tes of spot
ılated stereo
Units are in
tes of spot
rolled stereo
s. Units are
t elevations
om the first
aw that the
significant.
\gation and
1e results of
n the stereo
angulation.
runs and the
iking is the
, than in the
Table 5: Comparison of the spot elevations taken from
the first and fourth multi-sensor triangulation results.
Units are in feet.
Table 6: Comparison of the spot elevations taken from
the third and fifth multi-sensor triangulation results as
compared to the GPS controlled stereo models. Units are
in feet.
4. CONCLUSIONS
Both conventional triangulations produced more accurate
results than any of the multi-sensor triangulations. The
conventional aerial triangulation using the GPS surveyed
points as the ground control provided the most accurate
solution from which to compile the topographic maps.
The RMS error of the control points in this triangulation
was 0.558, 0.445, and 0.237 feet in X,Y, and Z,
respectively. But to achieve this level of accuracy
required 14 control points. We were also able to achieve
better results in a conventional aerial triangulation using
control points collected from a USGS 1:24,000 scale
topographic map sheet. Using 19 ground control points,
we achieved an RMS error in the control of 17.056,
20.664, and 14.43 feet in X,Y,and Z, respectively.
But based on the comparison of the spot elevations
collected from both the conventional and multi-sensor
triangulation generated stereo models, we saw that the
aerial photography can be controlled to support
topographic mapping within U.S. National Map Accuracy
679
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
Standards. For a 1:24,000 scale map, the National Map
Accuracy standards for horizontal positioning is 40 feet
CE 90%, and the height accuracy is 10 feet LE 90% given
a 20 foot contour. If we attempt to map at a 1:10,000
scale from the 1:24,000 scale photography, we can still
attain the horizontal positioning necessary, but the height
accuracy is outside the standards, assuming a 10 foot
contour interval. These levels of accuracy were
attainable using only 5 well distributed ground control
points within the SPOT stereo model, and just tie points to
the aerial photographs.
But in order to attain this level of accuracy the project
must be set up with the following factors:
(1) The project must use SPOT stereo pairs, and the
aerial photographs need to be tied to both SPOT images
through common tie points. If just a single SPOT image
is used, or if the photographs are tied to just one of the
SPOT images the accuracy in Latitude and height are
degraded significantly.
(2) The SPOT stereo pair must have a sufficent
convergence angle between their look angles. At lower
convergence angles, the solution of the Z component of
the stereo models is degraded.
(3) At a minimum, it appears that the block of aerial
photography needs to be tied to the SPOT imagery around
the perimeter of the block. This was evident from the
significant decrease in accuracy when we just tied one
corner of the block to the SPOT imagery, and the lack of
any significant change when we just tied the corners of
the block to the SPOT imagery. The decrease in accuracy
in Strip 7 also indicates that the accuracy of this approach
will decrease as the block extends away from the SPOT
imagery.
In areas where existing, up-to-date topographic maps of
suitable scale exist, it would be more cost effective to
digitize the control from the maps and create the stereo
models for compilation through a conventional aerial
triangulation. But in areas where the maps do not exist, or
are not current, this multi-sensor approach can provide a
method of creating accurate stereo models for compiling
small-scale topotgraphic maps. This was just a
preliminary study. Further study will include applying the
approach to smaller scale photography, using a more
accurate control point source, and bridging a larger block
of photography between two SPOT stereo models.
