"m in the late 
analysis with 
je precursors 
led primarily 
development 
tarting in the 
ystems have 
many map 
© Computer 
of data. GIS 
tal mapping 
nal database 
and by the 
y. The vector 
inant in this 
se line maps 
tion of GIS 
h as property 
tor mapping 
le, in digital 
of working. 
ient of GIS 
data model, 
e Computer- 
three origins 
asic range of 
nanipulation 
raster DTM 
f vector data, 
ydelling and 
e data in a 
ystem. Most 
erform basic 
ification and 
me systems 
> to share the 
cess to both 
tionality. In 
itectures can 
atasets in à 
cessing can 
nes and map 
: boundaries. 
2. CURRENT STATUS OF REMOTE 
SENSING/GIS INTEGRATION 
It is useful to consider briefly why this strong 
division of vector and raster processing has 
occurred: 
e The raster and vector data models are 
fundamentally different - one storing data as 
a regular array of values, the other storing 
data as a series of points, with implicit lines 
drawn between them (early vector data 
models were often simply a series of 'pen 
up’, ’pen down’ and 'move to XY' 
commands used for controlling a pen 
plotter). Even today, there is considerable 
uncertainty over how data held in these 
different models should be analysed 
together. 
e Remote sensing and GIS analysis are also 
fundamentally different processes. With 
remote sensing, the source data is spectral 
information (i.e. reflected radiation in one 
or more spectral bands), and the main 
processing tasks are concerned with the 
labelling of each pixel such as 'wheat' or 
'coniferous woodland'. With GIS, however, 
vector data usually represent the edges of 
objects, whose identity is known such as 
road edge’ or "woodland boundary’. 
Remote sensing is therefore primarily 
concerned with the production of data, 
whereas GIS is primarily concerned with the 
analysis of data for applications such as 
record management, route finding or 
locational analysis (although the analysis 
may well incorporate datasets derived by 
processing of remotely sensed data). 
* Historically, remote sensing and GIS 
technologies developed as separate 
disciplines for both cultural and information 
technology (IT) reasons. Information 
technology, in particular display hardware, 
Was strained to its limits to service the 
conflicting requirements of the two 
disciplines and so they grew up conditioned 
by bespoke environments, e.g. frame buffers 
for image processing. Computer hardware 
385 
and software advances, stemming from RISC 
workstations, dramatic improvements in 
affordable memory and storage and Open 
Systems standards, have resulted in an IT 
environment capable of supporting both 
disciplines. The first generation of systems to 
exploit this have been in fact hybrids, with 
combined display of imagery and vector data, 
but discrete environments for the two sets of 
functionality. 
e Due to the cultural separation of the two 
disciplines in the past, there has been a 
dearth of compatible data exchange formats. 
However, a number of recent trends in remote 
sensing and GIS have been influential in 
bringing the two disciplines closer together: 
e Environmental GIS applications are 
increasingly using data derived from 
remotely sensed images. These data may be 
an update or enhancement to a map (such as 
the mapping of forest boundaries) or the 
results of an image classification exercise. 
e Remote sensing analysis is increasingly 
using GIS data. Examples of this include 
the overlay of vector data for presentation 
and location, and the use of DTMs for 
terrain correction of imagery and 
perspective visualisation. 
e The rapidly improving price/performance of 
computers now enables both GIS and image 
processing to use the same workstation. For 
instance, image processing systems do not 
need to use specialised hardware or 
displays, and a standard workstation used 
for GIS has the performance required for 
image processing (both in power for 
processing large volumes of data and the 
graphics display capabilities for 24 bit 
colour images). 
e The convergence of the two disciplines, the 
growing commonality of user requirements, 
and the beginnings of effective 
standardisation activities (de facto and de 
jure) are improving the situation on data 
compatibility and availability. In addition,