. Is referred to
the Procrustes
the x and y
greater than 6
rent technique
ates a better
he Procrustes
o observe that
ors, Procrustes
1al to zero for
this case, the
ution, without
applying the
wy between 2
ithm, the root
r the classical
0 centimetres.
random error
t of the whole
or. Finally, if
1e differences
le for the two
the root mean
f fixed points
points for the
n 1, first, the
'ordinates and
lgorithm (see
veen the true
a traditional
oper symbol).
> error is only
ie Procrustes
ues, while the
20-30 cm.
s
onditions
X COORDINATE DIFFERENCES
WO aon oto po
[mj
Point ID
Y COORDINATE DIFFERENCES
T $^. et # a in A
Point iD
Figure 7. Residuals for mixed systematic and random error
conditions (Procrustes €, traditional )
The graphs of Figure 7 refer to the case in which systematic and
random errors are both present; in this case, a generally
significant better result is evidenced when applying the
Procrustes method with respect to the traditional one. It is worth
of mention the fact that the residuals are perfectly comparable
with the applied random error values, emphasising in this way
the capability of the Procrustes algorithm to compensate the
systematic component also in the case in which it is
superimposed to the random errors.
X COORDINATE DIFFERENCES
{m}
fel ee /
Point ID
Y COORDINATE DIFFERENCES
v Te INR
Im]
Point ID
Figure 8. Residuals for pure random error conditions
(Procrustes ®, traditional )
The above graphs (Figure 8) consider the case in which only the
random errors are present: the results are generally similar.
As the previous results show, the Procrustes technique seems
very powerful in the presence of systematic errors, and can be
used also in the case in which only random errors are present:
this is true if and only if the fixed points adequately cover the
network taken into account.
S. A NUMERICAL EXPERIMENT ON A REAL CASE
The Udine branch of the Italian *Dipartimento del Territorio*
made available the coordinates values and the description of the
fiducial points surveyed since 1990 in the North-West part of
the Udine municipality. The network is constituted by 64
fiducial points, covering an area of almost 2.8 km”; the number
of total polygons used for the various surveys is equal to 205
(see Figure 9, and Figure 10). The initial experimental phase
consisted in the GPS measurement of 18 fiducial point
positions. Of these points, 9 were considered like fixed points,
and homogeneously distributed over the entire network, while
the remaining 9 have been used to control the results.
Figure 9. Spread of the 205 adjacent, partially or totally
overlapping fiducial polygons composing the real network.