7. APPENDIX
In this section first results of the new algorithm implementation
are presented. As explained before, it is a double filtering,
based on Kalman prediction, which utilizes both constant
velocity model and constant acceleration model. The
acceleration addition term is very important in curves (just the
most problematic areas); to confirm that, we show the different
comparison between input data and outputs obtained with
constant velocity model in one case and constant acceleration
model in the other.
Figure 4. New algorithm implemented: input data compared
with constant acceleration model output.
It is clear how the two trajectories, input and output, are quite
overlapped adopting a constant acceleration model in curves.
Figure 5 shows how a certain shift persists adopting a constant
velocity model even in curves where there are the great
problems of distance between tracks. So the choice to
implement a differentiate filtering seems to provide good
results.