(484) 
the parallax measurements in several points and the introduction of computed 
corrections to the instrument. Though these manipulations should have given a 
gain in precision, this gain should not have been able to compensate the aggrava- 
tion of the operations and computations. The operator forms the stereoscopic 
model in the way he likes best and measures the parallax in the normal six 
points. These parallaxes must be zero with a tolerance of 0,02 mm; if not: the 
model has to be perfected by calculation. 
If more serious deformations are found these deformations are to be traced 
by the special method of Prof. Poivilliers. 
Reduction of the systematic errors in the distortion. 
These errors are still the most important ones. The method of combination 
of aerial cameras with plotting-cameras as indicated in the Hague gives no suf- 
ficient solution for the tracing of the small deviations in the symmetry. The 
measurements are therefore completed by a study, with the help of the photo- 
goniometer, of the variations of the angle under which the ray of a circle, cen- 
tered on the principal point of the camera, is seen. Generally it will be found 
that this circle is transformed to a, more or less regular, ellipse. The lens of the 
plotting camera must be orientated, in its mounting, in such a way that the two 
transformations (by the lens of the aerial camera respectively that of the plot- 
ting camera) coincide as good as possible or as adjustment-points of the image 
points have to be selected situated on perspective rays ending in four common 
points of the two curves. In this way the influence of the distortion is rigorously 
eliminated. The application of this method calls for the realisation of a new 
photogoniometer, the ones in use being not practical and precise enough for these 
operations. 
Output of an instrument: A double passage of 45 models per month in 
average. 
Second communication: Computation and Compensation. 
The presently used methods are substantially different from the methods 
exposed in the Hague; one has tried to apply simple methods, accessible for all 
common computers. Therefore certain correction terms are neglected and only 
very simple compensation processes are used. This is admissible and the result- 
ing precision is sufficient thanks to the precautions taken to reduce the opera- 
tional errors and thanks to the small bridged distances (about 5 models in aver- 
age). The case of long strips without reorientation (not yet practically applied) 
will be treated further on. 
The method can be divided in three successive phases: 
1°). Determination of the elements of scale and absolute orientation of the pre- 
pared models. 
2°). Computation of these elements for all models and compensation of the 
systematic errors. 
3°). Computation of the coordinates of the pass-points and compensation of 
the accidental errors. 
Determination of scale and absolute orientation of the prepared models. 
This determination must be executed with utmost precision; therefore the 
number of ground points in each model must be superfluous. With the help of