ES a crie
ci i
_
MULTISPECTRAL ANALYSIS OF DIGITAL IMAGE DATA IN ARCHITECTURAL PHOTOGRAMMETRY
Robert Godding
Institut fir Photogrammetrie
und Bildverarbeitung (IPB)
Technische Universitdt Braunschweig
Giesbert Sacher
Meßbildstelle, Gesellschaft für
Architekturphotogrammetrie mbH
Dresden
Gunnar Siedler
Institut für Photogrammetrie
und Fernerkundung (IPF)
Technische Universität Dresden
Germany
Commission V
ABSTRACT
A combination of the digital methods of close range photogrammetry for the determination of object geome-
try and remote sensing for a multispectral determination of object attributes by using classification
algorithms is useful for various applications. The image data can be recorded directly with a digital
device, for example with a low resolution CCD-camera or the high resolution Réseau-Scanning-Camera.
Additionally, an analogous recording on film with a later digitisation is possible. The different spec-
tral bands are selectable by using different filters during the image recording.
First results of a multispectral and multisensor data acquisition and evaluation in an architectural
application will be presented.
Key Words: digital multispectral image data, image acquisition systems, filters, classification
1. INTRODUCTION
Multispectral image analysis allows the determi-
nation, localisation and quantitative recording
of object attributes by analysing their special
emissions and reflections with multispectral
recording systems.
The main application of multispectral image anal-
ysis is the wide field of remote sensing. The
idea of terrestrial use was derived from geologi-
cal applications. Terrestrial multispectral data
recording has been applied since 1980 for open
cast mining and other geological sample sites.
Then, first analysises of buildings were effected
where the distinction of different building mate-
rials was the major field of interest.
Simultaneously, the experiences made in the field
of painting analysis were of use in this sector.
Photos in different spectral ranges were taken
for damage anaysis although the data material was
mainly evaluated visually.
At present, the spectrum used in close range
applications reaches from X-rays to the thermal
infrared light. In most applications, spectral
information from 350nm to 2000nm is applied. The
improvement of digital image processing methods
together with the availability of different kinds
of digital sensors for the image acquisition and
more powerful computers allow a wide range of new
applications and the development of new methods.
In the following, some possibilities of digital
multispectral image acquisition and evaluation as
well as results of a survey of an architectural
object will be presented.
2. RECORDING AND EVALUATION SYSTEMS
2.1 Cameras and Scanners
The recording of digital images is possible with
CCD-cameras on-line, or by digitising analogous
photos off-line.
The on-line usage of usual CCD-sensors with
approximately 768 x 512 pixel is restricted to
applications which only require a low resolution.
Digital cameras with larger solid state sensors
and a higher resolution are very expensive or are
not available on the civil market [Luhmann,
1991].
For this reason, the realisation of higher reso-
lutions and larger image formats is reached by
using different techniques, which in most cases
are based on the recording of parts of an entire
image with a moving sensor. The precise orienta-
tion of the partial images relative to the entire
image can be achieved mechanically (for example
with a piezo-unit at the micro-scanning camera)
or optical and numerical by using the réseau
technique (réseau-scanning) [Wester-Ebbinghaus,
1989].
For the digitising of analogous photos, several
instruments are available on the market. Here as
well, different scanning techniques are known.
Most scanners work with a CCD-line or with a CCD-
matrix sensor and the required accuracy can be
reached with a high precision mechanism or with
optical and numerical methods as used in réseau-
scanning.
For the image acquisition of photos, the use of
metric cameras or middle-format réseau cameras is
most efficient.
