If the block adjustment is based on the well proved planimetry- 
height-iteration, as used in the computer program PAT-M43 |14|, 
8 correction terms refer to the planimetric block adjustment and 
6 are used with height block adjustment, The corresponding para- 
meters, called pi to pg and hj to hg again are formulated as 
orthogonal to each other and with respect to the transformation 
parameters of planimetry and height. The formulation of the 
correction terms and their effect on the model points is shown 
in figure 4, 
- Figure 4 - 
By the additional model parameters of figure 4 the correction 
terms suggested in |12| are surpassed, 
With ordinary applications of aerial triangulation the functio- 
nal models as presented here guarantee a fully adequate compen- 
sation of the data inherent systematic errors and can serve as 
standard models. In case of essentially more points per image 
or model (e.g. cadastre) however, the use of further correction 
terms can be suitable. 
3.2 The Stochastic Model 
First of all it seems to be obvious to treat the additional 
parameters as free unknowns, as done in |8| and |11|. This 
would lead to the following formulation: 
V = Ax - By - f (la) 
In (la) f is the observation vector, containing the measured 
image or model coordinates, x denotes the vector of unknown 
terrain coordinates and transformation parameters and y is the 
vector of the unknown additional parameters. 
Two facts however, are ignored with the formulation (la): the 
relatively small size of the systematic errors and the fact 
that they vary from project to project with regard to sign and 
Size (the theoretical mean value is zero). Therefore it is more 
suitable to treat the additional parameters as observations of 
amount zero with appropriate weights. This can be done by 
keeping (1a) and adding the following set of observation equa- 
tions: 
Vo = yır 9 (1b) 
The weights of the additional parameters can simply be chosen 
according to the expected amounts of the correction terms or 
somewhat smaller, 
If some of the additional parameters can be derived from cali- 
brating data, which are representative for the actual practical 
project, the obtained amounts can directly be introduced into 
the corresponding lines of the observation equations (1b), re- 
placing the amounts zero, The weights of these parameters can 
then be determined from the accuracy of the calibration. 
The formulation (la), (1b), which is also used by Brown |9| and 
others, leads to banded bordered normal equations, with the 
additional parameters forming the border. In that way favourable 
computing times are guaranteed. 
-— ("9 (73 
in 
ape 
Foe 
ae gu 
pa.