces 
of the building, where the surfaces are represented by 3-D planes. Finally, the tri-intersections of the surfaces form the 3- 
D building corners and their relative orientations/associations. 
  
958 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.