uw
i ft, ve AU:
SURFACE RECONSTRUCTION FROM DATA OF DIGITAL LINE CAMERAS
BY MEANS OF OBJECT BASED IMAGE MATCHING
H. Diehll, C. Heipke?
! Department ZTA 21
Messerschmitt-Bólkow-Blohm GmbH
Post Box 80 11 09, D-8000 Munich 80, Germany
Tel: + 49-89-607 28651; Fax: + 49-89-607 25157
2 Chair for Photogrammetry and Remote Sensing
Technical University Munich
Arcisstr. 21, D-8000 Munich 2, Germany
Tel: + 49-89-2105 2671; Fax: + 49-89-280 95 73; Telex: 522854 tumue d
E-mail: heipke@photo.verm.tu-muenchen.de
Commission III
ABSTRACT:
This paper deals with automatic surface reconstruction
from data of digital line cameras in the framework of
object based image matching. In this approach grid
heights of a digital terrain model (DTM), the elements
of exterior orientation and grey values of an orthoimage
are simultaneously determined from the recorded image
grey values in a least squares adjustment. In order to
provide initial values for the unknowns a hierarchical
image representation is used.
For the investigations simulated satellite images from a
high mountain area (the Vernagtferner in the Austrian
Alps) recorded with a 3-line camera are used. The ele-
ments of exterior orientation are assumed to be given for
every set of three scan lines. The influence of white noise
added to the grey values of the three image strips and of
random errors in the exterior orientation is investigated.
The results show that this approach can tolerate a large
amount of noise in the image (up to a standard deviation
of more then 10 grey values), but is very sensitive to
errors in the exterior orientation.
Key words: Image matching, 3-line camera, object sur-
face reconstruction, algorithm
287
1. INTRODUCTION
In recent years the main topic in photogrammetric re-
search was the investigation of digital or softcopy pho-
togrammetry. There exist two different aspects: on the
one hand, classical photogrammetric methods are supp-
lemented or replaced by digital ones, and on the other
hand new possibilities are discovered, which are especi-
ally suited for digital data.
This article deals with a task of digital photogrammetry:
the extraction of height information from line imagery
by means of object based image matching. There already
exist a large number of investigations, publications and
implemented procedures for image matching. Therefo-
re, we first present a brief review of the different me-
thods.
The methods of analytical photogrammetry have at first
been transferred directly to digital photogrammetry, the
computer had simply replaced the human operato:.
That is, the computer had to find conjugate points in
stereo image pairs, but the subsequent calculations were
still the same, regardless whether the image points had
been found by man or machine.
For point matching well known methods from statistical
mathematics have been developed like correlation of
grey value distributions /Sharp et al. 1964; Helava 1976;
Panton 1978; Hannah 1988/. Subpixel accuracy was
