; Walk-
f North
for Vis-
> ACM,
play Al-
lization
3 Com-
orithms
Ph.D.
idimen-
us einer
Vermes-
c versus
ngs. In:
tomatic
1 Space
Model-
nterface
res L.F.,
D Geo-
t 95-14,
Design
or Data
ISPRS,
ation of
)5 ACM
pp. 95-
Prentice
a Dis-
an ATM
‘ess, pp.
k-Based
ISPRS,
he city:
. In GIS
>man J.,
or Archi-
1904.
als Ba-
ions-Sy-
Multiple
. In Pro-
jal Envi-
EXPERIMENTAL STUDY OF OPTIMAL DIGITAL MAPPING PARAMETERS
Kurt Kubik, Director, Space Centre for Satellite Navigation
Queensland University of Technology, Australia
Peter Harvey, Director, CSIRO SuperComputer Centre, Australia
Inter-Working Group IVIII - Digital Photogrammetric System
KEY WORDS: Softcopy, Photogrammetry, Accuracy, DEM/DTM
ABSTRACT
The paper evaluates optimal performance parameters for digital photogrammetric mapping. This mapping method is a relatively
new technique and the relationship between input parameters (flying height, pixel size and compression method) and quality of
results is as yet insufficiently known. The present paper documents a statistical evaluation of this relationship. The tests with the
VirtuoZo digital photogrammetry system demonstrated spot height accuracies of 1 pixel for pixel sizes up to 25 microns and 0.3 to
0.5 pixel for larger pixel sizes. The effect of JPEG compression for up to a factor 8 was insignificant.
KURZFASSUNG
Fuer Digitale Photogrammetrie wird der Zusammenhang zwischen Eingabe Parametern (wie Flughoehe, Pixelgroesse und
Kompressionsmethode) und der Genauigkeit des Resultates untersucht. Zu diesem Zweck wurde ein Test Feld in unterschiedlicher
Flug Hoehe ueberflogen und die Photo Bloecke mit dem digitalen System VirtuoZo ausgearbeitet. Es zeigte sich dass die Hoehen
Genauigkeit fuer Pixelgroessen bis zu 25 Mikron gleich zu 1 Pixel ist, und fuer groessere Pixelgroessen gleich zu 0.3 - 0.5 Pixel.
JPEG Kompression bis zu einem Faktor 8 hatte keinen merkbaren Genauigkeitseinfluss.
manner that the map is produced economically and to
specifications. Currently, these parameters are chosen
empirically, and inefficiencies exist by using high resolution
scanners while subsequently applying lossy compression, where
the same end result could probably be achieved with cheaper
low-resolution scans and lossless compression of the image.
Lossy compression results in pixel values of the reconstructed
image which may differ from the original ones (Wallace, 1991).
The relationships in choosing these parameters are currently not
well understood, and experimental work was required in order
to guide our further theoretical studies. In particular, image
compression is required in the production environment, as one
single digital image may already occupy some 400 Mbytes and
several hundred images may be necessary within one mapping
project. The compression should however not counteract costly
accuracy gains, which were obtained by using high resolution
scans and at low altitudes.
1. INTRODUCTION
The use of aerial stereo photography for cartographic map
production is an accepted standard. During the last decennium,
digital photogrammetric mapping was introduced, which uses
digital images, fully automated 3-D shell reconstruction from
the stereo images and automated derivation of Digital Elevation
Models (DEM) and contour maps. The following parameters
have to be chosen in the execution of the process, which
strongly affect both the quality and costs of the product:
* Flying height/image scale, at which the serial photography
is captured: The lower one flies, the more photographs are
required to cover an area and thus the cost increases, but
objects are visible in greater detail;
* Pixel size of the digital image: The smaller the pixel size in
the digital scanner used for scanning the images, the more
expensive is the scanner, and the larger is the recorded data
set, but the details are more accurately recorded;
* Compression method and ratio: With less compression,
large data sets remain which are costly to handle, but
accuracy is preserved.
The experiments were done on equipment available at the
Centre, such as the Helava or the VirtuoZo digital
photogrammetric systems. The data sets for the experiments
were aerial photography of a test field of appr. 4km*4km at
Queenslands Department of Primary Industry’s (QDPI)
Gayndah site in Central Queensland. The terrain at this test site
can be described as flat to rolling pasture land, with height
variations up to 20 metres. Tufted grass 20 cm tall was
abundant. A narrow water course with some steeply eroded
areas and stands of trees traversed the area. Image 1 shows a
3-D view of a representative section of the test area, which was
We are especially interested in the performance of these
parameters in connection with VirtuoZo, the Digital
Photogrammetry System developed at Wuhan Technical
University of Surveying and Mapping (WTUSM) with some
contributions from Queensland University of Technology
(QUT) and marketed through VirtuoZo Systems Pty. Ltd.
(Zhang, 1995). These parameters should be chosen in such a
203
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996