International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
bodies, built-up areas, wasteland etc. have values 
greater than zero, which are eliminated by this 
filtering operation. 
c) Extracting pixels having negative value only from 
original fallow season and crop season image. 
After this operation fallow season image consists 
of all types of vegetation and soil pixels. 
d) Generating WDVI, IWDVI image composite using 
filtered fallow season image for soil line estimation. 
      
    
  
    
  
    
   
       
  
  
  
  
  
  
  
  
  
  
  
        
  
  
  
  
  
  
  
  
  
  
  
    
  
  
  
  
  
  
The soil line generated in this stage is average 
representation of all types of vegetation and soil 
present in the fallow season image. 
e) Repeating Steps — (a) to (d) until slope of the soil line Ju wm 
; "ET a : i 2 E * 
becomes stationary. ap Fig.7(a) DOP: 10.11.2001 Fig. 7(b) DOP : 24.01.2002 
The repeat of the procedure eliminates shrub and (Fallow Season) (Crop Season) 
waterweeds. Soil line stabilizes only if vegetation and vid = ver Ed TE epu ? 
n hy Bo c9 sn Co TA MY | 4e 454 à 
soil pixels are present. The soil line at the end of ui >" 4e NER X a nuts od, d 
iterations is fit between vegetation and soil land cover ui * | Là e a EAM A : t re STO d 
types. Eo Le 3 Mir - Moe *. e. t is: ph. 
f) Ratio of NDVI crop over NDVI fallow is used to ER ca ie M a, wo. Sp Ia ( 
eliminate residual forest. ; : "k ^ W d Ë # sal + w 
: S s - 3 ; eh Ty A 
g) Extracting soil pixels from fallow season and dynamic Eo t Let d ne bu 4^ (ios 
vegetation pixels from crop season images yt b ai et 4v a - 3, , 
e The filtered image of fallow season contains only [wt bee ds. ^ ke ^ i 
soil pixels (bare agricultural fields). M a er, Vu NT 7. 
= : : peus t "ons iN 
e The filtered image of crop season contains only "2 58e TE Er RA 
dynamic vegetation (agricultural crops). ne Me b i Is. N Ha 
S ; 4 “te . x ye Ll. x A T imma , à eet ud 
The above sequences of operations ensures identification of Fig. /(c) Extracted soil Fig. /(d) Extracted vegetation 
agricultural plot soil pixels use of which can result in better From the multi-date WDVI profile shown in figure (8) it can 
estimation of soil line. The flow chart for automatic be observed that the image data acquired during 85th to 
extraction of dynamic vegetation is show in Fig 6. 105th day will discriminate opium poppy from other crops. 
Cosine/Atmosphericall : 
ol una Multi Date profile for WDVI 
Two date Imagery N 
200 
^ 150 e— Poppy 
2 
Regression A C Wheat 
(send = 100 —e— Garlic 
© 
© —e— Isabgol 
o 
v 50 —e— Mustard 
IWDVI image 
Generation 0 
4 35 60 85 110 
Days 
Eliminating IWDVI 
Zero & +ve values 
V \ Figure 8 : WDVI multi-date profile 
ds IRS-1D LISS-III data acquired on 5" December 2001 (35" day), 30^ 
Filtered | | December 2001 (60™ day), 24™ January 2002 (85" day) and 18" 
Vegetation February 2002 (110" Day) 
NDVI S uat ino: EM 
However the level of discrimination is not similar across all 
Extracted Py Extracted growth stages, but maximum discrimination is possible when 
Vegetation Soil opium poppy is at its peak growth stage, i.e. flowering phase 
  
  
  
  
  
  
; : : . : (85th to 110th day). 
Figure 6. Flow Chart for Automatic Extraction of Dynamic 
Vegetation 
Figure 7(a) —(d) shows section of IRS-1D LISS-III imagery 
covering part of Western Madhya Pradesh. 
7. CLASSIFICATION 
high WDVI and low IWDVI value. The color composite is 
used for training set identification and for performing 
maximum likelihood classification. 
The two images WDVI and IWDVI are displayed as color 
composite by assigning IWDVI image to red plane and 
WDVI image to green and blue planes. Pixels of opium 
poppy appear in unique hue (cyan), as opium poppy.crop has 
1130 
NY)