Data Points
200 Accuracy
35. 6% 42. 54 61. 5% 42. 0% 56. 34 59. 84 79. 9% 81. 0% 71. 8% 71. 3% 85. 1% 74. 7% 82. 2% 82. 6% 85. 1%
I-
pee] que | Ls mus um GM um ; : Excellent
150 F
Good
100 F
80 FS
Possible
1 2 3 4 5 6 7 8 9 | 11 12
The Number of Units on Hidden Layer
BB Correct in cach rank. [J [] Incorrect in each rank
Figure 2. The accuracy of networks with different number of units of hidden layer.
Input Layer Hidden Layer Output Layer
Wii fi
Area under
8 Degree Hg
Area under a Wi
15 Degree
H4
Area over Hs
15 Degree
Suitability for
4 . «1 Lits
" O Grassland Establishment
Elevation 5
T7
Ha
Topography
H,
Soil Hi
Productivity
Hu
Figure 3. The neural network to assess the suitablity for grassland establishment.
Table 2. Network parameters determined by NEUROS2.
Offset of H1 H2 H3 Faro H5 H6 H7 H8 H9 H10 H11
Hiddel Layer 1.834 -0.621 -1.279 -0.296 -1.270 3.704 -2.923 -0.225 0.466 0.416 2.941
Link. Coef. H 12 I3 14 15 16
H1 -1.061 -0.291 -0.427 -4.805 -2.459 2.089
H2 -0.858 -0.971 -0.126 -0.963 -0.428 -1.388
H3 -2.859 -13.304 -7.903 0.332 2.558 -2.512
H4 -3.168 0.673 -2.663 1.720 -0.522 -4.137
H5 0.608 3.060 3.803 -2.239 -0.561 -3.385
H6 2.126 -1.874 2.190 -6.164 -6.325 -7.022
H7 3.305 0.003 -3.463 0.296 0.666 -2.314
H8 1.108 -0.094 2.972 0.315 -2.641 -4.361
H9 0.084 -4.605 -7.942 -3.455 -2.058 2.467
H10 -2.978 -1.585 -6.215 -9.085 0.755 0.713
H11 0.887 -5.200 -1.446 -2.194 -4.085 -2.961
Offset of Output layer -0.264
Link. Coet. H1 H2 H3 H4 po —— H6 H7 H8 H9 H10 aim
0.448 -0.649 1.620 0.714 0.535 3.636 0.691 -3.248 0.909 -1.613 1.438
790
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
In t
graz
Gra
regit
exp:
hum
envi
agric
man
cons
This
using
(1)
Figur
map
parts
seco
the c
was r
First,
map (
direct
cover
Outpt
and e
Thus,
same
Using
NEUR
with di
