12 
eliminated isolated cells. The resultant grid was then converted into a series of polygons which were labeled 
“trees” (fig. 14). Contrary to the opinion of Wilkinson this was an easy process that produced a reasonable 
thematic map layer in a couple of minutes. Of course, the “tree” polygons are based on a very simple logic 
involving only one band of data. However, the high resolution 2.5 meter data provides a reasonable level of 
confidence in the sampling of “ground control” points. Using visual inspection and the same logic it would be 
just as simple to extract features in the athletic fields such as the track or baseball diamond, roads, and houses. 
2.2.2 Treating Remotely Sensed Data as Continuous Variables 
Another approach to extracting information from the CAMS data involved using some of the grid cell 
tools that are typically associated with digital elevation models. It is important to understand that once the data 
are in a grid structure there are no restrictions on which functions can be performed computationally to the 
data. Therefore, the brightness values for the thermal band of the data could be analyzed as if it were an 
elevation layer. In other words, instead of the “z” value representing elevation it represents a range of 
reflectance values of thermal energy. A simple way of viewing any type of continuous data is to generate 
contours. Within the desktop GIS environment contours can be directly generated from any grid. The contours 
of the thermal band clearly outline features that have significantly different temperature that those of 
surrounding features. For example, the roads, houses, buildings and athletic facilities stand out in the contour 
map (fig. 15). Interesting variations on the use of contours to display continuous data are the use of slope 
values (fig. 16) and a shaded relief map (fig. 17). The slope values highlight the steepest gradient in 
temperature. It provides a “feel” for the variation in the reflected heat more than do the contours alone. The 
hill shading procedure accentuates the high reflectance values and helps to distinguish the man made features. 
While the slope and hill shade procedures generate interesting visual displays the grid values that they produce 
are not as meaningful as the raw data. For example, the low or dark values of the hillshade grid only mean that 
the cell is on the southeast side of some feature that has high thermal reflectance values. 
2.2.3 Integrating Various Data Structures 
In final analysis it is interesting to see how the high resolution multi spectral data can be integrated into 
a GIS environment. Within this type of modem spatial data handling environment it is clear that data may be 
combined in a number of different manners based on the needs of the user. For example, the man-made 
features that are apparent in the grid of thermal data can be “density sliced” into a separate layer. That grid 
can be displayed simultaneously with the “tree” polygons (fig. 18). Since the near infrared values appear to 
reveal several distinct features in the school property they could be represented by their contour values (figs 
19 and 20). In fact, representation of three different data structures can be used simultaneously to highlight 
different thematic features from three different bands of the CAMS data (fig. 21). From a practical viewpoint, 
digital multi-spectral images can be incorporated directly into a GIS environment and handled as either grid 
or polygon data structures. 
It was also an important part of this experiment to demonstrate the difference in image content that we 
can expect in the later half of this decade. It is one thing to say that there are sixteen 2.5 meter data pixels in 
every 10 meter SPOT pixel. It is quite another to visualize what this level of increased spatial resolution means 
in terms of information content. One way to accomplish this is to combine the feature outlines from the CAMS 
data with the SPOT image (fig. 22). From this perspective it is clear that features such as the roads, track and 
buildings that were extracted from the CAMS data are indistinguishable on the SPOT scene for this suburban 
area. It is also apparent that the 10 meter data are not an appropriate source for editing an existing vector road