9 
March 19^4, the total number of 3»975 models of aerial triangulation were 
executed and 2,223 models (37%) of them were remeasured. Causes of re 
measurements are analized as seen in Fig 6. It is interesting that the 
considerable parts of the remeasurements occur due to the bad identifica 
tion of major controls. 
Next we will compare the residual errors at fiducial marks between 
the first and second or third measurements for the case shown in Table 1. 
We may see from the Fig. 7 and Table 1 that 1) in spite of the relatively 
high occurences of remeasurements, the cross form of fiducial marks at the 
corner measured by comparator have the highest precision and the standard 
deviation thereby is 6.9 microns for 3»952 times of remeasurements, 2) 
wedge type of fiducial marks at the lateral sides shows the standard 
deviation of +_ 10.8 microns and it is a remarkable fact that in this case 
the occurrence of the errors larger than 21 microns is quite large compared 
to others, and 3) we usually have to anticipate the occurrences of 3% 
of the errors larger than 10 microns, in practice. 
5. Discussion on relative orientation 
Purpose of relative orientation in our analytical system are besides 
the computation of outer orientation elements, to aim at the detection of 
mistakes or* errors of photo coordinates mainly in y direction. Mistakes 
or errors in x direction reveal themselves only for the points which are 
common in adjacent models. 
Confirmation of errorneous points are made half automatically on the 
computer. This phase of operation will be described in other paper in 
more details. 
The phase of this procedure is very usefull to detect the bad points 
definitely. But it appears that for this procedure the 6 points system 
for relative orientation in a model is insufficient in number. For by 
the omission of any one point, the solution become unique. Although in 
our system every points in the model are used for relative orientation 
and often there comes some points other than fundamental in a model 6, 
sometime it occurs that the model has no point other than 6 fundermentals. 
After some experiences we decided to use 8 fundamental points for relative 
orientation to avoid the inconvenience against the elastic operation of 
free rejection or insertion of any point before the electronic computer. 
In such a case the distribution of the points necessarily becomes as