International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
Fig. 1 GretagMacbeth ColorChecker chart mounted on a
vertical optical rail for testing purposes. Data obtained from the
bottom two rows of patches (primary and greyscale) are
reported on here
Since there is no universally defined standard reference colour
at which scanning reference distances are measured, for the
purpose of this work the Neutral 8 grey patch (second from
bottom left) was selected as the reference patch.
The key aims of the experiments described in this paper are as
follows:
e To investigate any variation in data distribution and noise
for each colour patch
e To investigate any variation in data distribution and noise
with rotation of each patch
e To investigate any changes in range for each colour patch
and to attempt to link these to the scanner output giving a
measure of the intensity of the return signal.
METHODOLOGY
Colour Checker Chart Scanning
The majority of scanning was carried out in the stable
environment provided by UCL’s instrument calibration
laboratory. A purpose built scanner mounting was manufactured
that allowed the scanner to be attached onto an optical rail at the
same height as the ColorChecker chart. Experiments were
divided into three groups, a close range set where chart rotation
and translation were carefully controlled, a near range set where
only translation was varied and a further range set scanned
outside the laboratory. The eight different distances employed
gave a measurement range of 4 to 9.5 metres in the laboratory
and up to 24 metres externally. In each case the chart was
scanned at a resolution of Imm by 1mm using the default
scanner settings for atmospheric correction.
Close range case
In order to allow repeatable rotation and translation of the chart
with respect to the laser scanner, the chart was securely clamped
to a small optical rail mounted on a motorised high accuracy
rotation stage (Newport RVI60HAT). The rotation stage was
located on top of a 600mm travel translation stage (Newport
IMS600CC) -which was in turn rigidly attached to an optical
table. The optical bench mounted laser scanner was aligned so
as to be orthogonal to the axis of the translation stage by auto
reflection of a laser beam off the scanner glass entrance
window.
The distances between the first three chart positions (one to
three) were recorded to better than 20 microns by the translation
stage unit. At these first three locations the chart was also
rotated by known angular increments using the rotary stage
(20°, 40° and 60°) and rescanned to investigate the effects of
incidence angle and colour on the measurement, Figure 2.
Translation stage, providing
chart locations 1, 2 and 3
Y
Rotation stage
GretagMacbeth
«4—— ColorChecker
Cyrax chart
2500
Laser Location of
Scanner, ColorChecker chart
rotated for longer
for longer à : à . range tests
range 5m Optical rail providing ™a
tests sa chart positions 4, 5 and 6 |
Fig. 2 Optical arrangement showing rotation of the
~ ~ ~ ~ 0,
GretagMacbeth ColorChecker chart from 0" (normal to
0 : -
the scanner) to 40 and 60°, and location for near range tests
Near range case
For the near ranges (four to six) laboratory space restrictions
dictated a different arrangement. In this case the colour chart
was located on a survey tripod and orientated orthogonal to the
optical rail on which the scanner was located.
Far range case
The far ranges (seven to eight) were scanned outside and
therefore were not conducted in such stable conditions as the
laboratory. However, it was considered valuable to scan at these
ranges for completion.
Data processing
Each scan of the colour checker chart was divided up into small
point sets cach of which occupied a single colour chart patch.
For the purposes of this paper, a section of the central area of
each greyscale and primary colour patch was manually selected
ensuring that all spurious points were eliminated, for example,
trailing edges along the boundary of the patch [7]. Point data for
each patch were exported as 3D coordinates together with the
scanner measurement quality number for each point in an
ASCII file format. The coordinate data were input into a least
squares shape fitting package [4] in order to independently fit
planes to the data and determine the range to each patch. In each
case the coordinate datum was preserved since the exported
coordinates of each point were defined with respect to the laser
scanner orientation.
To investigate the distribution of the data in each patch the
residual values, normal to each fitted plane, were used to create
a series of histograms the parameters of which would indicate
the quality of the data. This process was performed at each of
the eight ranges tested when the chart was orthogonal to the
scanner and at ranges one, two and three with the chart rotated
at known angular increments. This latter set of tests allowed à
check to be made on scan data acquired at different angles.
Statistical checking was performed using t-tests lo confirm
whether there was a significant difference between the range
from the scanner of an overall plane fitted to the entire chart and
the small planes fitted to cach colour patch. A series of f-tests
1032
ar
ar
CC
sc
pe
el
CC
ac
Li
Ri
pc
ex
th
gr
be
st:
of
fit
on
on
foi
of
To
bet
we
of
shc
gre
ren
edi
wa
sm:
wa
fitt
thai
Fig
dist
obs
whi
blac
Wh
