4 
artificial neural networks that can include fractals, multi space and multi-temporal data sources are needed 
to substantially improve on the current level of classification accuracy. He also believes that the GIS user 
still demands polygon based data layers that are more efficient to store than their raster counterparts. 
2.0 DESKTOP GIS IMAGE PROCESSING EXPERIMENT 
In order to see how far we have come with the integration of GIS and remote sensing in the desktop 
computing environment an experiment was undertaken to process standard SPOT 10 meter panchromatic data 
and a nine band 2.5 meter scene of Calibrated Airborne Multi spectral System (CAMS). The CAMS data 
was provided by NASA Stennis as part of its commercial applications program (Jensen, et al., 1994). The 
data provide a useful glimpse of the next generation of high resolution multi spectral remotely sensed data 
that are likely to be available in the next few years. The experiment was based on using the latest version 
of a desktop GIS (ArcView 3.0) that incorporated the ability to handle both vector and grid cell data 
structures (ESRI, 1996). The system runs on a standard Intel based personal computer under Microsoft 
Windows 95. In other words, the experiment was performed on the same type of computer used for home 
multi-media applications. The user interface was a standard Windows 95 graphical user interface and was 
fully integrated with a word processing system through simple cut and paste operation. In fact, all of the 
illustrations in this paper are simple screen captures pasted into a wordprocessor and output on an inexpensive 
inkjet printer. 
2.1 PROCESSING SPOT DATA 
The first part of the experiment involved the integration of preprocessed SPOT 10 meter 
panchromatic data (SPOT, 1988). In order to meet the objectives of the study a small section of a SPOT 
scene for Columbia, S.C. which had been geographically registered and stored as an ERDAS LAN file was 
selected. This standard ERDAS image format was directly incorporated into the GIS without any conversion. 
Using the histogram in the legend editor it is possible to graphically enhance and manipulate the single 
panchromatic band to accentuate the desired features (fig. 1). In this manner the image can serve as a visual 
back drop for editing and other GIS processing. From a spatial analysis perspective, it is important to be able 
to directly convert an image into a grid. This function is implemented through a simple pull down menu 
(convert to grid). Once the SPOT image was converted into a grid it could be treated as a matrix of numbers. 
In this case the numbers happen to correspond to the 256 brightness values of the SPOT panchromatic band 
(green-red). This grid can be manipulated and displayed in a number of ways. In effect, the brightness values 
can be classed and symbolized with a wide range of classing methods (equal interval, equal area, natural 
breaks, standard deviations etc.) and number of class intervals. For example, by dividing the range of data 
into five equal area classes it is possible to enhance and distinguish specific types of geographical features 
based on their reflectance values (fig.2). The equal area classing procedure has the effect of stretching the 
histogram by insuring that each of the gray shades is represented. In addition to providing good visualization 
capabilities the grid structure enables the data to be classified and mathematically manipulated into any 
number of subsets. For example, by selecting the values greater than 56 it was possible to create a separate 
layer of features that reflect the most light in the SPOT panchromatic band (fig. 3). This process places 
roads, buildings and bare ground into a separate grid file.