Paul Pope
3.3 Pre-processing
3.3.1 Reference Image and DEM Creation: The reference image and DEM required by this method were created
from data provided to the authors by two different agencies. The reference image was created from panchromati,
digital orthophotos of the study area. The DEM was derived from another DEM of the study area. The digital
orthophotos have a smaller ground sample distance (GSD) than the source DEM. Therefore, a nominal GSD of |g
meters was chosen for this study. This value lies between the GSD's of these two raster data sets, and is equal to tj,
nominal GSD of other airborne scanner imagery of interest to the authors.
The digital orthophotos were obtained from the Wisconsin Department of Natural Resources. The format of these dig
is TIFF with ancillary TFW (georeference information) files. These images are tiled by township, have a GSD of |
meter, are georeferenced to the Wisconsin Transverse Mercator (83/91) coordinate system (SCO, 1995), and have |
byte per pixel quantization. A 1:24,000 scale digital elevation model (DEM) for this area was obtained from the Unite]
States Geological Survey (USGS). The 7.5" quadrant name of this DEM is "Cross Plains, WI". The format of they
data is the Spatial Data Transfer Standard (SDTS). The DEM has a GSD of 30 meters, is georeferenced to UTM Zope
16 (NAD 1927), and has 2 bytes per pixel quantization. Elevations are in units of feet, and are referenced to the NGVD
of 1929. The overlap between the digital orthophotos and the DEM were used to define study area extents expressedi
WTM (83/91) coordinates. The upper left and lower right corners of the study area were set to (N 294,440 [m], E
540,400 [m]) and (N 284,400 [m], E 550,400 [m]), respectively.
These data were pre-processed by using a hybrid GIS called TNT Mips V6.2 (Microlmages, Inc., Lincoln, Nebraska)
The digital orthophotos and DEM were imported to the GIS. The digital orthophotos were mosaicked, resampled tos
GSD of 10 meters using cubic convolution, and subset to match the study area extents. The output raster was used
the "reference image". The USGS DEM was resampled to a GSD of 10 meters using cubic convolution, elevations
were rescaled from feet to meters, and the raster was subset to match the study area extents. The output raster was used
as the DEM. The minimum and maximum elevation of this DEM are 253 meters and 373 meters above MSL
respectively. Each raster contains 10,000 rows by 10,000 columns of pixels. These raster data were then exported t
simple array files (no header; rows stored sequentially), accompanied by ancillary georeference and size informatim
files (Figure 1).
El
(a) hs
Figure 1. The reference image (a) and the DEM (b) of Cross Plains, Wisconsin.
tx
3.3.2 Spatial Autocorrelation Analysis: Spatial autocorrelation will reduce the ability of the PMIIM method ©
synthesize unique signals to be used in the image-to-image matching process. This will in turn reduce the ability of the
method to resolve the necessary trajectory corrections. However, it can aid significantly in reducing the computation
workload. A spatial autocorrelation analysis was conducted on the reference image by using a minimum correlation
value of 0.7. The average lag value was 1.3 pixels with a standard deviation of 0.47 pixels. This means that on averag
a unique scan line (i.e. correlation value less than 0.7) is not encountered until 1.3 scan lines away from any particular
scan line. This effect is symmetric about a scan line. Therefore, unique scan lines can be synthesized only if te
average distance between their ground traces are 2*1.3 + 1 = 3.6 GSDs away from each other. This value is called tht
734 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.
"spi
pert
3.3.
Ap]
(NA
is d
of A
Tw«
the
num
the
33.
in a
to
5,00
met:
at tl
met
at th
each
to 7:
3.3.
repr
mea
traje
seco
gene
aeriz
and
to 1
Phas
The
degr
of fi
devi
rise
inde,