ration techniques.
On the shortest
inica, 14(10), pp.
rnational Journal
997. Hierarchical
ingle Perspective
ttern Recognition
. 870-883.
Rucklidge, W. J.,
ff Distance. [EEE
'hine Intelligence,
S., 1996. Object
" Transactions on
Y(3), pp. 267-277.
Hypothesentests in
chology. Harcourt
and fast pattern
| inspection. Real-
ation and Visual
nd Company, San
> objects using the
>
f Computer Vision,
clusion, clutter, and
; Mustererkennung
‚er, Berlin, pp. 148-
isual motion. MIT
ıd Ebner, H., 2001.
ages for industrial
| 3-D Measurement
irical performance
ds. In: Empirical
. I. Christensen and
Society Press, Los
ation in Perceptual
hology, W. D. Ellis
3. On the Role of
e Vision, Jacob Beck
Is), Academic Press,
COMPARISION OF DIFFERENT SENSOR TYPES AND VALIDATION OF AN
APPROACH FOR MULTI SENSOR FUSION
A. Wendt, C. Rosing, M. Weisensee
Institute for Applied Photogrammetry and Geoinformatics, University of Applied Sciences
Oldenburg/Ostfriesland/Wilhelmshaven,
Ofener Strafe 16/19, D-26121 Oldenburg, Germany - (a.wendt, weisensee) (a) fh-oldenburg.de
Commission V, Working Group V/1
KEY WORDS: Data Acquisition, Comparison, Combined Adjustment, Multi-Sensor Fusion, Surface Reconstruction, Accuracy
ABSTRACT:
In the presented paper an approach for data fusion and simultaneous adjustment of inhomogeneous data is used, which is intended to
increase the accuracy and reliability of surface reconstruction. The theory of the approach and first examples for applications have
been given by the authors beforehand. The aim of the approach is to adjust any kind of data in a combined adjustment and to give
adequate weights to each measurement. Therefore an assessment of the quality of different sensor data is needed.
Thus, a comparison of different sensors is given. A test will outline the differences between surface reconstruction results by stripe
projection and photogrammetry and the differences will be discussed. The validation of the approach for multi sensor fusion for the
reconstruction of surfaces is described and demonstrated by using two different objects from the area of close range
photogrammetry. Both objects represent free form surfaces and are to be captured with an adequate resolution. The first object is an
artificial reference surface used for control purposes and having geometry precisely defined by CAD-data, the second object is a tile
made of concrete. The results of the numerical experiments are interpreted and conclusions for the processing chain of the approach
are drawn.
KURZFASSUNG:
Im vorgelegten Beitrag wird ein Ansatz zur Datenfusion und gemeinsamen Ausgleichung ungleichartiger Messwerte verwendet,
welcher die Genauigkeit und Zuverlássigkeit der Oberfláchenrekonstruktion erhóhen soll. Die zu Grunde liegende Theorie und erste
Anwendungsbeispiele wurden durch die Autoren veróffentlicht. Das Ziel des Ansatzes ist es, verschiedenste Datenarten in einem
gemeinsamen Ausgleichungsprozess zu verarbeiten und dabei jedem Messwert ein adáquates Gewicht zuzuordnen. Dazu ist im
Allgemeinen eine Beurteilung der Qualitát der verschiedenen Sensordaten erforderlich.
Darum wird zunächst ein Vergleich der verschiedenen Sensoren durchgeführt. Ein Versuch zeigt die Unterschiede zwischen der
Oberflächenrekonstruktion mittels Streifenprojektion und Photogrammetrie. Die Differenenzen werden dargelegt. Hierbei wird der
Ansatz der Sensor-Fusion zur Oberflächenrekonstruktion anhand von zwei Objekten aus der Nahbereichsphotogrammetrie erläutert.
Beide Objekte repräsentieren Freiformoberflächen und sind mit einer angemessenen Auflösung zu erfassen. Das erste Objekt ist eine
glatte Referenzoberfläche, welche zu Kontrollzwecken hergestellt wurde und eine präzise definierte Oberflächengeometrie aufweist,
das zweite Objekt ist eine in der Tiefe strukturierte Betonplatte. Die Ergebnisse der numerischen Experimente werden dargelegt und
interpretiert und Folgerungen für die Oberflächenrekonstruktion mit dem vorgestellten Ansatz werden gezogen.
Distinctions have been made between sensors and methods
primarily giving position data on surfaces in form of a more or
1. INTRODUCTION
In the field of optical measurement techniques, which are
applied in industrial, architectural and medical measuring tasks,
a manifold of different sensor types has been developed and is
used in order to acquire the geometry of surfaces of various
objects. Here, high technology standards of data acquisition are
achieved.
The selection of any measurement technique and sensor type
strongly depends on the special requirements of the respective
task concerning the size and environment of the object, time
required for data acquisition and plotting, robustness,
resolution, accuracy and reliability of the results and numerous
other factors.
t Corresponding author.
less dense cloud of 3D points, e.g. stripe projection and laser
scanning, and those registering radiometric information of the
material in form of two or more digital images from which the
geometry has to be determined, i. e. digital photogrammetry.
These distinctions loose their meaning, as further improvements
in sensors and systems deliver positions in 3D space as well as
radiometric information about the surface using either active or
passive scanning systems.
Besides the obvious application of such radiometric information
for texturing object surfaces in a visualisation there are several
other benefits concerning the requirements of measurement
tasks mentioned above. Here, of greatest importance seem to be
that the fusion of independent clouds of points is strongly
—105-
