\TER , SOIL and
dium and High
W , Medium and
\TER , SOIL and
ve distinct char-
knowledge, only
isfy the conver-
whereas, mixed
| using the over-
ddressed by the
ssibilistic calcu-
s NDVI and HUE
ively represents
s both the labels
9e i.e. soil, so by
le, it can be
; SOIL.
egory, the mem-
are found to be
1.1). By applying
erator, the mem-
ver is found to be
| and HUE values
; found that NDVI
nedium i.e., the
ind soil, whereas
1 to High i.e., the
egetation. As the
ver types, so in
e fuzzy variable
value is 22. The
im and High, thus
re of soil & water.
ence” principle, it
le is a mixture of
vegetation is ex-
ase.
>mbership of the
es for water are
ese are 0.88 and
ariables. Using
erator, the mem-
» found to be 0.12
possibilistic OR
as SOIL.
a 1996
ES
EXAMPLE 3.
A sample, X3 (Refer fig.1) has NDVI and HUE
values of -0.17 and 180 respectively. It is found that
NDVI gives a fuzzy label of Low & Medium i.e., the
sample is water & soil, whereas HUE gives a fuzzy
label of Medium to High i.e., the sample contains
both water and vegetation. In this case as the labels
represent different land cover types, so in order to
resolve the ambiguity, the fuzzy variable TONE is
examined and has a value of 15. The fuzzy label Low
& Medium assigns the pixel as a mixture of soil,
water & vegetation. By applying “ Convergence of
Evidence" principle, itcan be inferred thatthe sample
is a mixture of all the three land cover types.
For determining the category membership of the
pixel, the different membership values for water are
0.55, 0.95 and 0.67, for soil these are 0.45 and 1.0
and for vegetation 0.08 and 0.17. Using Possibilistic
Function of AND operator, the membership value of
water, soil and vegetation are found to be 0.55, 0.45
and 0.05 respectively. The hard ciass, using
possibilistic OR operator, of the pixel is classified as
WATER.
The summarised results of the examples as ex-
plained above is shown in Table 3.
6. CONCLUSION
The proposed method is less numerical intensive in
comparison to standard and widely used classifica-
tion methods like Bayesian Maximum Likelihood
method and any other sub-pixel classification meth-
ods. This method can be utilised to develope a fully
automated image understanding system. In doing
so, it may be generalised to find the different types
and numbers of constituent classes. In fact, it will be
required to prepare a meta-level knowledge module
to resolve the types of VARIABLES and their opti-
mum combinations to be considered for different
types of mixed pixel. However, to provide the exact
category membership of the constituent classes,
the membership functions of fuzzy labels may be
defined on the basis of in-situ conditions as these
may vary both spatially and temporally. The source
of imprecision in this method lies in the manner in
which fuzzy labels and fuzzy algorithms are applied
to the formulation and solution of the problem.
7.REFERENCES
Dubois, D and H.Prade, 1988. Possibility Theory.
Plenum Press, NY.
Estes, J.E., E.J. Hajic and L.R. Tinney. Fundamen-
tal of Image Analysis: Analysis of visible and ther-
mal inferred data. Manual of remote sensing, 2nd
ed. American Society of Photogrammetry. Vol.1.
Chp.24, pp 987-1123.
Fisher, P.F. and S.Pathirana, 1990. The Evaluation
of Fuzzy Membership of Land Cover classes in the
Suburban zone. Remote Sensing of Environment,
34, 121 - 132.
Foody, G.M., and D.P.Cox, 1994. Sub-pixel landcover
estimation using a linear mixture model and fuzzy
membership functions. International Journal of Re-
mote Sensing, Vol 15, No.3, pp 613 - 631.
Jain, A.K.1989. Fundamental of digital image pro-
cessing. Prentice Hall of India Pvt. Ltd.
Jasinski, M.F. and P.S.Eagleson, 1990. Estimation
of Sub-pixel vegetation cover using red-infrared
scattergrams. IEEE Transactions on Geoscience
and Remote Sensing, 28, pp 253-267.
Metzler, M.D. and R.C.Cicone, 1983. Assessment of
Technologies for classification of mixed pixels. 17th
International Symposium on Remote Sensing of
Environment, Ann Arbor, Michigan, pp 1051 -1021.
Settle, J.J., and N.A.Drake., 1993. Linear Mixing
and estimation of ground cover proportion. Interna-
tional Journalof Remote Sensing, Vol 14, pp 1159 -
1177.
Townshend, J.R.G., 1980. The spatial resolving
power of earth resources satellites. NASA Technical
memo 82020.
Wang, F.,1990. Fuzzy Supervised Classification of
Remote Sensing Images. IEEE Transactions on
Geo-Science and Remote Sensing, Vol 28,
No.2,1990, pp 194-201.
Zaded, L.A.,1965. Fuzzy Sets. Information and con-
trol, Vol.8, pp 383 - 353.
Zadeh, L.A.,1973, Outline a new approach to the
analysis of complex systems and decision process.
|. E.E.E.Transactions on Systems , Man and Cyber-
netics, 3, 28-44.
Zadeh, L.A.,1975, Calculus of fuzzy restrictions. In
Fuzzy Sets and their applications to Cognitive and
Decision processes, edited by L.A.Zadeh, K.S.Fu,
K.Tanaka and M.Shimura (AcademicPress, NY), pp
1-40.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996