Frame ground coverage (GW) = 1760/0.4 = 4400 feet
Recommended average flying height (H) = 0.5 GW/tan20 = 6044
feet
Thirty percent side lap between alternate flights ^ 0.3 x 4400 —
1320.0 feet
Pixel ground resolution = 4400/2000 = 2.2 feet
B/H = 1760/6044 = 0.3
In order to have a priori estimate for the contour interval possible
with the digital camera at the calculated parameters the following
photogrammetric criteria were utilized (Light, 1990)
Piel size 5 HKxB/H 10; bom noiimmaih-datinto A Round (1)
where:
K = non-dimensional number expressing the degree to which
Stereocorrelation can be achieved. (Typical bounds for K are
0.2<K<1.5)
0, = precision of a single height observation which is assumed to
be one third of the contour interval (CI) (0, = 1/3(CI)).
B/H = the base to height ratio.
Substituting B/H=0.3 and K=0.3 (or the parallax errors=0.3 pixel
size) in equation (1) will result in the following value for CI.
Pixel size = 1/0.3x 03x03 x CI =
or CI 4 2.272032 73'- 22m
This figure served only as a priori estimate but the possible
contour interval should exceed the expectation.
2.3 Data Acquisition
The digital aerial photo was acquired during May 1995. The
project area is located at Desloge, Missouri and it is characterized
by variation of terrain patterns from the very steep slopes along
the river banks to rolling and flat areas. The digital frame camera
was installed on board the Photo Science, Inc. Piper Navajo
Chieftain plane. Four flight lines and a supplementary line were
flown at about a 6500 foot flying height which resulted in a slight
increase in endlap and sidelap coverage. The data was then
processed at the ground facility of Daedalus Enterprise, Inc. at
Ann Arbor, MI, where about 100 black and white digital images
were obtained in a TIFF format ready for the softcopy
workstation processing. A few days prior to the flying, the
ground surveying crew established an accurate network of nine
control points using GPS techniques and the already established
NGS points. Each ground control point was marked with an
aerial photography target in the shape of two crossed panels one
and a half feet in width and six feet in length.
2
2.4 Digital Photo Triangulation
A total of 24 frames that covered the testing site and a total of
eight paneled ground control points were used to form the block
of photos that is later oriented and triangulated on the
Autometric-Vision International softplotter. Considering all the
problems that are inherent in the interior parameters of the DFC,
the stability of the triangulated block was of concern. Therefore,
a smaller block of seven frames with a suitable number of ground
controls was used for the purpose of this analysis.
The interior orientation was achieved by measuring the pixel of
the four corners and the image coordinates system origin was
assumed to be at the point of intersection of the two axial lines
connecting the opposite corners. Although this assumption is not
free from error, it was necessary as no accurate information was
available about the principal point location.
2.5 Image matching autocorrelation, and surface modeling
The triangulated pairs were then correlated and a digital elevation
model was generated using the available tools of the softplotter.
The generated surface was then modeled using both the terra
model and the Intergraph MGE Modeler. A set of two foot
contours was generated. Figure 1 shows a clip from the generated
contours.
3. RESULTS AND DISCUSSION
In examining the pixel resolution on the softcopy station, the 0.7
m pixel ground coverage seems to be adequate for revealing
ground details. The ground targets were easily distinguishable.
Due to an appreciable amount of error propagation in the larger
block which resulted from camera interior instability, number and
distribution of control points (as some of the control points
disappeared prior to flying due to local land use activity), a
smaller block of seven digital photographs was triangulated and
autocorrelated to produce the digital elevation model (DEM).
Examining the obtained DEM and the derived contours reveals
a great agreement with the set of contours derived conventional ly
from a photogrammetric stereoplotter. However, it is found that
there is a bias in the digital surface generated from softcopy
which is clearly reflected in Table 1. The table shows the data
from twenty points selected on a variety of terrain patterns that
were measured stereoscopically from the three-dimensional
digital stereopair displayed in the softplotter and the same
locations were selected and measured in the surface generated by
conventional photogrammetry. It is noticed that discrepancy is
reduced in magnitude as the evaluated point is closer to a control
point. Some other points deviated from the root mean squares of
error or bias and that is believed to be either because of the
changes in the ground surface as there were a few years between
the two mapping dates, or it is because of some trees or brush
covering the ground surface.
It is also noticed that the discrepancy pattern changes
considerably at the step slope at the river banks as the correlation
technique faces more ambiguities due to slope relief and dense
tree coverage.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B1. Vienna 1996
The stu
frame
photog
develo}
technic
The stu
digital
a mapp
param
proved
4000x
essenti
triangi
proces
to enh:
The au
offerec