probability
ae.
the case for
| total of 6L
1erefore, for
sulting form
it algorithm,
m 6L to 6.
e redundant
inity of the
to allow the
flexibility in
dancy is less
ocess whose
mage staring
ters from 6L
r that uses a
ixel, which is
ituation, it is
»ach scan line
yroom model.
litional frame
ey or a GPS
triangulation
(x,y) are line
; are abundant
gery, point to
oround, linear
round control
includes such
veral different
ts. At a given
is defined by
ve center.
t 1995 (Figure
he straight line
waviness".
Edward M. Mikhail
The second data set, flown over the
urban area of Fort Hood, Texas in
October 1995, is shown in (Figure 8).
Its ground sample distance is 2.2
meters, respectively. Its flight height
was about 4430m. As can be seen
from Figure 8, straight roads along the
in-track direction are severely wavy.
Figure 7 HYDICE Imagery, Washington, DC.
Using the first Gauss-Morkov model,
the orthorectified images
corresponding to Figures (7) and (8)
are shown in Figures (9) and (10)
respectively.
Figure 8. HYDICE Imagery, Fort Hood
3 URBAN FEATURE EXTRACTION
3.1 Hyperspectral Analysis
Remote sensing techniques have been used for many years for the classification of multispectral imagery. However,
until recently, most of this type of imagery did not have sufficiently fine spatial resolution to make it useful in urban
environment. Now imagery such as HYDICE and HyMap discussed in section 2 offer excellent urban data. Figures 7
and 9 showed the original and
orthorectified 3-color images of an
airborne hyperspectral data flightline
over the Washington DC Mall.
Hyperspectral sensors gather data in a
large number of spectral bands (a few
10's to several hundred). In this case
there were 210 bands in the 0.4 to 2.4
m region of the visible and
infrared spectrum. This data set
contains 1208 scan lines with 307
pixels in each scan line. It totals
approximately 150 Megabytes. With
data that complex, one might expect a
rather complex analysis process,
however, it has been possible to find
quite simple and inexpensive means to
do so. The steps used and the time
needed on a personal computer for this analysis are listed in Table 3.1 and described as follows:
‘Figure 10. Ortho-rectified Image (Gauss-Markov, Fort Hood)
Define Classes. A software application program called MultiSpec, available to anyone at no cost from
http://dynamo.ecn.purdue.edu/-biehl/MultiSpec/, is used. The first step is to present to the analyst a view of the data set
in image form so that training samples, examples of each class desired in the final thematic map, can be marked. A
simulated color infrared photograph form is convenient for this purpose; to do so, three bands are used in MultiSpec for
the red, green, and blue colors, respectively. (See Figures 7 and 9).
Feature Extraction. After designating an initial set of training areas, a feature extraction algorithm is applied to
determine a feature subspace that is optimal for discriminating between the specific classes defined. The algorithm used
is called Discriminate Analysis Feature Extraction (DAFE). The result is a linear combination of the original 210 bands
to form 210 new bands that automatically occur in descending order of their value for producing an effective
discrimination. From the MultiSpec output, it is seen that the first nine of these new features will be adequate for
successfully discriminating between the classes.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 897
