Ammatzia Peled
TTT
aphic and were implemented. These differ by the gray level groups, the “estimated” ranges and the filter size. These analyses
Stem wa resulted with the edges of the building being detected. Yet, the problem was that roads were detected also, in addition
efined by to the building's edges. The best result, was achieved with the following parameters: (a) 16 groups that were
ature type determined automatically, (b) "estimated" ranges that were calculated automatically by the formula:
t with the «Meant2.5*Standard Deviation", and (c) 21X21 filer size.
Tr sections,
Therefore
eters, it jg |
ture class. ion,
generated
20 +
vel values
157
acteristics | 40
statistical
'lemented,
5
e
r all other jt
Figure 1: Histograms of gray level values for the pixels within buildings.
arameters, 2.2.2 “Transportation” Coverage
The “transportation” coverage includes seven different types of roads. According to the distribution of gray level
values, for each road type that fall in “unchanged” regions, it was difficult to define gray level interval for each road
class, uniquely.
Therefore, separate histograms were established for each road segment (see fig. 2 and fig. 3). According to several
filter size. classification processes, the results show a mixed detection of roads and buildings, similar to the “buildings”
classifications. The best result, was achieved with the following parameters: (a) 8 groups that were determined
automatically, (b) "estimated" ranges that were determined manually, and (c) 7X7 filer size.
el groups,
TYPE-CODE = 106
BO reson
yverage #7 40
rages.
30 f /
/ J.A e\ |
20 / J| Y
; tested. À / ly (1) |
stablished. 10 / f J
I Éd |
cterize the ;
0 m e —Á mem a 7
es and the ee RE rms =
v, or other o € e $ > 2 & = e = 3 2 9 S A
heck for a - ==
egions. As
emn Figure 2: Histograms of gray level values for the pixels within *under-paving" roads.
the chang
gray level
| processes
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 715