246
To be economically acceptable, DEM generation
should provide accuracies matching those of manual
plotting (i.e. between 0.1-0.2 %o of the relative flight
height h r ) in a shorter time. Since the point
measurement rate is much higher than that of the
human operator (and bound to increase further with
processing power), large DEMs with hundreds of
thousands points may be generated in an hour time or
less, depending on hardware and software
performance. As far as accuracy is concerned, figures
range from 0.1 to 0.8 h r (Bacher, 1998; Balsavias and
Kaeser, 1998, Duperet, 1995), therefore not always
matching manually produced DEM. Off-line editing is
necessary, clearly pointing out that self-diagnosis tools
are not as reliable as they should. Indeed, in large scale
photogrammetry and problem areas the percentage of
points to edit may be as large as 10% or more, making
it a key feature of the system the performance of the
editing tools available to the user.
Any DEM generation system has self-diagnosis
capabilities at the matching level as well as at the
interpolation level. At the former, systems mostly
provide a quality index of each matched point which
can be as simple as the correlation coefficient
(confronted with a terrain and/or image dependent
threshold) or may be based on a classification scheme,
reporting or providing hints in case of failure, in more
sophisticated systems. Overall, though, they are not
always reliable, since they may label as good wrong
points or may discard them as wrong when they are in
fact good ones. At the interpolation level, self-
diagnosis is understood here mainly as the capability of
filtering out non-terrain objects, rather than blunders in
the matching process, which may be best dealt with by
geometric constraints supporting the matching.
Despite this drawbacks, the advantage of automation is
spreading the use of automatically generated DEM.
Recently we have been involved in a project in the
framework of the management of tens of marble
quarries in northern Italy, in the province of Brescia.
The goal of the investigation is to check the accuracy
level of automatic DEM generation in these areas and
what kind of project parameters (image resolution,
scanning accuracy, additional terrain information,
matching techniques) are best suited. This paper
reports on the results of DEM generation in the
repeated survey of a rock quarry, to figure out the
volume of material excavated, using aerial images.
This information is used in planning the exploitation of
the quarries by the authorities in charge; accuracies of
some percentage point on the volume’s estimate are
acceptable. A direct control of the excavation volumes
to fix taxes has been for the time being ruled out, since
weigthing the lorries in and out of the quarry is more
accurate.
2. THE PILOT PROJECT BOTTICINO
The project goals were set as follows;
1) to study which selection of the flight parameters
would be the best so that the performance of DEM
generation could be improved;
2) to study the accuracy of the DEM and of DEM
changes, with respect to DPW system (scanner as
well as software) and what gains could be made by
first measuring manually a low resolution DEM.
The marble quarry in Botticino, a small village near
Brescia, was chosen as a test area, because a recent
aerial survey was available and because it is quite
representative of the quarries in the area; they are
located in steep hills, with height differences from top
to bottom up to 400-500 m; their fronts range from
several hundreds meters to more than 1 km.
The excavation proceeds in stages, cutting the hill side
in banks 10 to 30 m high, slightly inclined downhill;
though more or less you may recognize a main front
running parallel to the contour lines, large blocks are
extracted here and there, leading to a pattern of
“holes”. Debris is spread around in several areas,
depending on which front is currently active. The hills
around the quarries are covered by bushes and trees,
while the excavation area is rather bright, resulting in a
scene with an overall high dynamic range but (though
not everywhere) a low contrast within the quarry, in
full daylight and clear sky.
The first flight, executed in summer 1997, consists of a
single strip with 4 images (numbered 49 to 52) at the
average scale 1:5600 and lead to the compilation of a
1:1000 map of the quarry, which is contained in the
model 50_51. Although the flight was flown around
midday, sharp shadows are projected from the banks,
making sometimes difficult to identify the exact
location of the base of the walls.
2.1 The new flight plan
As already mentioned, the second part of the project is
supposed to complete a new flight over the same area,
to highlight volume changes in the time span and to
allow an accuracy evaluation thanks to a topographic
survey.
The new flight plan has been designed in order to
improve the accuracy of automatic DEM generation by
proper choice of the flight parameters. Three aspects
have been taken into account: flight height and camera
focal lenght, endlap and sidelap values along and
across strip and finally strip direction versus terrain
morphology.
For a constant image scale, a lower flight height means
using larger focal lenghts, increasing areas prone to
occlusions and perspective distortions. A longer focal
lenght would reduce the occlusions and also help
image correlation to account for perspective
differences; if image scale is taken constant, this would
on the other end change the H/B ratio, worsening the
elevation accuracy. Larger image scales, compare to
the well established standard values adopted in
mapping with analytical plotters, would undermine the
economies obtained by digital methods, as only for the
larger surveying work implied. Based on this
reasoning, either image scale and focal lenght have
been retained to the values of the preceeding flight.
Though it is hard to figure out their effect in advance,
experience shows that occlusions and terrain
discontinuities degrade the accuracy of DEM generated
