^ 
j 
   
but it is my experience that the distortion of most quality-lenses can be expressed with this 
equation, The parametres a3, ag and a7 are to a great degree correlated as described in [10] 
but as long as only these 3 terms are used, the correlation problems are not prohibitive. The 
ortogonal expression suggested in [7] may be worth considering, but I haven't tried it yet. 
Calibration 
The general idea of making the total camera calibration from control points on the object 
to be surveyed, is of course feasible, but rather unpractical. Some of the camera parametres do 
not alter from picture to picture. These parametres can more conveniently be calibrated in a 
laboratory testfield once and for all, leaving only the unstable camera parametres to be calibrated 
for every picture. 
It is also recommendable to test the camera in a laboratory testfield before using it for ‘ 
photogrammetric measurements. The stability of the camera parametres can be tested, and a 
a priori value for the residual error can be found. Laboratory testfields have been constructed 
with varying degree of sophistication [2] and [12]. 
If the purpose of the testfield is to calibrate cameras for measuring purposes and getting 
a a priori knowledge of what a camera will perform, it is important, that the conditions of the 
testfield are similar to the operational conditions under which the photogrammetric measurements 
are to be made, I have been working with testfields consisting of 40 to 70 points placed in two 
planes with the size ranging from 1 x 1 x 0,5 metre to 5x 8x 3 metres. To achieve operational 
conditions I have used testfields with approximately the same size as the object to be measured, 
thereby escaping the problem of variable distortion [1]. The coordinates of the testfield points 
can be measured by intersection using 2 minute theodolites for about 4 hours with a method 
described in [5]. The accuracy of the testfield points is not very good (0. 1 mm to 0. 25 mm) but 
will have the same accuracy as the control points used later in the object. € v 
In such a laboratory testfield all camera parametres are calibrated. Taking several 
pictures, the stability of the different parametres can be investigated. Testing different kinds of 
film material can also be done in the testfield. 
The radial distortion of the lens does not alter from picture to picture and can be 
calibrated once and for all. For many cameras the housing and focal plane frame are so stable, 
that the angles à and B can be given the fixated value of zero. 
Remaining to be calibrated in every picture are the 6 exterior orientation elements 
X ; Yo ;Z9,0,9,k ), the principle distance c , the coordinates to the principle point Xd 
and yo and in the cases where a bellow-camera is used, the two angles a and B, giving 9 or 
11 unknown parametres. Instead of the usual minimum of 3 control points in every model, the 
minimum requirements are increased to 5 or 6 control points. The control points must, however, 
be distributed in such a way, that 2 of the control points are situated outside the plane of the 
others. 
Computing and programming 
  
Renouncing the advantage of the calibrated, stable and accurate photogrammetric camera 
and introducing cameras with very variable characteristics, the demand on the analytical 
treatment and the requirements of the computing program rise considerably. 
Not only the cameras vary, in the non-topographic field two surveying procedures are 
seldom alike, and new measuring tasks will exceed the capability of existing programs. It is 
therefore essential to have a program system, where alteration and extension are relatively 
simple to perform. 
With the modern electronic computer it is possible to solve any photogrammetric 
problem, if you have time and patience enough. It is therefore essential to set up some priorities. 
If not, you can end up with a very sophisticated but very unpractical program system. 
I have set up the following priorities : © - 
1) To incorporate error-finding procedures in the programs, so that gross errors can be found 
and automatically removed, and failures in the surveying procedure will be expressed clearly in 
the output. 
Th on