316 
the operator can supervise the entire process at the same time to 
tune the parameter values and to intervene when necessary. These 
advantages, however, do not outweigh the limitations of the 
processing strategy and algorithms. 
Digital systems are potentially most versatile, flexible and 
efficient for medium-fine and coarse terrain relief modelling. 
Time-relaxed operation allows implementation of sophisticated and 
complex procedures. On the other hand, however, a huge storage 
capacity is needed for images; moreover, high-speed storage, 
retrieval and communication, and fast processing and display are 
required. 
Performance of image matching can be increased by exploiting the a 
priori geometric knowledge, such as the collinearity or coplanarity 
(epipolar) relation. Self-adaptive capabilities are indispensable, 
and in certain circumstances the least squares fit improves the 
matching accuracy. 
Image matching can be linked with or based on feature extraction. 
Features extracted from images can support matching of the adjacent 
areas and vice versa; knowledge of the local heights gained from 
image matching improves automatic image interpretation. 
At present, automatic terrain relief modelling is one of the major 
areas of research and development in photogrammetry. The trend is 
from medium towards fine DTM. The interactions between matching, 
analysis and understanding of images are gaining in importance. 
"Least squares matching" plays a minor role in relief modelling. 
5.3 Image transformations 
Image transformations concern various conversions of the input into 
the output images. Most common transformations in GI applications 
are from perspective to orthogonal geometry by means of the 
orthophoto systems. Such systems can operate in real-time, i.e., 
parallel with the relief modelling, or in time-delayed stages. For 
massive production and updating, the time-delayed or "off-time" 
systems are better suited. 
Analytical systems, such as the Wild 0R1 Aviophot and Zeiss 
Orthocomp use digital geometric transformation and tracking control, 
and optical transfer and transformation of a thin image window. The 
technique is known as digitally controlled optical-mechanical 
"differential rectification". In rough terrain, the window length 
should be short enough to keep the remaining distortion small, 
which, on the other hand, impairs the production speed. Another 
limitation of the analytical systems with optical image transfer is 
the inability to handle the image intensity/colour. An exception is 
the control of illumination intensity, which does not suffice to 
obtain homogeneous output images as desired for mosaicking. 
The major advantage of the analytical systems with optical image 
transfer is high image quality of the transformed images, which has