. o, [pixels?] N. of iterations Max error
Test
Slope 0 1.38 107? 10 110?
Slope 0.005 1.58 102 10 1102
Slope 0.03 1.15 102 10 21072
Table 2: bundles adjustment results.
The determination of the ground point coordinates in the
simulations lead to the following results
(AZ = Z (estimated by the plotter) —Z (a priori known)):
Test Max error | mean AZ mean formal
AZ 0, Std Dev of Z
Slope 0 0.0 0.0 : «107?
Slope 0.005 0.296 -0.004 0.0029 0.006
Slope 0.03 0.042 -0.003 0.0004 0.006
Table 3: plotting results.
5.2 Tests on real objects
In these tests only approximate orientation parameters
and object shape are known.
Three images of a cilindrical object with radius of
approximately 30 cm are taken by a CCD camera, as
shown in the following scheme.
E Y Y 7 7
\ ‘ ‘ i :
bd
Figure 4: taking scheme
The digital images are obtained with a frame grabber with
no special care for clock alignment and a resolution of
640x480 pixels.
Least square matching results are satisfactory, with a
standard deviation of the adjustment among 107$ and 5
1073, and standard deviation of the image coordinates x
and y less than 0.110 e 0.085 pixel respectively.
Images orientation is performed setting a conventional
reference system fixing all the orientation parameters of
the first image and the Xo coordinate of the third. The
procedure is performed with a standard deviation of
0.102 [pixel] and 10 iterations.
In this test a "ground truth" is not avilable to check the
real precision of the procedure. To evaluate the results,
besides a first qualitative judgment of the plot of the
obtained points as shown in the image below, a
parametric model of the surface has been built.
422
Figure 5: obtained surface.
Assuming that the plotted points belong to a cylindrical
surface it is possible to esimate in a least square
procedure the five parameters that describe the cylinder's
size and its position in space.
It is now possible to estimate the distance of each point
from the surface. This value can be considered the error
of the determination of that point. For the points in the
tests this error has a mean of 0.147 mm and a standard
deviation of 5.056 mm, the number of points is 518.
6. CONCLUSIONS
The system that has been described can be considered a
prototype of a complete digital photogrammetric system.
Several test of different digital photogrammetric
techniques, not described in this paper, have been
carried out to select those suitable for real applications.
The development of the user interface is carried out
thinking of a real production enviroment, giving the user a
simple, consistent and standard interface, as well as
default values for most of the required parameters.
The simulated tests show the functionality of the system;
the real test shows the precision of the overall procedure
that, taking into account the quality of the images, can be
considered satisfactory. A further step in the
development of the system will be the integration of the
digital sensor calibration techniques, currently under
study at our laboratory.
7. SELECTED BIBLIOGRAPHY
1. Papers about researchs developed in the same
department and related to the present work:
Benciolini, B. 1990. The Observation Equations of Digital
Phototgrammetry. ^ Proceedings of the ISPRS
Intercommission WG 3/6 Tutorial: Mathematical Aspects
of Data Analysis, Rhodes, Greece.
Benciolini, B., Sguerso, D., 1996. Digital image matching
in photogrammetry. Bollettino di Geodesia e Scienze
Affini, anno LV, n. 2, pp.178-190.
Zateli, P., 1994. Progetto, realizzazione e
sperimentazione di un sistema fotogrammetrico digitale.
Bollettino della SIFET, n. 3, pp. 135-154.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996
nS
—— pease Mm NN —
IN "TS 73 “WW
PP
Q) Mm
Q) hmmm OD) YY
O £73
mo oO