affine distortions. The other scanners deliver great differences
between affine error and Helmert errors. Consequently this
factor may be an additional criterion in an assessment of a
scanner’s geometric quality. A large RMS error difference be-
tween Helmert transformation and affine transformation leads
to the the conclusion that the evaluated scanner produces
a difference in pixelsize between x and y directions. When
scanning with the single line principle this may be caused by
unstable or incorrect speed of the moving CCD.
7.6 Repeatability measurements
The repeatability of geometric accuracy results for a DTP
scanner and a high performance film scanner is assessed in
Table 4. For the repeatability measurements every subse-
quent test scan has to be executed at the same and highly
stable conditions. Therefore it is essential to place the test
target for every scan at exactly the same position on the
scanning region when doing the repeatability test. Further
possible vibrations of the scanner caused by the scanners en-
vironment, which can lead to additional unwanted digitizing
errors, have to be avoided.
Scanner trans- res. Mean Stdev
form. of RMS of RMS
error error
um dpi um / pixel um
DTP helm. 63.6/400 108.0/1.7 0.3
scanner aff. 63.6/400 17.0/0.3 0.3
high perf. helm. 8.5/3000 0.5/4.7 0.1
scanner aff. 8.5/3000 0.4/3.2 0.01
Table 4: An investigation of measuring repeatability for a
DTP and a high performance scanner. Geometric accuracy
evaluation is taken for 10 mutually independent scans, where
the test target for every scan is placed exactly on the same
position.
The repeatability of the investigated DTP scanner seems to
be very good. The Mean of RMS errors and its standard
deviation are computed for 10 mutually independent scans of
the geomeric accuracy test target. The evaluation is based
on a Helmert transformation as well as for affine transforma-
tion. It is evident that the geometric error produced by the
DTP scanner has a large stable portion and is therefore highly
reproduceable. The remaining random variation of the error
is in this case very small. Consequently an increase of geo-
metric accuracy properties of the DTP scanner could possibly
be achieved by the use of a calibration grid and built in cali-
bration procedures. However the maximum resolution is too
low to use the scanner as a photogrammetric device. When
investigating the data delivered by the high performance geo-
metric accuracy scanner, the change of geometric RMS error
vectors between subsequent images is infinitely small, which
shows the high geometric performance of this type of scanner.
8 CONCLUSION
A system for the geometric accuracy evaluation of scanners is
presented. It can be used to obtain the geometric properties
of various types of scanners, from the low cost DTP to high
performance film devices. The analysis is possible by man-
ual visual or fully automated methods using algorithms and
achieves an accuracy which is sufficient for detection of local
186
as well as global geometric behavior. It can be shown that
the precision of the scanner investigation is highly dependent
on the appearance of the underlying geometric accuracy tar-
get. Visual analysis may result in a bias by a specific human
analyst. Therefore, and for reasons of economy automated
algorithms are prefered to produce test results. Of course vi-
sual analysis will continue as an important part of a scanner's
evaluation because visual analysis is the only way to detect
visual artifacts. Scanner evaluation results may by used for
comparison between scanners as well as for detection of long
term stability and can serve as a basis of geometric scanner
calibration. The experimental results show the usability of the
proposed algorithms implemented in the SCANEVAL system
for geometric accuracy evaluation.
9 ACKNOWLEDGMENT
This work was partly supported by the 'Osterreichische Na-
tionalbank Jubilaumsfond’, project 5821.
REFERENCES
[Baltsavias, 1994a] Baltsavias, E.P., 1994. The Agfa Horizon
Scanner - Characteristics, Testing and Evaluation, Pro-
ceedings ISPRS, Vol. 30, Como, pp.171-179
[Baltsavias, 1994b] Baltsavias, E.P., 1994. Test calibration
and procedures for Image Scanners, Proceedings ISPRS,
Vol. 30, Como, pp. 163-170
[Baltsavias, 1994c] Baltsavias, E.P., Bill, R., 1994. Scanners
a survey of current technology and future needs, Proceed-
ings ISPRS, Vol. 30, Como, pp. 130-143
[Bethel,1994] Bethel, J., 1994. Calibration of a photogram-
metric Imaging Scanner. Proc. ASPRS, Reno, Nevada, pp
81-88
[Leberl, 1992] Leberl, F, Best, M., Meyer, D., 1992. Pho-
togrammetric Scanning with a Square Array CCD Cam-
era, International Archives of Photogrammetry and Re-
mote Sensing, Washinton D.C., Vol. XXIX, Comm. ll,pp.
795-800
[Maalen, 1993] Maalen-Johansen, 1.,1993. On the precission
of subpixel measurements in videometry, Opticai 3D Mea-
surement Techniques Il, WICHMANN, pp. 169-178
[Seywald, 1994] Seywald, R., 1994. Requirements of a sys-
tem to analyze Film scanners, Proceedings ISPRS, Vol. 30,
Como, pp 144-149
[Simonis,1991] Simonis, Roland, 1991. Scanner analyzer tar-
get, Proceedings SPIE,Beam Deflection and Scanning
Technologies, Vol. 1454, pp. 364-369
[Trinder, 1987] Trinder, J.C., 1987. Measurements to Digi-
tized Hardcopy Images, Photogrammetric Engineering and
Remote Sensing, Vol. 53, pp. 315-321
[Trinder, 1989] Trinder, J.C., 1989. Precision of Digital Tar-
get Location, Photogrammetric Engineering and Remote
Sensing, Vol. 55, pp. 883-886
[West, 1990] West, G.A.W., Clarke, T.A.,1990. A survey and
examination of subpixel measurement techniques, Proc.
SPIE, Close-Range Photogrammetry Meets Machine Vi-
sion, Vol. 1935, pp. 456-463
[Wolber, 1991] Wolber, J. W., 1991. Measuring the geomet-
ric accuracy of a very accurate, large, drum type scanner.
SPIE Proc. Camera and Input Scanner Systems, Vol. 1448,
pp. 175-180
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B1. Vienna 1996
Till
proc
moti
syste
data
attitu
freqi
In
track
anot
Furt
Fig
