Geoscience Reference
In-Depth Information
Table 2.21 Software modules of the simulation system for classi
cation of the phenomena on a
terrestrial surface
Software
The functional characteristic of a software
REICM
Reduction of the experimental information in a computer memory
DRHIM
Data reconstruction by means of the harmonics interpolation method
SIDSM
Spatial interpolation of the data by means of a spline method
DRMOT
A method of optimal interpolation
CASRS
Cluster analysis focused on sings space of remote sensors
CALRS
Cluster analysis focused on the account of local reading of sensors
RSC
Research of spottiness characteristics
RRAMDA
Realization of recognition algorithm by a method of the discriminant analysis
CSAIRM
Sorting and accumulation of the in-situ and remote measurements
CMS
Computer mapping of the spots
In the system designed to classify phenomena on a terrestrial surface automat-
ically the two variants of this approach are program modules CASRS and CALRS,
distinguished by the way in which attribute spaces and comparison algorithms are
organized.
Software item CASRS is focused on the attribute space of remote sensors. The
CALRS module allocates areas of equal instability based on local variations in
sensor data. Distinction between the algorithms of comparison in these items
consists in the taking into account or neglecting of interrelations of the neigh-
bouring counts of sensors. Item CASRS forms clusters without taking into account
a geographical generality of radiometers indications. Item CALRS forms continu-
ous spatial clusters.
CASRS and CALRS automatically exclude from consideration the calibrations
information. Item CASRS consists from six groups of operators. The first group
carries out the organization of data
nal
set of radiometer readings is made by the second group of operators. The third
group of operators
files. Averaging of the information using
finds minimal cluster distances then two points on which this
distance is realized are allocated. The fourth group of operators investigates
changes of minimal cluster distances on the given step of procedure and in case of
occurrence of sharp changes carries out visualization of clusters structure. The
fth
group of operators unites the nearest clusters and recalculates their characteristics.
The sixth group of operators develops a criterion for the algorithm to stop.
Analysis of statistical characteristics of
“
spottiness
”
for three types of areas of
Atlantic and Paci
c oceans was conducted. These statistical characteristics were
determined for the most informative thresholds. At that time statistical character-
istics of
“
spottiness
”
for the same areas, selected using criteria of minimal value of
coef
cient of correlation for joint sample of positive and negative spots. Analysis of
these characteristics showed,
”
coincide for areas with temperate sea roughness and storm zones. Minimum for the
coef
that
the statistical characteristics of
“
spottiness
ˁ
min
is run down for a case of most informative thresholds.
But for quiet area the situation is different.
cient of correlation
Search WWH ::
Custom Search