1.1 DEFINITIONS 
The first step in any form of automated geographical processing consists of the transformation of 
analog models of objects on the Earth's surface into machine readable formats. Remote sensing provides a 
valuable way to collect and transform such data. It is interesting to look back more than thirty years to 
understand the objective of digital image processing:. 
Before the computer can process and analyze them, the images must somehow be translated into [an] 
analytic form acceptable to it. This is a rather difficult problem without satisfactory solution as yet 
and in computer programming comes under such various terms as "artificial intelligence" and 
"pattern recognition." (Kao, 1963) 
Functionally nothing has changed since that time. We only seem to argue about the definition of when we 
are performing a remote sensing, GIS or cartographic operation (Cowen, 1988a). The GIS community tends 
to view remote sensing as a data input technology, while the remote sensing specialists see GIS data as useful 
ancillary information to improve their classification accuracies. 
1.2 DEVELOPMENTS IN THE 1990’S 
The article by Ehlers et. al. (Ehlers et. al. 1989) serves as a good starting place to evaluate the state 
of integration of remote sensing and GIS technology. Basically, they characterized the state of the art at that 
time as one of ‘separate but equal’ systems that could exchange data formats between the two systems. In 
fact, research published about that time was just beginning to include discussions of the conversion of 
Thematic Mapper data into polygons for direct analysis with other GIS layers (Cowen et al., 1988b). The 
vectorization procedure was rather cumbersome and required data conversion in a remote sensing system and 
subsequent file transfer to the polygon based GIS. In other words, at the end of the last decade sophisticated 
user could incorporate remotely sensed data into GIS applications regardless of the data structure. 
Ehlers and the others viewed the next step in the integration process to be a “seamless” environment 
in which image based data could be handled as entities that could be combined with a wide range of other 
spatial and temporal data sources in a single system. In such a seamless computing environment a user would 
not be required to convert data formats when moving among procedures and would not even be aware that 
tools from different applications were being used. Their view of a totally integrated system extends the GIS 
and remote sensing integration to include real time photogrammetric measures and robust spatial query 
languages that could operate on any spatial data structure. They did not believe that any commercially 
available system existed at that time to handle this seamless level of integration 
In the middle of this decade it is useful to take stock of where we are in this evolution towards the 
closer coupling of remote sensing and GIS toolboxes. It is difficult make a straight forward comparison. In 
many ways systems do not measure up to the “seamless integration stage” outlined by Ehlers. However, it 
is clear that, functionally, some vector based GIS software systems have incorporated a full suite of grid cell 
modules and tools for integration of image based data. At the same time, the remote sensing systems, have 
expanded their ability to display and convert vector GIS layers. There is no need to convert either data 
structure to display it. There are also simple procedures to rasterize or vectorize spatial data in both types 
of systems. Furthermore, fairly complete implementations of Tomlin’s (1990) grid based cartographic 
analysis are now common to both GIS and remote sensing systems. One of the major distinctions between