xsslıc moomoo: pat xsl
XSSJTC "eec? pat xs2
Xss$3r¢ cucooc? pat xs3
. Producing a ground control points file (xtie.ctp) by
control points coordinates to satellite images. Our
reference for ground coordinates was the existing map
of area.
The control points file (xtie.ctp) is shown in figure 4.
iion of
37 were
hophoto
saicking.
Os Were
| area of
isplayed
rd copy,
y in the
Mage of
adapted
sensitive
p placed
e been
and the
des are
1ask and
digitized
order to
unique
g to the
IN
images
vot have
Line* Col* x! Y4 Delta. L& Delta, C&
287 349 840050 3172900 -0.178833 -0.012420
194 385 841225 3174575 0.147858 -0.131226
139 222 838263 3176275 -0.138992 0.399384
158 162 837000 3176138 -0.008987 -0.418976
243 211 837538 3174300 0.178940 0.163208
figure 4 - Control points file
- Correction of control points for relief displacement.
For doing that we used CTPTODTM program:
Usage:
CTPTODTM inctp outcoe dtm alpha col ref
Inctp: Control points from ILWIS (.ctp)
outcoe: Transformation coefficients for 'imtodtm'
(.coe)
dtm: ILWIS DEM file (.mpi and mpd)
alpha: Looking angle in degrees
o: Vertical, pos: East
col: Image column to which alpha applies
ref. Reference height (no relief displacement)
The actual command which we entered was
CTPTODTM xtie outcoe pat dtm 5.6 -351 150
The looking angle (5.6) and the column number (-351)
have been extracted from the original image
descriptions.
4.1.1 Geometric correction with relief displacement.
We did the geometric correction for relief and
transformation to the selected coordinate system
(existing map of area) by use of GCD program:
Usage: GCD res in out coe dtm alpha col ref [pix]
or
Usage: GCD res in out coe pix
res: resampling method
n: nearest neighbour
b: bilinear
in: patched input image (.mpi and .mpd)
out: patched output image (.mpi and .mpd)
coe: coefficient file (.coe)
dtm: patched ILWIS DEM file (. mpi and .mpd)
alpha: looking angle
0: vertical
pos: east
col: image column to which alpha applies
ref. reference height (no relief displacement)
pix: output pixel size (m)
539
The actual commands that we entered were:
GCD n pat xsl pat s1 outcoe pat dtm 5.6 -351 150 20
GCD n pat xs2 pat s2 outcoe pat dtm 5.6 -351 150 20
GCD n pat xs3 pat 53 outcoe pat dtm 5.6 -351 150 20
4.1.2 Geometric correction without relief
displacement. We executed GCD program with the
original images and the orginal control points to
produce the anaglyph vision:
GCD n pat xsl pat_slst xtie 20
GCD n pat xs2 pat s2st xtie 20
GCD n pat xs3 pat s3st xtie 20
Then we unpatching all images to ILWIS format by use
of UNPATCH program:
UNPATCH pat sl sl
UNPATCH pat_s2 s2
UNPATCH pat_s3s3
UNPATCH pat slst slst
UNPATCH pat s2st s2st
UNPATCH pat s3st s3st
4.1.3 Anaglyph stereo vision. We produced a color-
composite image by use of COLORCMP program in
ILWIS:
COLORCMP slst sl sl $; or
COLORCMP s2st s2 s2$; or
COLORCMP s3st s3 s3 $;
After completion of one of these commands, and with
help of a pair of anaglyph glasses, we are able to see the
stereo model of area on the screen.
4.2 Classification
4.2.1 Train and test sampling. We divided the ground
truth into two separate sets, one for training and one for
testing the classifiers.
4.2.2 Crossing and error analysis. Using the program
"crossing", cross tables were generated between the test
samples, e.g, TEST3, and their corresponding
geometrical corrected conventional classified satellite
image. Confusion matrices are shown in figure 5.
1 2 3 4 5 6 7 8 uncl ACC
1 § 235 0 0 0 0 0 0 1 97 0.71
2]o0 153 ^ Q g o9 9 B 9 0 1.00
310 0 0 0 2 0 0 10 3192 | 000
410 0 0 0 A 62 0 0 119 | 0.00
510 120560 051419 [32 2 X9 | 0.33
6/0 73 0 0 41 0 0 0 6 0.00
7 | 28 3 Ü 0 0 0 2 0 2 0.04
slo 0 0 2 8 0 0 à 70 0.00
5
REL | 089 069 ? 0 0 055 1 0.08
average accuracy = 26.00 %
average reliability = 31.67 %
overall accuracy = 32.31 *
figure 5 -Confusion matrix of Maximum Likelihood
classifier with zero majority filtering
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996