2.1.1 Change detection with SPOT data 
6 
It is also possible to handle temporal image data within the GIS environment. For this study two 
SPOT scenes for the same area were processed for 1990 and 1992. Since they were both registered to the 
same projection and same coordinate system they could be treated as two grid layers. In a more rigorous 
study one would be very cautious about comparing imagery at different dates; it is intriguing to simply 
subtract them to determine where the greatest differences exist (fig. 4) (Jensen, 1995). A standard deviation 
classing procedure identified the cells in this difference grid that had the greatest changes in reflectance values 
between the two dates. The spatial distribution of cells that had greater than three standard deviations creates 
an interesting map of changes in this rapidly growing suburban area. Since these extreme changes form a 
series of clumps it was useful to convert them into polygons (fig. 5). The conversion from a grid to a polygon 
data structure was considered to be quite a sophisticated procedure at the end of the last decade. In the 
current desktop GIS environment this function is a simple option on a pull down menu. The resultant 
“change” polygons can be displayed with the 1990 SPOT data to pinpoint areas of greatest change (fig. 6). 
2.1.2 Summary of SPOT analysis 
The major conclusion of this part of the experiment was that the SPOT panchromatic data were easily 
incorporated into the desktop GIS. It was subsequently converted into grids which were then processed into 
other GIS layers. The system also supported temporal analysis and conversion of grids into polygons. In 
effect, this indicates that even small organizations with minor resources can perform rudimentary remote 
sensing tasks that can supplement the traditional windshield surveys performed by many planning 
organizations. Several remote sensing organizations are currently providing data that are preprocessed into 
an image format directly compatible with GIS. This implies that remote sensing is increasingly serving as 
a common input to GIS. All of the image rectification and registration procedures are essentially being 
handled by service or data providers, thereby eliminating much of the need for close coupling of the two 
technologies. This service provider type of remote sensing seems to intensify the debate about the 
linkage between remote sensing and GIS. For example, ten years ago Fussell et al. raised the following 
questions: 
What will be the role of remote sensing vis-a-vis the current trend toward Geographic Information 
Systems (GIS) technology? Is our future role to be reduced to providing input to GIS activities? 
(Fussell et. al. 1986) 
2.2 INFORMATION EXTRACTION FROM CAMS IMAGERY 
According to Wilkinson (1996) one of the goals of image processing is generalization. He suggests 
that the generalization process is critical to GIS data base development, 
“... that is to simplify the spatial structure based on a guiding principle that the thematic map should 
be made as visually simple as possible but on the basis that only information from the image domain 
enters the process. The generalization of the resulting pixel-based thematic map is a difficult problem. 
... Generalization is needed essentially to transform the map into a form which is suitable for 
vectorization and storage in a GIS. Generalization in the image domain can therefore be viewed as 
the process of preparing the thematic map data for entry into the GIS. This is needed essentially to 
do two things: (i) remove spatial noise arising from erroneous classification and (ii) to reduce the