Image Processing Reference
In-Depth Information
Table 6.3
Selected morphological applications of the V-I-S Model
Spectral
bands
Spatial
resolution (m)
Author(s)
Location
Year
Data
Techniques
Chung
Salt Lake
City
1989
SPOT-X
3
20
Subpixel,
visual
Rashed et al.
Cairo
2001
IRC-1C
4
24/10
Subpixel,
SMA
Card
Salt Lake
City
1993
TM
6
30
Subpixel,
visual
Phinn et al.
Brisbane
2002
TM, photog
6
30
Subpixel,
SMA
Ridd et al.
Pasadena
1997
AVIRIS
224
20
Per pixel,
ANN
6.4.1
SPOT-X Data: Spatial Variation in Urban Communities
Chung (
1989
) completed the first study using the V-I-S model in Salt Lake City. An
early use of SPOT-X data with its then-new 20 m resolution required precise spatial
rectification, especially to achieve sub-pixel identification. Key features of this
investigation are:
•
Large scale CIR photography for land cover calibration
•
Land survey monuments for ground control
Large scale CIR (color infrared) photography (1:4,600) provided very high quality
ground information. To fix the photos to ground control points, state plane coordi-
nates of road intersections were obtained and converted to a UTM projection.
Beginning with a 60-group unsupervised classification, print character maps in
transparency form were overlaid on the CIR photos. Each pixel was large enough
(about 1 cm) that the cover type could be accurately determined at the center of
each pixel quadrant.
Seven V-I-S cover types, plus water, were sought: Vts - trees and shrubs, Vg -
green grass, Ia - asphalt, Ic - concrete, Ir - roofs, Sl - light soil, Sd - dark soil, and
W - water. From 776 random pixels a total of 3,104 point observations were
recorded. Table
6.4
displays the composition of four sections of the city for the
cover types. Note the transition from industrial, to commercial (CBD), to a near
town residential area (Liberty), to a low density suburban area (Sugar). Vegetation
increases from 4%, to 10%, 47%, and 57% respectively, while impervious surface
is essentially a reciprocal: 96%, 90%, 50%, and 37%, respectively. Soil varied
between 0 and 5%.
Figure
6.9
displays a Red vs. NIR (x,y) spectral feature space plot (excluding the
brightest signatures), indicating the V-I-S composition of each spectral cluster.
Note the transition from 100% non-vegetation, mostly impervious, to 100% vegeta-
tion along each trend line. Note also that class 24 is 80% water, 20% impervious.
Figure
6.10
employs colors to group the original 60 clusters into V-I-S related
