o
*
It is not certain that all partial errors could be observed or studied separately. It will however always be possible
to study some partial errors and groups of them, Errors could be dependent on each other. In some cases a series
of partial errors are dependent on each other in a certain order. We mention diapositive, stereoinstrument, rela-
tive orientation, point identification and operator. The partial errors will in this case add up in a definite order.
It will hardly be possible to study one of them and notat the same time the others, especially if you wish to find
out their relative importance upon the total error. The analysis of variance, especially the type with hierarchal
classification, could be a useful tool for such a study. It will also give the "experimental error” which could be
considered comparable with the "noise" of radio or television. This experimental error is usually a mixture of
the residual errors, of all the sources of error. It will hardly be possible to subdivide this noise and to relate its
parts to special sources of error. That is one of the reasors why we do not believe in the usefulness of the study of
partial errors. Hence we will study the total photogrammetric error.
We will not use the mean of the residual y-parallax of an orientated model as an indicator of photogrammetric
accuracy, because of the following reasons. The y-parallax is a coordinate difference within the model. Con-
sequently it will hardly indicate'the constant and systematic errors of the y -coordinates of the model nor oí the
x- and z-coordinates, It is not probable that a one-dimensional y-difference will indicate the error within the
orthogonal three -dimensional coordinate system of photogrammetry. The y parallax can only give information
of some of the error sources in photogrammetry.
In the Reichenbach Experiment 1962-1964 it was stated, that the differences between geodetic and photogrammetric
xyz coordinates were independent of each other. Large x differences consequently did not indicate large z-errors.
In the Experiment Reichenbach 1962-1964 as well in the Experiment Pecny” 1966 -1968, large errors have been
excluded.
1, PROBLEM
The following problems should be studied.
The errorof a comparison point, for the stereomodels 1 and 2 for the three negative scales 1:3 500, 1:6 000 and
1:12 000,
The error of distance between two series of every fifth pair of all combinations of points for the first model for all
three negative scales.
The frequency and cumulative three -dimensional error distributions of the first models for the three negative sca-
les.
The general photogrammetric x distribution of the three -dimensional error distribution.