rking 
brace 
amera 
pecial 
mera, 
ligital 
non- 
aking 
. The 
joint- 
er the 
| the 
well- 
range 
were 
hard 
duced 
netric 
casts 
jxntrol 
taken 
| two 
built 
ndles 
some 
there 
ontrol 
-field 
X-ray 
y but 
] 1/b 
show the two different types. In the first 
case the additional distance 
measurements for controll points 
accuracy resulted 0,1 mm standard error 
of position in the second case this value 
  
Figure 1/a: The test-field 
was 0,02 mm. The accuracy values of 
the control point were proved by the 
index numbers gained from bundle 
adjustment. Can the control points be 
=] 
SO || 
/e 
/ s. 
Je = 
a AN 
  
  
Figure 1/b: The test-field 
161 
accurate within such a geometrical lay- 
out our not? In order to obtain the 
answer we examined the effect of the 
error in the image space taken as a 
function of the principal distance and the 
object distance. We considered our 
control points accurate in case when a 
maximum effect would cause a lower 
value than 5pm in the image space. If 
this condition was not fulfilled during 
least square adjustment even control 
points themselves were corrected. 
Concerning medical judgement deviation 
values below 1 mm may be corrected by 
the dentist, who treats the patient, so this 
valne was considered as a maximum 
error and our intention was to reach the 
mean square error of 0,33 mm of the co- 
ordinate concerning surface modell. 
In order to plan the accuracy of the 
close-range photogrammetrical network 
we need the index numbers 
characteristic of the accuracy of the 
image co-ordinates. If we know these 
values we can produce the condition 
equation of least square adjustment and 
it is possible to calculate the variance- 
covariance matrix previous to 
adjustment. If it is taken for granted that 
there is a network of high value of 
redundancy and it becomes evident from 
the covariance matrix previous to the 
adjustment that the needed accuracy is 
produced by the geometrical  networi: 
taken it can be examined whether the 
omission of certain images and control 
points would make any modifications or 
not and in what proportion. At the 
application of the method -supposing 
that there is no correlation between the 
images- the reservation was made that 
from one stand-point there can be made 
only one photograph. This method needs 
a comparatively great quantity of 
calculation and because our task is to 
develope a new technology and not only 
a single solution of the problem, our 
extra-work will be refunded. Figure 2/a 
shows the previously made photographic 
geometry and the coefficient matrix 
belonging to it. Figure 2/b shows the 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996