653 
FACADE DETAIL FROM INCOMPLETE RANGE DATA 
Jan Böhm 
Institute for Photogrammetry, Universität Stuttgart, Germany 
jan.boehm@ifp.uni-stuttgart.de 
Commission V, WG V/4 
KEY WORDS: Range Data, LIDAR, Building, Façade, Modelling, Matching 
ABSTRACT: 
In recent years, the demand for highly detailed building models has clearly risen. Most of the details, which characterize a building, 
come from its façade. It is obvious that such façade detail is ideally acquired from ground-based sensors. For any street-level data 
acquisition system, a common problem arises: that of occlusions and hence incomplete data. In the past we have shown how multi 
image coverage can overcome occlusions in image based façade recording. In this paper, we demonstrate new approaches on façade 
detail from range data in the case of incomplete data. The approach builds on our development of LASERMAPs, a simple and 
efficient way to use street-level LiDAR data to enhance existing prismatic building models. 
1. INTRODUCTION 
The impressive advances in the automatic generation of virtual 
city models achieved within the photogrammetric community in 
recent years, have contributed to the widespread dissemination 
of such models to various end-user platforms. So called ‘digital 
globes’ have become an especially popular platform for LODI 
(level-of-detail 1) models, which feature simple block models 
without roof structures. Lately L0D2 models have been 
released, which feature detailed roof structures, while the body 
remains a simple prism enriched by colour texture. See (Kolbe 
& Grôger, 2003) for a detailed definition of the LOD hierarchy. 
However, the use of highly sophisticated computer equipment 
and enormous manpower in the modem media and 
entertainment industry, which can be experienced in stunning 
computer graphics special effects in movies or highly detailed 
three-dimensional video games, has further raised the 
expectations of today’s audiences to the quality of computer 
generated content and visualizations techniques. These 
expectations can not be fully met by today’s city models. 
Furthermore the dominant business concept behind digital 
globes, the advertisement market, requires the spectacular and 
recognizable presentation of local businesses within the models. 
It is therefore foreseeable, that the level of detail has to be 
further increased and the time for widespread dissemination of 
LOD3 models is near. Increased detail can only come from 
three-dimensional façade detail, which in turn can only be 
achieved by ground-based sensors. An overview of how to 
create detailed 3D models for buildings in cities from terrestrial 
data is given in (Mayer et al., 2008). 
We have developed an approach which efficiently combines the 
coarse geometry of an existing LOD2 building model with the 
detailed features form ground-based LiDAR data which is based 
on our concept of LASERMAPs. In the case of a LOD2 
building model, the approach is straight forward, since the 
coarse geometry is described by a polyhedral boundary 
representation and facades are typically planar polygons. This 
property is also exploited when intensity images are used for 
texture mapping. In order to use a similar 2D-2D mapping for 
the integration of terrestrial LiDAR data, we have to derive a 
two-dimensional representation of the point cloud. Obviously, 
this is not possible for the whole point cloud. We rather have to 
split the point cloud into groups with respect to the facades of 
the building using a simple buffer operation for each facade 
polygon. The portion of the point cloud that belongs to a 
particular facade can then be interpolated into a regular raster, 
in a fashion very similar to digital elevation models derived 
from aerial LiDAR. 
In ground-based façade scanning, as in any ground-based data 
acquisition, incomplete data acquisition is a major problem and 
a primary reason for insufficient data quality. Incomplete data 
acquisition can occur for multiple reasons. One cause is the 
partial occlusion of the façade by other objects, such as cars, 
trees, pedestrians, street signs and so on. Figure 1 shows an 
extreme case for an occlusion by a tree and the incomplete data 
it causes. Another cause is the self occlusion of the façade due 
to an oblique viewing angle. Protruded balconies or indented 
windows will cast shadows along the direction of measurement 
and cause incomplete data acquisitions. While such effect can 
be minimized by proper station planning, we have to keep in 
Figure 1. Incomplete ground-based LiDAR data due to an 
occlusion by a tree. The example is taken from a dataset 
shown figure 2Figure 2.