In: Wagner W., Szekely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part7B 
374 
whose coordinates in both TWD67 and TWD97 were known, 
were selected as check points. 
By comparing the registered and the known coordinates of the 
check points in TWD97 datum, an rms value for the coordinate 
differences is of about 11 cm in both x- and y- coordinates 
(Table 1). This indicates that the results of the registration 
method developed here are satisfactory for the subsequent 
application. 
Difference 
Ax(m) 
Ay(m) 
max 
0.326 
0.309 
mean 
0.048 
-0.030 
rms 
0.109 
0.110 
Table 1. Coordinate differences of check points. 
The quality of reconstructed building models is evaluated by 
manual check. In Figure 5, a closed polygon represents a 
building. Our results have shown that in total 34 of 108 
buildings are incorrect building models showed with gray 
polygons in Figure 5 using our automatic procedure. 
Figure 5. Incorrect building models (gray polygons) in the test 
area. 
These incorrect polygons always arise on certain cases. These 
cases can be divided into two categories: (1) Roof faces are 
partly or fully covered by neighbouring buildings or trees, as 
shown in case A in Table 2. (2) Building outlines are not 
detailed enough, as shown in case B in Table 2. In cases B, 
small buildings are inside a big building, but the outlines of the 
small buildings are not drawn in the map. In all these cases, 
boundary LiDAR points are not sufficient to match with 
building outlines. This shows that these discrepancies between 
the boundary points and the outlines of that building influence 
the results of the reconstruction of building models. All of these 
incorrect models should be refined manually, by 
photogrammetry, or even field work, for instance. The complete 
result of automatic reconstruction of 3D building models is 
drawn in Figure 6. 
Table 2. Fusion of LiDAR data and topographic map after 
registration and the results of building model reconstruction. 
Figure 6. Complete result of automatic reconstructed building 
models. 
5. CONCLUSIONS 
This study has presented a novel method to construct building 
models by fusing LiDAR data and topographic map information. 
A procedure for registering the boundary points of building roof 
patches extracted from LiDAR data and outlines of buildings 
acquired from topographic maps has been proposed by a robust 
least squares method. 
The experiments have shown that the proposed method for the 
building reconstruction procedure with LiDAR data and 
topographic map information, including feature extraction, 
registration and reconstruction, can be processed automatically 
and yields good results. Although manual editing is needed in 
order to achieve refined 3D building models, the results have 
shown that our method takes advantages of both surfaces and 
boundary information and improves the building reconstruction 
process. 
6. REFERENCES 
Filin, S., 2002. Surface clustering from airborne laser scanning 
data. International Archives of Photogrammetry and Remote 
Sensing, 34(3A), pp. 119-124. 
Filin, S., Y. Kulakov and Y. Doytsher 2005. Application of 
Airborne Laser Technology to 3D Cadastre. FIG Working Week 
2005 and GSDI-8, Cairo, Egypt. 
Gruen, A. and D. Akca, 2005. Least squares 3D surface and 
curve matching. ISPRS Journal of Photogrammetry and Remote 
Sensing, 59(3), pp. 151-174.