In: Wagner W., Szekely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
371 
BUILDING MODEL RECONSTRUCTION WITH LIDAR DATA AND 
TOPOGRAPHIC MAP BY REGISTRATION OF BUILDING OUTLINES 
B. C. Lin 1 *, R. J. You 2 , M. C. Hsu 3 
Department of Geomatics, National Cheng Kung University, 1 University Road, Tainan City, Taiwan - 
1 p6889102@mail.ncku.edu. tw 
2 rjyou@mail.ncku.edu.tw 
3 p66984095@mail.ncku.edu.tw 
KEY WORDS: Tensor Voting, Feature Extraction, Registration, Robust Least Squares, Data Fusion 
ABSTRACT: 
This study integrates LiDAR data and topographic map information for reconstruction of 3D building models. The procedure 
includes feature extraction, registration and reconstruction. In this study, the tensor voting algorithm and a region-growing method 
with principal features are adopted to extract building roof planes and structural lines from LiDAR data. A robust least squares 
method is applied to register boundary points of LiDAR data with building outlines obtained from topographic maps. The 
registration accuracy is about 11 cm in both x- and y- coordinates. The results of the registration method developed here are 
satisfactory for the subsequent application. Finally, an actual LiDAR dataset and its corresponding topographic map information 
demonstrate the procedure for data fusion of automatic 3D building model construction. 
1. INTRODUCTION 
The needs for building models are growing rapidly in 3D 
geographic information system (GIS), and hence a large number 
of accurate building models have become necessary to be 
reconstructed in a short period of time. Recent developments in 
airborne LiDAR have made it a new data source for 3D 
building model reconstruction, since LiDAR can quickly 
provide a large number of highly qualitative point clouds to 
represent building surfaces(Maas and Vosselman 1999). 
However, the LiDAR data has poor texture information so that 
the accurate building boundary extraction from LiDAR data 
may be difficult. Therefore, data fusion involving both LiDAR 
data and the existing topographic maps can improve the 3D 
building model reconstruction process. 
A number of researchers have studied the problem of feature 
extraction from LiDAR data to reconstruct 3D building models 
(Vosselman and Dijkman 2001; Filin 2002; Overby et al. 2004). 
In general, building roof patch features are first extracted from 
LiDAR data. Many methods (Filin, 2002; Maas and Vosselman, 
1999; Overby et al.,2004) can be used for the extraction of roof 
patch features from LiDAR data. Next, building models are 
reconstructed by combining the building boundaries obtained 
from ground plans and intersection lines of adjacent planar 
faces derived from LiDAR data. 
These approaches, however, may produce unreliable results in 
3D building model reconstruction if the coordinate systems of 
LiDAR data and the ground plans are not the same. To 
overcome the problem of coordinate systems of various data 
sources, data registration is a critical step for fusion of LiDAR 
data and the topographic map information (Schenk and Csatho 
2002; Filin et al. 2005; Gruen and Akca 2005; Park et al. 2006). 
In this study, plane segments in LiDAR data are extracted in the 
feature space based on the tensor voting computational 
framework (Medioni et al. 2000). The tensor voting algorithm 
implements features such as faces, lines and points through a 
symmetric tensor field directly derived from data. All geometric 
structures (surfaces, lines and points) can therefore be inferred 
simultaneously. This method also offers extra information about 
the strength of features which can indicate the main geometric 
characteristic of a point. 
For registration of LiDAR data and topographic maps, a robust 
least squares method (RLS) is adopted to estimate the 
transformation parameters in this study. After registration, 
height information and roof ridges extracted from LIDAR data 
are introduced to topographic maps and then the spatial 
positions of building outlines can be reconstructed. 
In the following, the tensor voting method and the registration 
method are first described. Finally, an experimental result based 
on an actual airborne LiDAR dataset is analyzed. 
2. FEATURE EXTRACTION 
2.1 Tensor communication 
The geometric feature of a point can be described by a second- 
order symmetric tensor which is expressed as follows: 
0 
0' 
"V 
0 
0 
e 2 
0 
0 
¿3. 
A. 
(1) 
* Corresponding author.