74 
e Flexibility in focusing range, 
e Some are motor driven, allowing for a quick 
succession of photographs, 
® Can be hand-held and thereby oriented in any 
direction, 
€ The price is considerably less than for metric 
cameras. 
Disadvantages: 
€ The lenses are designed for high resolution at 
the expense of high distortion, 
€ Instability of interior orientation, 
€ Lack of fiducial marks, 
€ The absence of level bubbles and orientation 
provisions precludes the determination of ex- 
terior orientation before exposure. 
The increasing use of non-metric cameras 
for photogrammetric purposes is, to a great 
extent, due to the research and development, 
both within and outside the ISP, aimed at the 
elimination of these disadvantages which in 
most cases is accomplished by combined 
calibration and evaluation techniques. As 
pointed out by Faig?, “the amount of object 
space control is directly related to the calibra- 
tion and/or the evaluation approach, and to 
the accuracy requirements.” He also divides 
the photogrammetric equipment systems, for 
non-metric cameras, into five groups accord- 
ing to the type of calibration that is included 
in the system. 
Some of the disadvantages can be over- 
come by introducing fiducial marks, often 
very easily. Orientation aids such as spirit 
levels can also be arranged in the workshop 
without too much effort. The instability of the 
interior orientation can be difficult to reduce 
because this necessitates more severe alter- 
ations in the camera itself. Examples of this 
are the introduction of glass plates in the 
image plane and film flattening by hand- 
driven suction devices in the film magazines. 
When these suggested or similar alterations 
are made to the camera, it will be more or less 
completely rebuilt, and after a complete 
calibration, may be regarded as a metric cam- 
era?! However, because the stability of the 
camera and the reproducibility ofthe interior 
orientation is limited, Faig calls the calibra- 
tion of such a camera "partial", even ifall the 
parameters used for the calibration of metric 
cameras are included. 
The other extreme to the calibration of 
non-metric cameras is the use of methods 
which do not require any calibration at all. 
One example is numerical and graphical re- 
ctification based on projectivity between the 
plane ofthe image and the plane ofthe object. 
Another example is the direct linear trans- 
formation (DLT) approach!. In both cases the 
solution is principally for the interior orienta- 
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING, 1976 
tion of the actual images but the elements 
themselves are not explicitly obtained from 
the solution of the equation systems. 
Between these two extremes of complete 
and no calibration, Faig introduces three 
other categories of calibration, namely: par- 
tial, self- and “on-the-job” calibration. 
Partial calibration is usually performed in a 
laboratory with some kind of test area or 
target array.39? It could conceivably encom- 
pass all the parameters of interior orientation 
used in metric camera calibration, but, as a 
rule, only some of them are used. The selec- 
tion of parameters depends on the accuracy 
required and the data reduction procedures 
applied. 
Self-calibration is based on multiframe 
photography and measurement of a larger 
numberofunknown points. Parameters forthe 
interior orientation are included in the ad- 
justment of the observation equations. Be- 
cause the redundant points are also unknown 
this results in certain geometric conditions 
being placed on the number and arrangement 
of photographs and points in order to achieve 
a strong solution. 
"On-the-job" calibration also determines 
the interior orientation in the same adjust- 
mentas the other unknowns (exterior orienta- 
tion and/or points) but in this case the solu- 
tion is based on a sufficiently large number of 
well-distributed and known control points. 
Self- and "on-the-job" calibration deter- 
mine the interior orientation ofthe same pho- 
tographs as used for the object measurement, 
which eliminates the effect of instability of 
the non-metric cameras. These types of calib- 
ration are often combined with a partial cali- 
bration for the radial distortion, which re- 
duces the number of unknown interior 
parameters, and the number of observations 
necessary for a strong solution. 
Another approach to the problem of calib- 
ration of non-metric cameras, which has not 
been used regularly in close-range photo- 
grammetry, is the method of linear least- 
squares interpolation and the use of statistical 
prediction and filtering. This is used in aerial 
triangulation and digital models for contour 
line plotting. Several other applications in- 
cluding calibration have been demon- 
strated.2° Closely related to this are the mul- 
tisurface fitting methods described by 
Hardy!° and Rauhala.27 It would seem to be 
very worthwhile to try these methods with 
the use of non-metric cameras because they 
can reduce the effects of instability and non- 
reproducibility of the geometry of the im- 
ages.