hi 
b 
planning permission was given in 1990) and 
also new residential areas (b), coded as bright 
red, can be seen so clearly. The change 
on vegetation has appeared in the change 
image as green (c). Land undergoing urban 
development in 1990 and 92 can be identified 
as magenta color (d). However, the success of 
this method mostly depends on the change 
end points responses. If the change end points 
have dissimilar responses in the selected 
spectral band, than a single band change 
image may show many changes. 
  
Figure 2. 1984-1990-1992 false color 
composite of the study area (Band 11/3/7). 
Image differencing : Image differencing was 
carried out on a pair of co-registered images of 
the study area taken at 1984 and 1992 for 
band 3 to assess the degree of change that 
has taken place in 8 years. The difference 
image is shown in Figure 3 . In Figure 3, areas 
of no change were represented by a value of 
127 while areas that were darker in 1984 than 
they were in 1992 had values between 128 and 
255 (new settlements and industrial complex). 
Areas that were lighter in 1984 had values 
between 0 and 126. 
Principal component analysis : The 
standardized principal component analysis 
(depending on a correlation matrix between the 
bands) was applied to multidate data set. 
From analysis of correlation and eigen 
matrices, it was found that the first four 
components contain more than 97 per cent of 
the total scene variance (Table 2). 
From the evaluation of the eigen vectors, it 
was found that PCA 1 image was identical to 
a total brightness image while PCA 2 were 
interpreted as attributable to changes in 
brightness occurred in overall scene between 
dates. PCA’4 and 5 were also attributable to 
changes in brightness and greenness (Figure 
4). The remaining PCA's were accounted for 
1.24 percent for inter-image differences due to 
atmospheric and sensor variations. 
  
Figure 3. Differenced image result 
(Band 3/ 1984 and 1992). 
Table 2. Eigen Values 
  
Eigen Values 
  
  
  
680 
SOONG DWN 
11 
  
————— 
Figure 4. Color composite of PC 5/2/4. 
Eigenvalues Var. % Total 
       
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996