The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008
measured with sub pixel accuracy by using centroiding method.
Then target points were matched using epipolar plane angle
method which was proposed by Sabel (1999).
8 B «_
m o o
mm 0
o ** it m
a m
* *
0 D
** o a
m-r
a
if
Q
.
» _
a
* B t g 8
i * m
0
m
i№
i*
Figure 1. Test field
After these procedures were completed, the monitoring stage of
signalized target points on test material under load can be begun.
The entire process is composed of three steps.
• After Each load application, acquisition of the multi
image sets of test material,
• the measurement of image coordinates of target
images which was matched in previous steps,
• The computation of the 3-D coordinates of these
points by forward ray intersection using the results of
the calibration process
•
An algorithm similar to the method which proposed by Maas
was used in order to remove the refraction effect. In loading
tests, test field was placed as parallel to transparent Plexiglas
which was used in front of the test box. The principles of
algorithm are shown in Figure 2. If the point P (X, Y, Z) in
object space is shifted to P (X, Y, Z), the collinearity condition
can be applied for P using the object coordinates of the shifted
point P.
Figure 2. Radial shift for compensation
Only a radial shift by AR parallel to the XY plane has to be
computed for each point relative to the nadir point of each
camera (Maas, 1995).
From figure 2 can be derived:
F 0 tan /?, +1. tan /? 2 + Y p tan /?,=/?
(F 0 +1 + Y p ) tan J3 X = R
With Snell’s Law
n x sin /?, = n 2 sin ß2
(3)
the system describing the multimedia geometry is complete.
Equations (1), (2) and (3) can only be solved iteratively due to
the trigonometric functions. If P is chosen as a first
approximation for P and
R m = ) 2 +(z„-z 0 ) :
(4)
the angle of incidence in the medium nl in the first iteration
becomes
/?, = tan
l (0)
V^o +/ + y p j
(5)
The angle of incidence in the plexiglas media according to
Snell’s Law and the correction for R<o) are computed by using
following equations.
ß 2 = sin 1 sin /?, j
(6)
AR _ ( sin (A - ßi )
cos /?, cos ß2
(7)
R — R(Q} + A R
(8)
The equations (5)-(9) are used iteratively until the difference
between consecutive AR values are equal or smaller than the
predefined threshold value (0.001 mm). And finally the
coordinates of the radially shifted point is computed.
X p =X 0 +(X p -X 0 )R/R
l p =Z 0 +(Z p -Z 0 )R/R
The collinearity condition for cameras Cj can then be used with
the radially shifted point Pj (X, Y, Z) instead of P (X, Y, Z)
(Maas, 1995).
720