inbul 2004
pating in
letermined
;, or based
> function,
ls just on
r position.
e function
(1)
ind K for
es) can be
points (over
lated and
covariance
Unknown
| based on
f "clean?
is equal to
this value
'alues will
calibration
coordinate
sidered in
>diction is
covariance
needs of
tended for
but some
1g are also
Ns IS Very
2000 has
olynomial
models for
tering. All
ell as the
using this
bility and
ned during
et for the
] value of
ic errors,
errors
olution of
1 scanning
I'he calibration process itself has several steps:
|l. multiple glass grid plate scanning,
2. measurements of the grid plates and processing of the
results,
3. analysis of achieved results and acceptance of
transformation parameters
If the results confirm that scanner satisfies the accuracy
standards, then every scanned image is resampled in batch
procedure based on accepted transformation/calibration
parameters.
. 3. CALIBRATION OF EPSON EXPRESSION 1640XL
SCANNER FOR THE PHOTOGRAMMETRIC
REQUIREMENTS
3.1 Epson Expression 1640XL
This scanner (Figure 1) is selected because it belongs to the
higher class of desktop scanners and it comes from very well
known manufacturer. The market price varies from 2000$ to
30008.
Figure 1: Scanner Epson Expression 1640XL
The basic technical features are given in the Table 2.
SCANNER TYPE Flatbed color image scanner
SUB SCANNING METHOD Movement of the mirror
PHOTOELECTRIC DEVICE Color CCD line sensor
SCANNING AREA
Scanning Platen 310 x 437 mm (A3+)
Transparency Unit 290 x 420 mm
LIGHT SOURCE Xenon fluorescent lamp
1,600 x 3,200 dpi with Micro
Step Drive
OPTICAL RESOLUTION
OPTICAL DENSITY 3.6 D
PIXELS/LINE 58,560 pixels
OUTPUT RESOLUTION
GREYSCALE 14-bit (16,384 grayscale levels)
COLOUR 42-bit (4,398 billion colors)
SCANNING SPEED
MONOCHROME (Draft A4 300 dpi - 14 sec”
Mode)
Table 2: Technical features of EPSON Expression 1640XL
scanner
3.2 Grid plate scanning
Glass plate (25.5cm x 25.5em) has been used for the geometric
calibration of desktop scanner. The plate has 625 grid points
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004
with lem grid spacing. Geometric accuracy of the grid is
checked by measurements using analytical photogrammetric
plotter.
Scanning of grid plate is performed for several spatial scanning
resolutions (1000 dpi, 1200 dpi and 1600 dpi) and for two
radiometric resolutions (grayscale and RGB) Grid plate
scanning was performed regularly on every hour, with scanning
of other images between these scans. In total, it was 24 scanning
of grid plate, with different scanning parameters.
3.3 Processing and result analysis
After grid plate scanning, grid points are measured in DigiScan
2000. The result of the measurement are the pairs of grid points
coordinates, that is pixel coordinates of grid points and
appropriate theoretical coordinates in plate coordinate system,
based on point position in the grid. The pairs of coordinates are
subject of further analysis on scanner geometric accuracy.
3.3.1 Assessment of overall scanner error and systematic
part
The first mathematical model applied on measurements 1s
Helmert transformation (eq. 2). This provides simple
transformation of coordinates from raster to plate coordinate
system.
A sabtatx-bt*y
Yebtatysbsx
where: X,Y — grid point coordinates in plate coordinate system
(in Um)
X,y - pixel coordinate of grid points
a,b,al,bl - coefficients
Achieved RMSE based on residuals after transformation is 130
um. Overview of errors vectors after applied Helmert
transformation is given at Figure 3.
eset i WAAL
Figure 3: Error vectors after Helmert transformation.
Figure 3 shows the nature of scanning errors obtained with
EPSON Expression 1640XL scanner. It can be concluded just
by visual inspection that there is a huge systematic influence,
which couldn't be modeled and removed Helmert
transformation. Also, obtained RMSE leads to the conclusion
