84 
in practically all phases of the process. This 
fact increases the universality of the system 
and makes it flexible for a wide range of ap- 
plications. The immediate interaction capa- 
bility is, of course, a feature completely miss- 
ing in off-line analytical procedures and, al- 
though present in analog systems, its poten- 
tial there is rather limited. Most of the mod- 
ifications and changes in the on-line system’s 
function, as decided on by the operator at the 
time of execution, are built into the pro- 
gramming software supplied by the manufac- 
turer of the system, and can be potentially 
expanded or further developed by users. Ob- 
viously, the key component of the system is 
the computer, and its performance limits the 
function of the system. Fortunately, contem- 
porary minicomputers, such as the PDP 
series 1l, are powerful and fast enough to 
handle practically everything needed in 
close-range photogrammetry. 
The full potential of an analytical on-line 
system can be exploited only ifthe closed-loop 
design is used. The open-loop version has 
limitations both in the scope of its functions 
and in the lower accuracy because of lack of 
feedback. 
TYPICAL FEATURES OF ANALYTICAL CLOSE 
RANGE PHOTOGRAMMETRY 
One of the main characteristics of close- 
range photogrammetry is the impossibility of 
preserving standard conditions in the data 
acquisition phase. For example, the range of 
photo scales is substantially different in such 
applications as scanning electron microscopy 
and terrestrial photogrammetry. Also, the use 
of photogrammetric cameras is, for various 
reasons, not entirely universal. In some in- 
stances, the practical aspects prevail and a 
preference is shown for ordinary non-metric 
cameras which may be more readily availa- 
ble, more versatile, less bulky, or easier to 
operate. Fast moving objects call for photog- 
raphy with movie cameras or with other 
high frequency systems. Finally, some 
photogrammetric evaluations must simply 
rely on given photoimaging systems, such as 
scanning electron microscopes, X-ray 
machines, ophthalmologic instruments, line 
scanners, etc. As a rule, these cannot be mod- 
ified for metric use or replaced by metric 
cameras. 
This variety must be reflected in the way in 
which the image geometry is defined for the 
on-line analytical processing. In most appli- 
cations the image is considered to be a central 
projection, with some systematic deviations 
from the concept due to lens distortion and 
film deformation. For non-metric cameras 
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING, 1976 
these parameters are either unknown or unsta- 
ble and must be derived with the use of 
self-calibrating procedures (Kolbl, 1972) orin 
an on-the-job calibration (Faig, 1975). In 
some instances, the central perspective may 
not be an adequate imaging model because 
the distortion may exceed certain reasonable 
limits, e.g., in the fish-eye or anamorphotic 
lens design. Sometimes, the perspective 
bundle becomes very narrow or is rep- 
resented even better by a parallel beam of 
imaging rays (Kratky, 1975a and b). Evident- 
ly, the character of images should be re- 
flected by various modified projection equa- 
tions. 
Another generalization is required if the 
imaging geometry becomes time dependent. 
In contrast to a simple frozen model of an 
instant exposure, one is confronted with the 
dynamics of sequential imaging typical for 
TV cameras, scanning electron microscopes, 
and other systems employing line-scanning 
principle. All these geometries, atypical in 
photogrammetry, can be handled in on-line 
analytical systems, provided that the calibra- 
tion of the dynamic projection is feasible. 
Although metric cameras should be 
applied wherever possible, one cannot al- 
ways guarantee their arrangement in a regu- 
lar setup which yields a normal or nearly 
normal photogrammetric case. The pictures 
to be photogrammetrically treated may be 
taken individually, in pairs, or in larger 
groups depending on the form ofthe object. It 
is also quite typical for close-range photo- 
grammetry that a set of pictures produces a 
complete view of an enclosed three- 
dimensional object from directions around 
the full circle. With the use of mirrors all 
partial images can be contained in a single 
stereopair. In this instance, and very often 
also in otherwise standard stereopairs, the 
area of interest for the photogrammetric proc- 
essing covers only a minor part of the over- 
lapping photographs. The configuration of 
control and intersection points used for the 
model reconstruction is then not too suitable. 
Consequently, a standard solution turns out 
to be unstable or inaccurate and may ulti- 
mately even fail. To ensure a reliable photo- 
grammetric reconstruction one has to use 
some additional information on the exterior 
orientation of cameras. These auxiliary data 
can then be used in the form of constraints to 
the solution when applied and enforced with 
properly assigned weights. In on-line analyt- 
ical systems the operator can enter the con- 
straints at execution time, immediately check 
their effect, and possibly make modifications 
in a rerun. In a similar way the operator can