108
Mus
i.
LE
108 b t
107 4
d
ios r1
s
105 =
8
a3
108 2
x
LE
103 si
3
Yi /
Figure 12. Graphic representation of the comparison, carried
out with the above mentioned program, between the
photogrammetric sections and the corrisponding sections on the
registered cloud of points
We have chosen to use the array of sections, obtained by photo-
grammetric models, as a reference for testing laser data.
In fact, they correspond to the following requirements: they are
representative in a homogeneous way of a tested area, the accu-
racy is known and they are indicative of the shape of the object
with a sufficient density, able to be compared with laser scanning
data. On the other hand, the measurments taken with topographic
and photogrammetric techniques are normally referred exactly to
those elements (edges, cornices, ...) not well-defined with laser
scanning. Therefore, these deviations are not related to the nodes
of the photogrammetric plot, but to those points which are able to
LtegaíTileFot() íegatifiletssii
Photogrammelric plotting Laser scanning acquisition data
acquisition data (section) (strips of points)
OráinaFunttlase:sl
Ordering laser scanning deta by 2
decreasing
Interpularet{}
Medium point definition for each
plotted segment (calcule z valo for
following comparison)
5ráinabatiCoepistiil)
Compilation of ordered st of all
L—— points {photogramenetric and ker
scanning data}
input. ange width for searching
laser scanner data ta be considered
Calcoialinterpolartionilaserfesata(]|
1 ch 4 Interpolation CalcolalnterpolarionilaserMirnguad(]
Vyesaht decraising with square ^
distance from reference point ee Sarna dola wee Least square interpolation
eriteria
ViscsiitfrfsRiseltatif]
Output result report
Figure 13. Flow chart of the program aimed at the
comparison between the sections
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004
better describe the surface. The aforesaid points were chosen as
medium points between two subsequent nodes of the plotted
polylines. At the end, we evaluated the deviations between the
photogrammetric sections and the corrisponding sections of the
registered cloud obtained as described in the sequel.
The residuals are < +1 cm.
4.4.1 Laser scanning sections: To make a point by point
comparison between the photogrammetric plots and the laser
scanning, with an automatic procedure, a Visual Basic program
was developed. The flowchart in figure 13 shows its working.
1. To have an important sample of laser data, strips of points, 1.5
- 2 cm wide, were cut out. The corresponding photogrammetric
section was plotted at x value equal to the center of the strip.
2. After having verified that there are not considerable variations
in x direction, in the examined thickness of the strip, x value has
been ignored: this is like projecting the points of every strip on zy
planes.
3. Photogrammetric data are ordered according to the sequence of
the plotted polyline vertices. For each segment a medium point is
calculated and its z value is stored for consequent comparisons.
4. A subroutine orders laser scanning data by decreasing z. This
criterion is true in the portions with a vertical tendency but it is
not appropriate in the horizontal one, where z value does not
indicate the correct subsequence of the laser data. In our case, the
sections describe prevalently vertical surfaces; other criteria should
be defined to generalize the approach.
5. Different search ranges can be defined around the previously
calculated medium point (step 3); in order to better define it, both
the quantity of plotted points and the object shape have to be
considered.
6. A laser data interpolation is made with the purpose of defining
those points with an y value corresponding to the previous stored
z value. Interpolation can be performed according to two different
criteria:
a - the points included in the search ranges are weighted with
inverse square distance from the point used as reference;
b - a linear regression based on least-squares method is applied.
7. At the end, for each calculated z value, both photogrammetric
and laser scanning y value, are compared and visualized.
5. CONCLUSION
In this paper an accuracy evaluation of an architectonic 3d model
by laser scanning has been carried out. The main phases of elabora-
tion have been tested and a comparison with reported accuracy val-
ues by the producer has been done. In general, we can say that the
results are confirmed, also after registration, on the cloud of points
but that the value proposed as *Modeled surface precision" (4 2
mm)is valid only in case of simple surfaces (level surfaces, spheres...)
while with complex surfaces (as the most common cases in the ar-
chitectural survey) it is not quite favorable, even though it remains
in the accuracy range required in the cultural heritage survey.
ACKNOWLEDGMENTS
The research has been financed by Italian Ministry of Education,
University and Research (MIUR) project COFIN 2002 (Research
Group Resp. Prof. Bruno Astori). The authors greatly thank: prof.
Franco Algostino for his helpfullness in the data management
ing. Marco Nardini (Leica) for laser scanner aivailability: arch.
Luigi Venezia and arch. Andrea Violetti for their cooperation in
the apse modeling.
KE
AB
Un
Ho
con
and
Th
arc
T
obs
des
prc
OV(
gra
CO
Se
sol
the
gel
the
gre
an
ap]
fea
art
sit
the
un
ac
