Of course there is a second way to judge the precision of the
adjustment result, that is by simply looking to the outcoming
standard-deviations of the x- and y-coordinates. This method, however,
has one disadvantage namely the dependency of the chosen S-base. This
means that an other S-base results in other standard deviations, so
one has to be very careful.
However, to be honest, it should be stated that applying the general
eigenvalue problem has also a disadvantage that is in terms of storage
capacity of computers. For each of the, relatively small, blocks about
5 Mb is needed. This results in the fact that in practice the precision
evaluation would not be carried out with the mentioned eigenvalue
problem, but by judging the outcoming standard deviations.
But still for research purposes the eigenvalue problem is a very handy
tool, because not only the precision of the whole network can be looked
upon, but also parts of it. In this way weak parts of the network can
be traced. However, there is one condition that has to be fulfilled.
When dealing with partial networks or blocks both the G- and the H-
matrix has to undergo an S-transformation to a base (and again the
same base) within the partial block, but this can be done quite easily.
5. THE RESULTS OF THE EXPERIMENTS
As mentioned the experiments were carried out with three different
block configurations. The total number of experiments was six, that
is two experiments with block 1, one with block 2 and three with
block 3.
For all experiments the following results are presented.
|. precision of x- and y-coordinates after the free block adjustment
(first phase) in terms of eigen values om
2. precision of the x- and y-coordinates after the pseudo least squares
adjustment in terms of eigen values
3. reliability of ground control points, calculated in the second phase
of the adjustment, in terms of boundary values.
The precision results are shown in two tables. Table 1 shows the
maximum eigenvalues resulting from all experiments. |
In tables 2 to 5 the boundary values of the control points are listed
for each experiment.
Before evaluating the results of the six experiments first some
additional information about each experiment will be given.
experiment |
block used: see figure 2.
distribution of tie-points: 4 per model (see chapter 3.1).
number of tie-points: 120
number of control points: 30
distribution of control points: see figure 2
chosen partial networks with respect to the eigenvalue
calculations:
- 59 -