ul 2004 
g linear 
clusions 
errors 
be used 
MSE is 
less of 
iges in 
as been 
affine 
X axis). 
stinctive 
ition of 
lometric 
r image 
ocedure 
EPSON 
; stated 
digital 
able for 
f 12 um 
Iso, it 1s 
canning 
g rather 
s. 
or large 
ry, they 
uracy is 
rs could 
Hebert 
OEEPE- 
mmetric 
r RMI. 
immetric 
GEOMETRIC PRECISION OF SCANNED IMAGERY FOR PRODUCTION 
PHOTOGRAMMETRY 
Andrew Calarco*, Alex Dam? and A. Stewart Walker? ' 
“ Leica Geosystems GIS & Mapping, LLC, 61 Inverness Drive East, Suite 200 
; Englewood, CO 80112, USA 
“ Leica Geosystems GIS & Mapping, LLC, 10840 Thornmint Road, Suite 100 
San Diego, CA 92127, USA 
(andy.calarco, alex.dam, stewart.walker)@gis.leica-geosystems.com 
Commission I, WG I/2 
KEY WORDS: Scanning, Digital, Film, Photogrammetry, Error, Precision, Workflow 
ABSTRACT: 
Little research has been performed to determine the quantitative effects of the geometric imprecision of scanned photographs in a 
photogrammetric production setting. While photogrammetric scanners are most commonly used to scan aerial photographs, desktop 
and graphic design scanners have also been applied. It is recognized that the precision of non-photogrammetric scanners is highly 
variable and their errors larger than those of photogrammetric units, but these values have seldom been quantified. In addition, the 
practical effect of these errors upon the photogrammetric solution has not received much study. 
A precision calibration plate was repeatedly scanned on photogrammetric and non-photogrammetric scanners to determine the 
variability and magnitude of errors. These errors were plotted spatially to illustrate the inherent instability of non-photogrammetric 
scanners. These models of scan accuracy were then used to determine the effects of the random scan errors in a photogrammetric 
production environment. While appropriate for qualitative uses, ultimately it was observed that the characteristic unpredictability of 
the errors rendered this imagery unsuitable for quantitative applications. 
During the investigation a significant degree of consistency was observed in the non-photogrammetric scanners, i.e. the errors, 
though large, included a systematic component that did not vary through time and changing conditions. Therefore a small block was 
scanned in such a scanner and the imagery re-sampled using the calibration parameters. The results were compared with those 
obtained from the same block scanned in a photogrammetric scanner. 
1 INTRODUCTION 
With the introduction of airborne digital sensors, much 
attention has shifted away from traditional means of image 
capture. However, at this point only a very small percentage 
of airborne imagery is collected from digital sensors with the 
overwhelming majority of images being subjected to the 
conventional softcopy methods of data processing. Images 
are collected with a metric aerial camera, processed, and 
converted to digital format on a film transparency scanner 
for use in photogrammetric production. 
The choice of film scanner is critical since any errors 
introduced during scanning will be carried through the 
photogrammetric workflow and can negatively influence the 
results obtained during aerial triangulation, Digital Terrain 
Model (DTM) extraction and registration of the rectified 
imagery to existing datasets. 
1] Justification of Study 
A distinction should be made between photogrammetric 
(PG) scanners and non-photogrammetric (NPG) scanners in 
the market. PG scanners typically meet very robust 
standards in terms of the optical alignment, throughput 
capability and the accuracy of analog image conversion to 
digital form. NPG scanners are available at a fraction of the 
cost of PG units, and this has led to their periodic use in 
photogrammetric production. While NPG scanners are 
typically able to produce imagery of high radiometric 
quality, uncertainty remains regarding the ability to 
consistency produce digital imagery of high reliability. 
There is no mechanism to calibrate the movement of the 
scanner stage / sensor line and therefore no quantitative 
values exist to the appropriateness of the imagery for 
photogrammetric measurement. 
The purpose of this paper is to explore and compare the 
achievable precision of imagery scanned on PG and NPG 
film scanners. Should the results show markedly different 
results between the units tested, the quantification of 
attainable precision may serve to preclude the use of NPG 
scanners in photogrammetric work. In an effort to establish 
the precision of both PG and NPG scanners, a fixed 
methodology was established. 
1.2 Research Design 
There are three types of error: systematic, random and 
egregious. Of the three, only the distinction between 
systematic and random errors was focused upon.