It can also be concluded that full automation is 
still a distant goal, computer assistance can 
play an important role in many applications 
but the results achieved by the analytical 
power of the human operator cannot yet be 
reproduced with sufficient success to remove 
the operator from the process of feature 
extraction and object recognition. 
4. Integration issues 
It is now widely recognised that more 
information can be obtained from imagery if 
data from several sources can be used. In 
order to use such techniques it is essential that 
these varied images be registered into a single 
co-ordinate system. Several examples have 
been quoted above. The image data also has to 
be used with map data or other data in vector 
form. To make use of this potentially vast 
source of data automation has to be used to 
synthesise the data and extract useful 
information. For example in the process of 
extracting features the support of a GIS is 
becoming essential, a number of methods 
under development for change detection 
require the use of existing information to help 
identify change. In a system for detecting 
changes in forest under development at UCL 
under contract to Earth Observation Sciences, 
edges extracted from imagery are compared to 
boundaries from maps in order to determine 
where change has, or has not taken place. In 
the top down method of feature extraction 
information about the objects is required and 
can come from a GIS. Ehlers (1994) 
identifies a number of issucs for investigation: 
data models 
integrated classification procedures; 
integrated error models; 
automated registration and map revision; 
integration with modelling techniques. 
These are currently the active arcas of 
research. 
5. Validation issues 
The concept of accuracy is well established in 
the evaluation of photogrammetric data. There 
is no problem in establishing the accuracy of 
co-ordinates derived from imagery as long as 
control in the form of check points is available. 
Moreover the accuracy is usually related to 
the scale at which a map is produced. Once 
data from different sources is merged the 
origin may be lost and if data or images of 
different scales are merged then assumptions 
may be made which are not valid. For this 
reason data should be given an attribute which 
indicates its source and accuracy. Little effort 
has been made to do this at present although 
342 
work has been done to assess problems caused 
by generalisation at different scales (Jaoa, 
1994). 
In the area of Radar data which has been 
geocoded products are produced to indicate 
certain aspects of accuracy. A pixel area map, 
also called an energy map, can be produced to 
indicate the surface area of a pixel when 
recorded, thus allowing correction for slope 
effects to be made by the scientist who is 
interpreting the data. Layover and shadow 
maps can also be produced (Schrieir, 1993), 
Error budgets are also produced with geocoded 
data which include an assessment of the 
accuracy of the DEM used for the geocoding. 
At the Remote Sensing Institute of the 
European Union's Joint Research Centre at 
Ispra, a geophysical processor is being 
developed which aims to provide an 
application scientist with all the information, 
and tools to use that information, to allow 
accurate interpretation of the data. 
6. Conclusions 
Digital photogrammetric systems are now 
established as production tools although the 
economic advantages are not yet well 
established. Photogrammetric software is also 
available on digital image processing systems 
although care must be atken in using this and 
accuracy assessments made. The use of the 
latter systems with GIS enables data derived 
from remotely sensed data to be directly passed 
to the GIS and this offers significant avantages 
in data acquisition. 
The big break through will be when 
automation can be applied as a production tool. 
At the moment this only applies to elevation 
models and automatic feature extraction and 
model orientation is still a very active research 
area. 
As these new algorithms are being developed 
and incorporated into digital systems it is 
important that quality is considered and 
validation techniques built in. 
References 
Ahac A. A., Defoe R., Wijk M. C. van (1992); 
"Considerations in the Design of a System 
for the Rapid Acquisition of Geographic 
Information', PE&RS, Vol. 58, No. 1. 
Benny A. H. (1981); 'Automatic Relocation of 
Ground Control Points in Landsat 
Imagery’, Paper presented at a RSS society 
conference on "Matching remote sensing 
technologies and their applications", 
London, 16-18 December, 1981.