QUALITY ASSESSMENT OF MAPPING BUILDING TEXTURES FROM INFRARED
IMAGE SEQUENCES
L. Hoegner, D. Iwaszczuk, U. Stilla®
Technische Universitaet Muenchen (TUM), Germany
Commission III Working Group 5
KEY WORDS: thermal imagery, relative orientation, image sequences, texture extraction
ABSTRACT:
Generation and texturing of building models is a fast developing field of research. Several techniques have been developed to extract
building geometry and textures from multiple images and image sequences. In this paper, these techniques are discussed and
extended to automatically add new textures from infrared (IR) image sequences to existing building models. In contrast to existing
work, geometry and textures are not generated together from the same dataset but the textures are extracted from the image sequence
and matched to an existing geo-referenced 3D building model. The texture generation is divided in two main parts. The first part
deals with the estimation and refinement of the exterior camera orientation. Feature points are extracted in the images and used as tie
points in the sequence. A recorded exterior orientation of the camera s added to these homologous points and a bundle adjustment is
performed starting on image pairs and combining the hole sequence. A given 3d model of the observed building is additionally
added to introduce further constraint as ground control points in the bundle adjustment. The second part includes the extraction of
textures from the images and the combination of textures from different images of the sequence. Using the reconstructed exterior
camera orientation for every image of the sequence, the visible facades are projected into the image and texture is extracted. These
textures normally contain only parts of the facade. The partial textures extracted from all images are combined to one facade texture.
This texture is stored with a 3D reference to the corresponding facade. This allows searching for features in textures and localising
those features in 3D space. It will be shown, that the proposed strategy allows texture extraction and mapping even for big building
complexes with restricted viewing possibilities and for images with low optical resolution.
1. INTRODUCTION
The analysis and refinement of buildings in urban areas has
become an imported research field in the last few years (Pu &
Vosselman, 2009; Mayer & Reznik, 2006). Buildings in city
models normally consist of simple facade structures and optical
textures. To refine and automate the extraction of geometry and
textures in urban scenes different solutions have been proposed
during the last years (Mayer, 2007; Heinrichs et al., 2008; Lo &
Quattrochi, 2003; Pollefeys et al., 2008). Those strategies are
extracted geometry and textures together to form a new building
or facade and are not merged with an existing geo-referenced
city model.
Much urban information cannot be extracted from normal
optical images but from other optical domains like infrared.
Ground cameras are recording the irradiation of building
façades (Klingert, 2006), for the specification of its thermal
behaviour for thermal building passports. IR data of buildings
are collected from several photos and analysed directly in the
acquired images without any geometric or 3d processing.
The focus of this paper is the integration of pre-known building
models in the texture extraction process from image sequences
to improve the textures' quality. There are two reasons to add
infrared textures to existing building models instead of
generating a building model and extract the textures from the
same infrared image source. The geometric resolution of
infrared images is quite low compared to images in the visible
domain and geometric features like edges show different
appearances which lead to problems in correct extraction of
facade planes. For many buildings, accurate building models
from databases like CityGML or a building information model
(BIM) already exist and the given model has to be improved.
Infrared textures are an additional data source for the extraction
of windows (Iwaszezuk et al, 2011; Sirmacek et al., 2011).
Thermal cameras record electromagnetic radiation in the
invisible infrared (IR) spectra. Thus, surface characteristics of
object can be detected, that stay invisible in normal visible
spectra. For recognition of objects with little difference in
temperature and for identification of small details from distance,
thermal cameras must be able to resolve temperatures with an
accuracy of 0,01 Kelvin. High-quality infrared cameras are able
to record image sequences with standard video frame rate (25
fps) or even higher. Because of special cooling technique,
camera optics and the low production numbers, the expenses of
infrared cameras are very high compared to normal video
cameras. Today, thermal image data is used for many different
applications. Typically, IR data of buildings are collected from
photos and analyzed directly in the acquired images. Bigger
building parts are acquired by combining several images. The
results of the analysis are stored in the 2d photos without any
geometric or 3d processing. This can be a problem, when
images from different cameras or views are combined and
stored for further processing.
In contrast to conventional IR inspection of buildings, in this
paper an automated strategy is used for texturing an entire
building model. Narrow streets and the low resolution and small
field of view of IR cameras are serious problems. Only small
parts of the building façade are visible in one image. Direct line
matching of the images and the projection of the model's edges
is used especially for aerial images (Frueh et al., 2004; Avbelj et
al., 2010) fails because of the lack of visible facade edges in
many of the images. In this case we have to deal with the fact,
that structures found in IR images do not always have
correspondences in the building model and vice versa. This
1h
Ns GN —— mM
m CT o
