The photographs were restituted on a 
ZEISS PSK. For the computations we used 
the computer program for a rigid bundle 
solution (multiple stations, convergent 
photographs), cited in /3/. 
In contrast to Hottier /5/, who discusses 
the obtainable accuracy of close-range 
analytical restitutions, we study here the 
differences in the results between the rigid 
solution and a two step solution with 
neglect of correlations. Thus we obtàin 
some details about the circumstances under 
which both methods can yield equivalent 
results in the field of control surveys. 
II. DATA REDUCTION 
In order to eliminate the influence of film 
deformation and the lack of flatness, we 
transform the image coordinates by means of 
the calibrated coordinates of the reseau. 
For this transformation we consider 3 
models, suggested in /6/: 
1) pseudo-affine transformation 
1 
x = à, + 85X * a4y + a,xy 
1 
y 
ag + ax + any + agxy 
2) perspective transformation 
: 2 
X 7 84 * & x * azY + a,x + agXy 
1 
= 2 
yg = a6 + ax + agy + a,xy + aos 
5) deformational transformation 
: 2 
x 2 84 + asx * B4y * &jyy - agXy 
1 
y = ag + &,X * agy - a,XJ * a 
with x,y measured image coordinates 
? 1 
= , y corrected coordinates 
a; transformation parameters. 
The transformation parameters are determined 
for each mesh of the reseau using the 
measured image coordinates of the reseau 
points and the calibrated ones. With these 
parameters the coordinates of the points in 
the corresponding mesh are transformed. 
The photogrammetric computations with the 
rigid bundle solution are performed with 
the corrected coordinates according to the 
three models as well as with uncorrected 
coordinates, reduced to the central reseau 
point. Although the results of these 4 data 
versions do not differ significantly from 
each other we use, according to /6/, the 
corrected coordinates out of model 3 for 
all further computations. 
COMPARISON OF THE RESULTS OUT OF THE 
BUNDLE AND THE 11-PARAMETER SOLUTION 
TII. 
The base for this comparison are the 3- 
dimensional coordinates of all object 
points, which result from the rigid bundle 
solution. There the orientation parameters 
and the unknown object space coordinates of 
points are adjusted simultaneously, using 
32 
all image coordinates and the coordinates 
of the control points as observations. 
These object space coordinates are com- 
pared with the corresponding coordinates 
out of the 11-parameter solution /2/, which 
is programmed as a two step solution. Here 
the parameters for an orientation or 
calibration are computed in a first step by 
means of object space control. In the 
second step the object space coordinates 
are determined, using these parameters as 
constants. 
For our comparison we use on one hand the 
coordinates out of the orientation version 
and on the other hand those of the on-the- 
job calibration. For the presentation of the 
results we consider the differences of the 
coordinates of all points, represented by 
the mean-values: 
+ Î 
MDX = = Dx 
wy si Y |py| 
n 
wz i Yn 
n 
MDP = = >Voz2 « ny? « pz? 
with DX, DY, DZ differences of the co- 
ordinates 
n number of points in the 
comparison (about 40 ). 
These mean-values are listed in the 
Tables 1-4. In addition we display in the 
last column R of these tables the ratio 
between MDP and a mean object distance of 
about 55 m in per mille. 
The results in the case of the orientation 
are given in Table 1. We differenciate 
between the two numbers of 6 and 18 control 
points on one hand and between the 5 con- 
figurations of camera stations: 
- eonfiguration 1: all 5 photographs 
- configuration 2: photographs 2 and ! 
- configuration 5: photographs 1 and 5. 
The Tables 2-! contain the results of the 
on-the-job calibration for the configurations 
1-5. As the values in column MDP of the 
versions with 6 and all control points 
differ much more as the corresponding ones 
in the case of the orientation, we consider 
4 additional versions with 8, 10, 12 and 14 
control points, by adding further points 
to the original ones which are 12, 14, 21, 
22, 23, 25. 
The comparison of the results of the orien- 
tation and the on-the-job calibration for 
the 3 configurations can easily be carried 
out regarding the Figures 3-5. 
We see that the on-the-job calibration yields 
the same accuracy as the orientation if ten 
or more control points are used. As expected 
configuration 2 with the long base is better 
than configuration 3. 
This increase of accuracy can again be 
gained by using all 5 photographs. For this 
configuration 1 the two step solution, 
whether we use it as orientation or on-the- 
job calibration, differs in the average about 
4 mm in the object points from those out of