Also the case where an extra point is added in the center of the 
model was investigated. This point does not participate in the ad- 
justment. It can be considered as a detail point (cadastral) point, 
a situation usually encountered in photogrammetry. 
The following stochastic models were considered for the observa- 
tions. 
For the model points: 
- equal precision of 10 cm on the ground for x and y coordinates 
with no correlation between them. These will be referred to as 
uncorrelated model points. 
- a full covariance matrix is used for model points. The determina- 
tion of this matrix is based on the work of Ligterink [9], taking 
only the observational errors into account. These errors originate 
from inner orientation, relative orientation and from the measure- 
ments of model coordiantes. The measured points are considered to 
be signalized. These will be referred to as correlated model 
points. 
The latter could be a better model than the former, but it should 
still be considered with care, since correlation between models is 
not taken into account [4]. 
Ground control points 
- not stochastic 
- equal precision of 5 cm for the x and y coordinates, and no corre- 
lation between them. 
The generated independent models have 207 side overlap and 602 for- 
ward overlap. The scale of the photography has been assumed to be 
1:10,000 and the principal distance to be 152 mm. 
3.1. Analysis of the results 
The values of the parameters of the choice functions which generate 
the best substitute matrices are summarized in tables I and II. The 
ground control is not stochastic. 
The first vertical column in these tables refers to the type of 
choice function used. The second column refers to the block configu- 
ration e.g., Al, A2, etc. (see also fig. 2). 
The first horizontal line indicates the tie point configuration. 
The second line indicates the parameters of the choice functions. 
In all the cases {7 as = |. 
From these tables the following conclusions can be drawn: 
- the exponential and logarithmic choice functions give comparable 
results, with the logarithmic function slightly better in most of 
the cases. The linear choice function gives very large values for 
the ratio (Al. x pin: 
- when double points are used, the ratio Ixdaax/i^]ala 
becomes smaller compared with the corresponding cases of single 
points. Further, the ratio increases with a decrease in the number 
of control points in the block (relaxed control). 
- when six points per model are used, the ratio becomes larger com- 
pared to the ratio of the corresponding cases with four points per 
model. This may be caused by the non homogeneity in the