You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

Systems for data processing, anaylsis and representation

Scott Miller and Alex Dam
Helava Associates Inc., 10965 Via Frontera, San Diego CA 92127
Digital photogrammetric products can be produced from virtually any type of imagery. Presently, most digital
photogrammetric products are produced from scanned diapositives. The image scanner can typically be the most
influencing factor on the accuracy of data collection. A Digital Photogrammetric Workstation (DPW) cannot recover
from a poor quality scanner. This paper discusses methods for geometrically calibrating scanners and suggests that
a standard should be made.
KEY WORDS: Scanner, Accuracy, Standards, Digital Image
Photogrammetric products can be produced from
virtually any type of image. The image can be from hand
held 35 millimeter cameras, video cameras, satellite
imagery, panoramic cameras and many others. The
accuracy of products produced is largely affected by
the input source. There is a general expectation that
photogrammetric products produced from high quality
mapping cameras, with distortions in the micron range,
would produce products commensurate with input image
quality. If we are to use scanners to digitize high quality
films for photogrammetric data production, we must be
careful to preserve radiometric and geometric quality if
we are to compete with existing photogrammetric
products such as analytical stereo plotters and high
quality comparators. If one does not wish to preserve
the fidelity of the input image and is cognizant of this
(i.e., it is not required) then there is no problem. If one is
not cognizant, and believes they are getting a product
equivalent to the film camera output, then there is a
serious problem. What standards are required? Should
we have a rating system? |n this paper, we are limiting
our discussions to geometric quality, there is of course
a very related issue of radiometric quality.
Today, many firms view photogrammetry in terms of C-
factor or similar rules. Through experience and
practice, firms learn what is achievable from a given
instrument type.
C-factor = Flying Height of Image / Contour
Interval of Map
Analytical plotters are typically achieving in the 1200 to
well over 2000 range for C-factor. Stage accuracy's for
first order analytical plotters are generally better than 3
microns. For the most part, a DPW should be limited in
accuracy only by the quality of pixels given to it and the
quality of control points used. Assuming the pixels are
of good quality, algorithms and the human can easily
measure to better than 0.5 pixels. Many studies prove
that precision better than 0.25 pixels are routinely
possible. To obtain accuracy, we need not just
precision, but absolute knowledge of where those pixels
are with-respect-to the original film focal plane of the
camera. Much has been done in photogrammetry to
assure good geometric fidelity of film when placed on
the stage plate of an analog or analytical plotter. Why
should photogrammetry require camera calibrations if
the scanner does not require it? A scanner should
preserve and adhere to the quality of analytical plotters
and/or should state the accuracy to be expected under
typical or standard operating conditions. If the
scanners geometric quality is as good as the analytical
plotter stages, then a user's C-factor concepts can be
used for DPWs as well (assuming the pixel radiometric
quality and size is good enough). If we scan an image
with a 25 micron pixel size, and assume that
measurements can be made to 0.25 of a pixel, we get a
desired precision of 6.25 microns. This tells us that our
scanner geometric error should be significantly less
than 6 microns to achieve the best results and not
unduly influence accuracy from a 25 micron pixel scan
General Scanner Types And Error Sources
Geometric qualities of scanned film primarily results
from the scanner design of which there are several
types. Stage types of scanners have a big impact on
scan quality. Drum scanners that primarily are derived
from the graphics art industries, typically use tension to
hold the film flat against the drum. This combined with
drum "roundness" make it difficult to maintain focus and
geometric position in relation to the scan head. Flat bed
stage designs can use hard optically flat cover plates to
hold the film flat during scanning. This is an easier
design to hold to higher geometric accuracy. Flat bed
stages with cover plates are less susceptible to
operator error in mounting the film. Scanners designed
for a typical application are normally designed with
sufficient accuracy for that application. In the graphic
art applications, desk top flat bed and drum scanners
are typical. Their accuracy is generally sufficient for the
application they were designed for. The Sharp JX-600
uses a flat bed design but does not use a flat cover
plate and does not use a high precision stage drive or
low distortion lens. Its geometric accuracy is in the 80
micron range (Sarjakoski 1992). It produces reasonably
good pixels and could be rated for photogrammetry,
perhaps by stating its limitation on C-factor.
Some sc:
or glass
Some use
4 or 8 fic
points wil
method si
The follov
the geom
en PR zn m pou
£1 T0 0—
A manufe
factors co
produce 1
can supp
image ac
A standa
Some sce
fidelity. T
analyze o
provide th
others us
quality. T
same thii
Scanned i
Besides tl
checked :
assure an
the sensoi
to the ca