nodern
e other
S WG
olution
digital
less in
s were
05
ar
US)
70
002
neras
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B-YF. Istanbul 2004
2. PERFORMANCE EVALUATION OF CARD SIZE
DIGITAL CAMERA
2.1 Experiments
In order to estimate accuracy of card size digital cameras,
experiments were performed using 7 kinds of card size digital
cameras. Figure 3 shows these cameras and Table 1 shows the
major components for these cameras. Stereo images were taken,
and camera calibrations were performed by the resection
method using 9 control points. After the camera calibration,
accuracy was estimated using R.M.S.E. for 33 check points.
Test target were made by the accuracy of =0.01lmm for plane
coordinates and 0.5mm for vertical. Figure 4 shows the test
target. 9 out of the 42 points on test target were used for camera
calibrations, and other 33 black points were used as check
points.
Stereo images for the test target were taken five times with each
camera under the same condition (distance from the center of
camera to test target, length of base line), and image
coordinates for each point were given as the center of area
gravity by image processing. Accuracy for 3D coordinates of
check points was compared with space resection using 9 GCPs.
The average accuracy for 3D coordinates of check points were
investigated in this paper.
pscri DSC
ie m
DiMAGE Xt -
E
“Optio
. IXY DIGITAL30
Figure 3 Card size digital cameras
Q : Control Points
Q : Check Points
==
360(mm Ya € à
360(mm)
Figure 4 Test target
2.2 Evaluation of Lens Distortion
In order to estimate lens distortion for card size digital cameras,
three different lens distortion models shown in Table 2 were
investigated. Model 0 means no correction for lens distortion,
model 1, 2 consider only radial symmetric distortion, and
model 3 considers both radial and tangential distortion (Fryer,
1986).
Figure 5 shows the R.M.S.E. for 33 check points by each lens
distortion models, and R.M.S.E. for the other amateur digital
cameras with more than 3 mega pixels are shown together. It
may be seen from the results that lens distortion for the
armature digital cameras are improved by radial polynomial 3rd
degrees distortion model (Model 1), and model 2 and model 3
do not show any further significant improvements compared
with modell.
Similarly, R.M.S.E. for some card size digital cameras are
improved by modell. However, R.M.S.E for some card size
digital cameras are improved remarkably by Model 3. It may be
said that the reason are depend on some mechanical problems in
downsizing and lightening process such as lens storage
mechanism and multi-layer structural lens, and individual
problem of the specific card size digital cameras. Therefore, it
can be said that lens distortion for the card size digital cameras
should be corrected by using radial and tangentral model,
instead of lens distortion for the armature digital cameras can be
corrected sufficiently using radial polynomial 3rd degree model
as usual.
2.3 Relation between image resolution and accuracy
The relationship between the image resolution and accuracy is
shown in Figure 6 and 7, and the results for armature digital
cameras are shown together. Figure 6 shows the R.M.S.E for
the both camera groups corrected by Model 1. Figure 7 shows
the R.M.S.E. for armature digital cameras by Modlel, and
R.M.S.E. for the card size digital cameras corrected by Model 3.
It may be concluded from these figures that the significant
accuracy are achieved in the case of digital card size cameras
when lens distortion are corrected by model 3.
Table2 Lens distortion models
Model No Correction item Equation
0 No correction.
: ; > dx=krx
1 Radial distortion ;
dyzkry
à - dx = kr x + karly
2 Radial distortion 2 :
dv=krv+kr'y
dx = kr xe kn £2x! )+ ky 2x0
3 Radial and tangential distortion de ky kn 42 2x
Where, d, and d, : correction value,Æ, ,k; and k; :distortion
coefficient
. . 2 2 2
x and y : image coordinate, r” =x" +y
Table 1 Major components of the card size digital cameras
Producer [Product Name Number of Pixel CCD Sensor Size (WX H x D) Weight
(Mega pixcel)
CASIO EXILIM ZOOM EX-Z3 9.20 1/2.5 inch 970x570x229 mm 126g
MINOLTA |DiMAGE Xi 3.20 172.7 inch 84.5 X 72.0 X 20.0 mm 130
MINOLTA j|DiMAGE Xt 9.20 1/2.7 inch 85.5 x 67.0 x 20.0 mm 120g
Canon IXY DIGITAL 30 3.20 1/2.7 inch 85.0 x 56.0 x 23.9 mm 165
PENTAX jOptioS 3.20 1/2.5. inch 83.0 x 52.0x 20.0 mm 98
SONY cyber-shot DSC-F77 4.00 1/1.8 inch 92.6x 71.0x 27.0 mm 152g
SONY cyber-shot DSC-T1 5.10 1/2.4 inch 91.0x 60.0x 21.0 mm 155
