)04 
   
3 
Sichi 
   
Boundary 
ct Boundary 
Capital 
ct Headquarter 
  
Resouces 
and 
; for 
n be 
1gle, 
nean 
are 
[6] 
il of 
oe to 
tical 
etter 
data, 
like 
ir air 
6.1 Iterative extraction of Vegetation and soil pixels 
International Archives of the Photogrammetry, Re 
    
    
electromagnetic spectrum due to pigments (chlorophyll, 
xanthophyll, etc.), which occurs in leaves. It is especially so 
in the red region of the electromagnetic Spectrum, whereas in 
the NIR region plants are very bright because of weak 
absorption in this spectral range, but the plant scatters 
strongly in this part of the electromagnetic spectrum. 
Vegetation Indices are algebraic combinations of remotely 
sensed spectral bands that can tel] something useful about 
vegetation. The numerical value of a vegetation index brings 
out several important characteristics of a plant like its LA] 
and leaf type and structure, thereby indirectly helping in its 
identification and state of health. 
Generally all vegetation indices assume that all bare soil in 
an image form a line in spectral space and are primarily 
concerned with Red-Near Infrared Space, so a Red-Near 
Infrared line of bare soil is assumed. This line is considered 
to be the line of zero vegetation. The clustering behavior of 
vegetation and soil in Red-NIR space is show in Fig 4. 
x. 
ie Vegetation 
AF . 
  
Figure 4: Soil and Vegetation Reflectance Scattergram 
Vegetation indices like NDVI, RVI etc. measure the angle 
made with the soil line and are called as angular vegetation 
indices. Indices, which measure distance from the soil line 
like SAVI, WDVI, PVI etc. are called as distance based VI's. 
NDVI is the most extensively used vegetation index to 
characterize the amount of biomass, LAI etc.. But the 
difficulty with NDVI is that, if the crop density or LAI is 
high then NDVI value saturates and does not provide 
necessary discrimination among them. The main reason 
being that it does not minimize variable soil background 
effect and being ratio based tends to be non-linear exhibiting 
asymptotic behavior leading to insensitivity to vegetation 
variations over high LAI's and/or canopy covers. The 
limitation of NDVI in discriminating opium poppy from 
other crops is illustrated in Fig 5. 
mote Sensing and Spatial Information Science 
5, Vol XXXV, Part B7. Istanbul 2004 
Multi Date profile for NDVI 
  
200 
i 160 —— Poppy 
8 120 Wheat 
3 —o— Garlic 
© 80 —e— |sabgol 
o 40 —e— Mustard 
0 
35 60 85 110 
Days 
Fig. 5 Multi Date NDVI profile 
IRS-1D LISS-III data acquired on 5‘ December 2001 (35^ 
day), 30" December 2001 (60* day), 24" January 2002 (85" 
day) and 18" February 2002 (110 Day) 
The limitation of NDVI can be circumvented if an index that 
can minimize the variable background effect is used. WDVI 
[4,[7] which is a distance-based index provides the 
necessary minimization of variable soil background effect 
thereby improving the discrimination between opium poppy 
and other crops. 
WDVI is calculated as, 
WDVI=NIR-g*Red (1) 
where, g=slope of soil line 
Appropriate method to calculate soil line slope makes use of 
two atmospherically corrected images of the same area, 
where one image is taken during the growing season and 
other during fallow season. Using the crop and fallow season 
imagery WDVI takes the form 
WD VI (crop) =NIR crop)" & fallow) * Red, (2 ) 
WDVI value is directly proportional to the density of 
vegetative cover and number of leaf layers. 
Inverse WDVI (TWDVI) is calculated as 
l WDV (crop =R€dicrop)=Battow) *NTR (crop (3 ) 
Where gian is calculated by swapping the X and Y-axis. 
The IWDVI value is directly proportional to the amount of 
soil component seen by the sensor 
6. VEGETATION AND SOIL PIXEL EXTRACTION 
It is well known that opium poppy being an agricultural crop; 
only agricultural fields and vegetation that has developed in 
these fields from beginning of the crop season to middle 
through end of crop season are of interest. Agricultural fields 
can be static or dynamic due to slash & burn of forest. The 
static fields will be in the plains for eg. Madhya Pradesh, 
Rajasthan and Uttar Pradesh. The dynamic fields are 
prevalent in areas where jhum cultivation is practiced, like 
a) WDVI, IWDVI image composite creation for crop 
Season image using soil line estimated from fallow 
season image. 
The input images in this stage consist of all features 
types, viz. vegetation of all types, water bodies, built- 
up areas, wasteland, cloud etc. The soil line generated 
1129 
Arunachal Pradesh in India, Thailand, Myanmmar etc. Since, 
the image contains different kinds of land cover features like 
forest, shrubs, settlements, water bodies, agricultural crops 
etc. and as the interest is only in agricultural crops, other 
features can be eliminated from processing. Agricultural 
Crops are extracted iteratively using radiometrically and 
geometrically corrected fallow and crop season image data as 
discussed below: 
from crop and fallow season images: 
in the first iteration is average representation of all 
the land cover types present in the fallow season 
image. 
b) Filtering IWDVI by eliminating zero and positive values. 
All vegetation type features will be less than zero in 
IWDVI image. The other cover types like water