International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
have a distinct stepped appearance with the white points lying
well in front of the plane and the black points well behind the
plane. In the corrected point cloud this stepped pattern has been
flattened out.
Fig 11: Corrected colour chart points
In practice the range correction can be applied after the data has
been captured by the scanner, so that a correction for each
scanned point location can be computed. Ideally the correction
factor should be applied to the data within the scanner.
Further correction factor considerations
1. The correction factor was formulated for a limited
range (4 to 5.5 metres). It was noted during this
experiment that there appeared to be different
regression curves for different bands of ranges. This
assumption should be tested further.
2. Neutral 8 grey was selected as the optimum colour in
terms of reflectivity.
3. It is possible that this correction factor will be
different for individual scanners using similar range
measurement methods.
CONCLUSIONS
It is known from practical experience with the 2500 laser
scanner that materials of different colour and texture produce
point clouds of varying quality. At Plowman Craven and
Associates (PCA) the scanner is being applied to projects
demanding greater accuracy where errors imposed by different
materials may be significant. It is therefore important to
understand where possible erroneous points may be found in
order to be able to evaluate and make use of the resulting point
cloud in an effective manner.
Prior to these experiments it was assumed that point cloud data
representing different coloured surfaces would vary in density
and data quality according to the surface properties at the
scanner laser wavelength. This paper has demonstrated that,
over the limited range of test distances investigated, the noise
within each dataset to orthogonal surfaces appeared random and
largely insensitive in terms of distribution and data density to
variations in surface reflectivity. Significant differences were
however found when the planar colour chart was rotated by
angles of up to 60 degrees to the scanner. In such cases the
spread of the data distribution actually improved with scan
angle. Unfortunately at 60 degrees the density of the point data
began to noticeably decrease making angles of 40 degrees more
practical.
Experimental results demonstrate a systematic discrepancy in
range recorded by the scanner to different colour patches.
Results agree with the expectation that colours with a poor
reflectance at 532nm, for example black and red, reflect less of
the laser pulse causing the scanner to record a greater range than
the true position. Furthermore the observed difference in range
between colour patches closely correlates with the data point
measurement quality value recorded by the 2500 scanner. This
correlation appears stratified into groups of different ranges
which is likely to be related to the manufacturer's scanner
calibration methodology. A correction factor was formulated for
ranges of 4 to 5.5 metres and applied to selected colour patch
datasets to produce a set of corrected point locations. The
correction was found to move 3D data points derived from
different reflecting patches closer to physical location of the
complete colour chart thereby enhancing accuracy. The
correction did not improve the precision of the point cloud.
It is likely that all scanners which use a similar method of range
measurement will also exhibit variations in range when
scanning a variety of different colours. It may therefore be
preferable to incorporate an experiment testing the response of
the scanner to a variety of colours as a standard part of a
calibration procedure. A further paper is in progress which
describes similar experiments undertaken to analyse scanner
response to a selection of different material types at both close
and longer ranges.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the support of Plowman
Craven and Associates Ltd (PCA) and the UK Department of
Trade and Industry's Teaching Company Scheme without
which this work would not have been possible. Equipment
purchased as part of EPSRC strategic equipment funding
(GR/R06878/01) supported these tests.
REFERENCES
1. Barber, D. and Mills, J. 2001. Redefining the three R's —
Reflectance, Resolution and Reference: Important
considerations for Laser Mapping Systems. Surveying
World. May/ June. Pages 33-34.
Johansson, M. 2002. Explorations into the Behaviour of
Three Different High-Resolution Ground-Based Laser
Scanners in the Built Environment. Proceedings of the
CIPA WG 6. International Workshop on Scanning and
Cultural Heritage Recording. Sept 1-2, 2002. Corfu,
Greece. Pages 33-58.
Lichti, D and Harvey, B. 2002. The Effects of Reflecting
Surface Material Properties on Time-of-Flight Laser
Scanner Measurements. Symposium on Geospatial
Theory, Processing and Applications, Ottawa.
4. Fryer, J.G., Parberry, R.D. & Robson, S., 1992. Analysis
of as-built cylindrical shapes. Australian Journal of
Geodesy, Photogrammetry and Surveying 56:91-109.
n2
“I
5. www.brucelindbloom.com A selection of useful Colour
Calculators and Spreadsheets accessed July 2002.
6. Tucker, C. 2002. Testing and verification of the accuracy
of 3D laser scanning data. Symposium on Geospatial
Theory, Processing and Applications, Ottawa.
7. Boehler. W, Bordas Vicent, M. & Marbs A. 2003.
Investigating laser scanner accuracy. Presented paper XIX
CIPA Symposium at Antalya, Turkey, 30 Sep — 4 October
2003.
1036
ST
25
KEY
ABS”
Accu
data t
decis:
not. I
is es
perfo
techn
asses:
select
and c
synth
condi
prefe
The :
accur
geodz
Then
analy
asses:
refere
accur
assess
gener
and n
for ve
mean:
derive
and y
Can re
matri:
this p
simpl
Syster
and «
image
In this
invest
Synthe
appro;
Image
