Geoscience Reference
In-Depth Information
Table 9.9 An example of the database element of the decision making system
Solution
Q
Vector-identi
ers
CuSO
4
10 %
Q
Δ
21.6
0.17
0.1
143
0.67
0.12
2
0.16
0.21
Q
ʨ
43.1
0.89
0.04
201
0.59
0.09
3
0.65
0.01
—
means of search in the BDMS of vector
identi
ers which are minimal distance
er Q
¼
X
1
; ...;
from considered vector-identi
f
X
n¼
g
of given water solution. Dis-
tance between vector-identi
ers is calculated with the use of the following formula:
2
4
t
X
3
5
þ
2
X
n
n
1
2n
min
X
j
A
j
d
¼
min
n
q
ð
Q
Q
n
Þ
¼
X
j
A
j
ð
9
:
4
Þ
i
j¼1
j¼1
Use of Eq. (
9.4
) gives better result in comparison with the application of other
known criteria of closeness between spectral curves. That is why in this case there is
minimal risk to miss the situation with dangerous pollution of water reservoir. In
common case, usually the following three methods are used:
Cluster analysis. In this case two types of clusters are formed for Cos
ʔ
and
Tan
ʨ
. Decision is made by weighted value of
ʴ
from Eq. (
9.4
).
Algorithm of discrepancy between spectra. It is assessed average distance
between the ordinates for both spectra and studied case and decision is made
taking into account minimal value of this distance.
Algorithm of discrepancy between etalon vectors. In this case decision made
taking into considered minimal value of
ʴ
from Eq. (
9.4
).
Table
9.10
gives experimental results, which give possibility to compare these
algorithms. Dependence of risk assessment as function of solution concentration is
represented in Figs.
9.12
and
9.13
. As it follows from these results, the risk to have
Table 9.10 Comparatively assessment of algorithms for recognition of spectral images of water
solutions
Object for study
Identification algorithm and its error (%)
Cluster
analysis
Discrepancy
between spectra
Discrepancy between
vector-etalons with the
use (9.4)
CuSO
4
15
12
8
NaCl
17
11
7
NaHCO
3
16
10
5
NH
4
OH
21
13
9
ZnSO
4
22
12
8
Potassium iodite
13
10
6
Na + Cu + Zn + Mn + glucose
18
9
9
Furaciline
23
11
5
Bi
dumbacterium
14
10
4
Search WWH ::
Custom Search