OOP LPL LLL CLC Lt <<
CCL CLEP LD LP CL << <P TOPO >>>
OP <<<-<
»YYY
Y YC YAC Y YCADU
678901234567
ng
Combination of green and blue bands. Three catego-
ries, namely, paddy freshly planted {A1.1), paddy
harvested (Al.h) and fallow land (A10), are classi-
fied with sufficient accuracy. Paddy mature (Al.3)
is misclassified as sugarcane young ratoon (A2.2).
Comparison of the histograms (Fig.5) for these two
categories shows inadequate separability.
ii) Combination of blue and red bands. Paddy freshly
planted (A1.1), Paddy harvested (Al.4) and fallow
land (AlO) are classified with good accuracy, as
for the combination of green and blue bands.Histo-
grams for the various categories (Fig.5) suggest
that categories not separable in the green band are
also not separable in the red band.
iii) Combination of red and green bands. Paddy harvested
(Al.h) and fallow land (A10) can be classified with
good accuracy. However, about half of paddy freshly
planted (Al.1) is misclassified as sugarcane young
ratoon (A2.2). An important result is that this
band pair is possibly the best for the classification
of paddy well grown (Al.2).
iv) Combination of blue, green and red bands. Evidently,
there is no improvement in classification accuracy
with respect to paddy mature (Al.3), sugarcane fresh-
ly planted (A2.1) and sugarcane well grown (42.3),
over that obtained from a combination of any two
bands. As in the case of the two-category classifi-
cation result (Table 1) for paddy well grown (A1.2)
and sugarcane young ratoon (A2.2), for instance, here
also more than half of Al.2 is misclassified as A2.2;
most of A1.3 is misclassified as A2.2; more than half
of A2.1 is misclassified as Al.1l; and about half of
A2.3 is misclassified as Al.l.
4. Conclusions Digital analysis of aerial color infrared fra-
mes of agricultural scenes (containing paddy and sugarcane) appee
ars to be a feasible method for classfication (and hence for acre
acreage and yield estimates) of the crops in the area under study.
However, choice of optimum bands (even in the visible + near infra
red portions of the electromagnetic spectrum), which is definite-
ly dependent on the season (or time) of acauiring the data, for
all the various categories (in different stages of growth Y is
still to be established. As regards the widespread aoplication of
such an analysis, more basic work is to be done.
Further Acknowledgements This work was carried out under a resea-
rch project sponsored and funded by the Indian Space Research Or-
ganization, Department of Space, Goverment of India.
Thanks are due to Professor G.Krishna and Group (Indian Ins-
titute of Science) for their valuable assistance in the digitiza-
tion of the picture data.