International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B4, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
3.3 Results
Three kinds of parameter estimation were conducted in our
experiment. First, azimuth angles for the digital camera were
estimated with a given viewpoint taken from the laser scanner.
Second, the X and Y positions of the digital camera were
estimated for given azimuth angles taken from the total station.
Finally, both the positions and azimuth angles (3-DOF) of the
digital camera were estimated. These estimations were
conducted for both a wide spatial range (10 m) and a narrow
spatial range (1 m).
3.3.1 Azimuth angle estimation: Azimuth angles for the
digital camera were estimated using each matching point
detected from 1,800 candidates, as shown in Figure 10. The
horizontal axis refers to the camera image numbers described in
Figure 7. The vertical axis indicates differences from the true
value measured by the total station.
— OO DD DD © O
c 5 0 &icoli£mosoco
Difference from true value [deg]
1214343414515
Camera image number
Figure 10. Estimated azimuth angle
3.3. Camera position estimation: The X and Y positions
for the digital camera were estimated using each matching point
detected from 441 candidates. Figures 11 and 12 show results
for the wide and narrow spatial range, respectively. The
horizontal axis refers to the camera image numbers. The
vertical axis indicates horizontal distances from the true value
measured with the laser scanner.
m
* ^O x3 oo o C
ococococccoco
ox
e
Distance from true value [c
ro 4x
© ©
07:02:34 $ 413 9 ND
Camera image number
Figure 11. Estimated camera position (wide range)
m
Ur "Ovi oo oO cC
99090988
ce
Distance from true value [c
bs] +
e >
10:
Lakh
3% 4.85 48.7.3 9
e 10 n 12
Camera image number
Figure 12. Estimated camera position (narrow range)
3.3.3 The 3-DOF estimation: Azimuth angles and positions
for the digital camera were estimated using each matching point
detected from 73,800 candidates. Figures 13 and 14 show the
results for the wide spatial range, with Figures 15 and 16
showing the results for the narrow spatial range. The horizontal
axis refers to the camera image numbers. The vertical axes in
Figures 13 and 15 indicate horizontal distances from the true
value measured with the laser scanner. The vertical axes in
Figures 14 and 16 indicate differences from the true value
measured with the total station.
250
— 40
230
$ 20 i
oz
> LO i
Ey WH
E19
[5]
9.3.0
o
a -90.60
25 153 73-74 3-6 7:8 9 10 n3
Camera image number
Figure 13. Estimated azimuth angles in the 3-DOF (wide range)
m
vdi 00 OD C
SS SS SS So SS
Distance from true value [c
1 (25 (4 55.0, 7 9 9.19 H 12
Camera image number
Figure 14. Estimated camera positions in the 3-DOF (wide
range)
012 3453 1:514 9 1 WP
Camera image number
Figure 15. Estimated azimuth angles in the 3-DOF (narrow
range)
£ 100
= %
o
= 80
S 70
3 60
= do
o»
= 20
£ 10 I
uw
B e au 7 8 9 10H12
Camera image number
Figure 16. Estimated camera positions in the 3-DOF (narrow
range)
374
Inte
From the
our appro
cloud via |
parameter
angle esti
reliably t
therefore |
indoor sp:
positional
Second, tl
azimuth ai
spatial ran,
result also
12. From t
stand-alon
indoor po:
accuracy W
Finally, bo
were estim
independer
parameters
approach c
achieve hig
indoor pos
LAN at 1€
improve th
the positio
angles.
When we a
for image n
15 also sho
We assume
and the ren
because the
Which the |
laser-scanni
missing poi
Therefore,
missing poii
values in thi
then differ f
reflection or
reason.
Although w
other data c
number of «
values taken
be an effecti
example, gy
azimuth ang]
Although the
We could p
massive poir
In addition,
achieve higl
angles,
Currently, th
useful in pre
Issue. Our
matching fro
Image proce,
computing m
Procedure, V
follows, The
