the impact of GPS techniques, and the ready
availability of ground control points in an
effectively standard reference system (WGS
84), greatly facilitates the registration, in
practice, of remote sensing and GIS or map
data.
A review of recent literature in the fields of
GIS and remote sensing reveals that there is
considerable interest in the integration of
remote sensing and GIS. Sadler et al (1991),
for instance, use census data in an attempt to
improve the classification of urban areas.
Census areas held in vector from are processed
to generate a continuous surface raster dataset
of population density value. This is then used
as a pseudo 'image plane' in a classification
process.
Janssen et al (1990) describe the use of a
topographic map to classify an image on a
'per-object' basis. However, as the systems
used (leading GIS and image processing
packages) have no capabilities for integrated
processing of this type, they rasterise map
objects and then use these as a mask within a
self-developed program for assessing the
dominant class within the objects, from an
image classified using a standard per-pixel
classification process.
The conclusion inevitably reached, therefore, is
that a level of integration substantially beyond
co-existence of image processing and GIS is
required. The statement:
"The integration of image data into GIS is
one of the great ideas whose time has
come" (Faust et al, 1991)
is now extremely relevant, whilst the
statement:
"Remotely sensed images have been shown
to be a cost effective means for update of
GIS data" (Faust et al, 1991)
is clearly demonstrable. The chief operation
that can be said to date to effectively integrate
remote sensing and GIS is heads-up digitising.
However, this has been due to the limited
386
capabilities of software packages: further
benefits of integration remain to be realised.
3. THE FUTURE FOR INTEGRATION OF
REMOTE SENSING WITH GIS
The first important distinction to make is
between remotely sensed data and raster data
(which may, of course be, or be derived from,
remotely sensed imagery). Most references to
'integration' refer to the integration of raster
data with vector processing functionality.
Although integrated data processing can utilise
data in both raster and vector forms, it is still
necessary for the output data to be held in one
form or another. For instance, a function to
classify an image on a per-object basis will
output data as an attribute of a vector object,
and a function assessing land suitability for
development might output data as a single
raster ’suitability’ value.
Processing functionality that genuinely
integrates GIS data with remotely sensed data
can be divided into two groups:
e Functions that enhance image processing
operations.
* Functions that enhance vector operations.
These generally make no distinction
between the use of remotely sensed imagery
and of any other raster dataset.
4. FUNCTIONS THAT ENHANCE IMAGE
PROCESSING OPERATIONS.
There are three main areas where GIS data can
be used to enhance standard image processing
functions:
4.1 Geometric correction.
Conventionally, ground control point (GCP)
co-ordinates are input to an image processing
system as an image is being rectified. If the
system can display and manipulate GIS data,
however:
e GCE
data
e Vec
veri
e AL
imas
42. 1n
This is
image |
in a nu
e The
Inste
sets
data
data
used
* Obj
clas:
hom
mor
obje
vect
give
with
199
relie
mix
easi
e Ten
illur
can
illur
(eg.
1ma,
may
"che
soul
,
pse
e Knc
dev
(eg.
to u
by:
cou
incl
exis