Vittorio Casella
4.1 Results for the Optech sensor
The Optech sensor is able to measure the first pulse, as well as the last, for each emitted ray. So we verified, first of all,
the differences between the coordinates of the first pulse and the coordinates of the last pulse. Provided that they refers
to a flat surface, they should coincide. Please notice that there are 357 points inside the court, so our estimations are
good and reliable.
AX AY AZ
m -0.009 -0.021 -0.075
s 0.008 0.020 0.050
Table 2: Differences between first pulse and last pulse ( F-L), in metres.
Then we computed the differences between the measured height values and the true value, determined by GPS. The
following table summarizes results for the first and last pulse.
AZ, AZ,
m -0.246 -0.170
S 0.152 0.142
Table 3: Mean value and dispersion of the height errors, for the first pulse and last pulse, respectively (metres).
It is remarkable that both tables shows that the first pulse is lower than the last. Besides, there is certainly a small sys-
tematic error and, to finish, the random error has the correct order of magnitude.
4.2 Results for the Toposys sensor
The tennis court check-area is particularly interesting for the Toposys data we have, because four stripes lie on it, two in
the first pulse mode and two in the last pulse mode. Therefore it is possible to verify the agreement between them. For
each stripe, there are approximately 1200 points inside the court. The following table shows the main results.
AZ (31218) | ^AZ (31219)| AZ (31306)| AZ (31307)
m -0.275 -0.270 -0.228 -0.283
8 0.050 0.050 0.052 0.050
Table 4: Analysis of the errors for two last pulse stripes and for two first pulse stripes, respectively (metres).
We deleted five points from the first stripe, because they have residuals around 25 metres. Apart from this problem,
there aren't differences between the stripes. Also in this case there is a significant systematic error: it has the same order
of magnitude and the same sign of the error of the other instrument. The dispersion, on the other hand, is smaller than
we expected.
5 CONCLUSIONS AND ACKNOWLEDGMENTS
Even if the prevalence of systematic errors on the random ones will require a specific tuning of the our proposed meth-
odology, we think it is interesting and effective, so we'll go on, fully applying it to a bigger number of check-areas,
which we are already measuring.
162 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.