also
1 and
rain
jects
ble to
ences
s the
es us
ilities
ICESS,
ator's
ay to
ation,
ns in
mage
ration
ind to
ment
eying
icated
on. It
verall
iment
; and
e real
cially
ot it is
ctions
ystem
yds of
nition
t field
calibration. In any case, when system calibration is applied to
the photogrammetric mapping process, there has to be an
accurately measured test field, but there also have to be various
point classes so that there are targeted points, points on
various terrain surfaces, and details of man-made features. The
need for this kind of mapping test field became obvious during
experimental work in Kaukajárvi area, Tampere, Finland
(Niskanen, 1990, and Salmenperá, 1990).
Calibrated stereomodel of Rusko
There exists a photogrammetric test field for the purposes of
system calibration in the Rusko area, Tampere, Finland, near
Tampere University of Technology (Rantaniemi, 1993, and
Salmenperá, 1993). It has been measured and photographed in
1992. There are 61 control points with a standard deviations of
2 mm and approximately 1600 test points with a standard
deviation of 5 mm for the coordinates and the heights. The test
points are located on terrain surfaces or on man-made features.
The area has been photographed with Wild RC20 (c = 153.19
mm) on Kodak MS2448 and Agfa AvChr 200 films. The flying
height is 500 m, which gives the scale 1:3300. This scale is
usually converted to a mapping scale of 1:500 in urban areas
in Finland. Forward motion compensation has also been used,
which is necessary to produce high-quality images for large-
scale mapping. The images are available in digital form, too.
As the test field exists as these images, it is called the
calibrated stereomodel of Rusko.
The images form two stereomodels. One of them is used for
research and system calibration and it has approximately 1200
points. The other one is used for training purposes and it has
400 test points. The resolving power of the camera/film
combinations varies from 25 to 30 lines/mm as evaluated from
test figures. The geometric accuracy of the stereomodels has
been tested by bundle block adjustment, where the image
coordinates of all the targeted control points are adjusted with
known ground coordinates. The resulting standard deviation of
the unit weight equals 4.5 pm, which corresponds to 15 mm on
the ground.
TEST RESULTS
Testing of stereo-operators
The stereo-operator has to identify, classify and select points
which belong to model surfaces, lines and objects, and then he
or she has to point at them. Operators seem to have their
personal ways of doing this, which results in systematic errors.
Especially heights and unsymmetric objects show these kinds
of systematic errors. The calibrated stereomodel can be used
to test their capabilities to perform these tasks and to classify
them.
The operator starts the test by internal and external orientation
of the stereopair. Then he or she goes over to computer-
controll so that the measuring program brings the measuring
marks near the test points. In height measurement, the
operator has to set and record the height only, but when all
the coordinates have to be measured, the given position is set
randomly aside and the operator has to set the measuring
marks in all dimensions. Before the end, part of the
measurements are repeated. After one day's work the
487
measurements are processed and the operator's performance is
evaluated. In this stage long-term statistical means are used as
reference (table 1).
Identification of terrain points
Good definition of the points to be measured is the basic
condition for achieving high and uniform accuracy in a
photogrammetric survey. To define a point in a satisfactory
manner, ground objects or features must have symmetrical
form, suitable size and sufficient contrast with the surrounding
area (Blachut, Chrzanowski and Saastamoinen, 1979). Very
few natural or man-made ground features meet these
requirements. Relevant factors are also the scale of the
photography, the resolving power of the camera/film
combination and image motion compensation. Furthermore,
colour photos usually give better interpretability than black-
and-white photos, and details can be recognized more easily if
they cast shadows. On the other hand shadows may cover
details. And, to make the matter still more complicated, the
definition of terrain points to be measured is a matter of
subjective judgement with the stereo-operator.
Another side of this question is the role of quality control.
According to Finnish instructions, photogrammetrically
measured coordinates have to be checked against field
surveyed coordinates, using a random sample (The Finnish
Society of Photogrammetry and Remote Sensing, 1993).
Because of identification problems, the instructions define the
concept of the range of interpretability. If the deviations in the
photogrammetrical coordinates from the field surveyed
coordinates are less than the range of interpretability , the
deviation is set to zero, or taken as errorless.
Dimensional stability of photographs
Uneven heating of the photographs can be a problem when
using some analytical stereoplotters (Salmenperä and
Miettinen, 1993). For example fluorescent tubes have high
surface temperatures, they illuminate only part of the
photographs and the area of intense heating moves as the work
proceeds. This gives rise to local deformations. The way the
photos are placed on the picture carriers below the cover
glasses may strengthen these deformations. Then the pressure
of the cover glass may at certain spots hinder the deformations.
The resulting deformations are then both local and time-
dependent and may exceed tens of micrometres (figure 1). In
addition, the deformations at the locations of the fiducial
marks are generally not representative of the deformations
throughout the photograph. This means that corrections for film
distortion cannot be computed reliably from the measurement
of the fiducial marks unless special actions are taken. It is
possible to reduce these deformations. One way to do this is to
use a light source which is placed outside the plotter. Another
means is to let the temperature of the photos reach the working
temperature before the cover glasses are tightened. In
addition, one could produce cover glasses which would be
slightly lifted from the picture carriers, while still keeping the
photographs flat. In the future the expanding use of digital
images will improve the photogrammetric survey process in
this respect.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996