11 
be lengths, angles, or lengths and angles combined. Usually three lengths are 
used, one of which is the principal distance (camera constant) i.e. the length 
of the perpendicular from the center of projection to the image plane. The 
other two quantities locate the principal point i.e. the foot of the perpendicular. 
This is the classic definition of interior orientation. Very often the fiducial 
marks are used for other purposes too, such as control of and correction for film 
shrinkage. Then there are other criterions on their number and positions. 
In practical work the classic definition of interior orientation is sometimes 
too crude to give good results. First there are two projection centers, the exte 
rior and the interior. Roos [23], has discussed other definitions of interior orien 
tation. The next extension is to the physical definition which can be computed 
for a lens using the principles of geometric optics to determine the aberrations. 
This definition includes the radial distortion of the lens. Lastly Roos suggests a 
technical definition which also includes the manufacturing irregularities of the 
lens determined empirically for every lens. 
There are, especially in American literature, more sophisticated definitions 
of interior orientation. They contain information on asymmetric radial distor 
tion, so-called tangential distortion, point of best symmetry etc. The ISP recom 
mendations for calibrating cameras [5] contain concepts related to interior ori 
entation. It seems as if some of these concepts have been introduced because 
of the method of calibration rather than the necessity for the reconstruction of 
the bundle of rays. 
The interior orientation contains data from which it is possible to recon 
struct the bundle of rays on the object side of the lens. This has been emphasi 
zed by Roelofs [22]. The reconstruction can be done in a more or less sophistica 
ted way, depending on the number of data used to define the interior orienta 
tion. There has to be a correspondence between the data used to define the 
interior orientation and the data necessary to reconstruct the bundle of rays in 
a certain procedure. This is of importance for the accuracy, which among other 
things depends on the number of, and the nature of the elements of the interior 
orientation. We have to choose an interior orientation that corresponds to rea 
lity as well as possible. This means that the nature of reconstruction of the bund 
le of rays defines the type of elements used in the interior orientation. For exam 
ple, if the reconstruction is done using the classic definition, then the calibra 
tion has to be done using the corresponding three elements. If the reconstruc 
tion is done analytically in an electronic computer using corrections for radial 
and tangential distortion, then the calibration has to be done with these para 
meters included in the interior orientation. 
Choosing the mathematical model for the interior orientation to suit the me 
thod of reconstruction of the bundle of rays could be called a statistical defini 
tion of interior orientation. First we define a mathematical model for the in 
terior orientation, which corresponds to the intended evaluation procedure i.e.