ul 2004 
——— 
quality 
2,3, 37- 
Quality, 
Landuse 
y means 
sis, ITU 
Terms, 
html, 
graphic 
Central 
tandard, 
rway. 
tabases : 
holl, M. 
incisco. 
ructureal 
jon of 
on for 
blication 
ormation 
ice and 
ACCURACY OF MEASUREMENTS MADE WITH A CYRAX 2500 LASER SCANNER 
AGAINST SURFACES OF KNOWN COLOUR 
J. Clark and S. Robson 
(jelark(@ge.ucl.ac.uk, srobson@ge.ucl.ac.uk) 
Department of Geomatic Engineering, 
University College London 
Gower Street, London WCIE 6BT 
KEY WORDS: Laser scanning, point cloud, accuracy testing, colour analysis, correction. 
ABSTRACT: 
Several commercial manufacturers produce laser scanning systems capable of measuring the surfaces of objects to precisions of the 
order of a few millimetres at ranges of between 2 and 200 meters. Experience on a number of projects has demonstrated significant 
variations in the quality of point cloud data as a function of object surface reflectivity. This paper investigates the performance of a 
Leica HDS2500 laser scanner in making measurements to a variety of surfaces of specified colour characteristics under laboratory 
conditions. From the data obtained it is evident that significant systematic range discrepancies exist which can be broadly correlated 
against the colour of each surface with respect to the wavelength of the laser used. Over the distances investigated a high correlation 
between the point data quality measure returned by the scanner and the range discrepancy was observed, offering the possibility of 
applying a correction to the data produced by a HDS2500 scanner. Such a correction, which is demonstrated to be applicable for 
diffusely reflecting surfaces, can be carried out with data available to the end user in order to significantly improve scanner accuracy. 
INTRODUCTION 
The Cyrax 2500 laser scanner, recently renamed the Leica 
HDS2500, uses the time of flight principal to measure range by 
observing the two-way travel time of a short pulse of laser light. 
The range to the surface is measured if a sufficient amount of 
pulse energy is reflected so that a signal is registered by the 
scanner's photodetector. The electrical signal generated by the 
detector must exceed a predefined threshold in order for the 
return time to be observed [3]. 
The proportion of the pulse reflected depends on the properties 
of the surface material and wavelength of the laser, which in the 
case of the 2500 system is 532nm. Lichti and Harvey [3] noted 
that the reflector-less nature of the system, and others of its 
type, raise the possibility of range errors due to pulse 
attenuation by the reflecting surface. Low reflectance results in 
poor quality data capture with some areas missing data 
completely or operating only effectively over a shorter range 
[1]. The system can operate over distances of up to 100 metres, 
although the recommended maximum range is 50 metres. Tests 
have shown that the system is not capable of measuring 
distances to materials with a low reflectivity beyond 50 metres 
[2]. In addition, Johansson [2] also detected some loss of points 
on these surfaces within the 50 metre range. In general it can be 
assumed that a material with a good diffuse reflectance, at the 
wavelength of the laser light being used, allows the capture of 
good quality data. 
For the 2500 system, Leica specify a single point positional 
accuracy of +6mm (16) over a range of 1.5m to 50m, a distance 
accuracy of +4mm (16) and an angular accuracy of +12”. The 
modelled surface precision is quoted as 42mm subject to 
modelling methodology. This specification can be accessed at 
Www.leica.com and has been reported on and evaluated by 
Tucker [6] and Boehler et al [7]. The work described in this 
paper will use these values to assess system performance. As a 
matter of note, the laser scanner used for this work had just 
Teturned from a factory service. 
Experiences at Plowman Craven and Associates (PCA) 
Experience with the 2500 system on a variety of projects has 
shown that scanning materials with different colours and 
textures produces point clouds of varying quality. Visual 
impressions, gained through modelling operations performed on 
the point cloud data, convey the impression that surfaces of 
lighter colour with higher reflectance provide the most 
favourable results in terms of high point density and minimal 
noise. Whilst this is to be expected, significant differences in 
measured range between surface types have also been observed. 
An example of this effect is the scanning of a large grey 
aeroplane with a dark stripe painted along the side. The 
resultant point cloud shows a false indent in the region of the 
stripe of the order of two to three millimetres deep. This 
suggests that the surface properties are affecting the laser’s 
return signal in some way and therefore the range recorded by 
the scanner. 
Overview of Experiments 
To investigate scanner measurement quality to different surface 
types, a selection of standardised colour patches and selected 
building materials were scanned at varying ranges and angles to 
the scanner. In each case the resultant point cloud was analysed 
to examine the effect of different colours and textures at each 
position. 
This paper reports on results obtained from a small portion of 
these laboratory tests namely those designed to investigate the 
quality of the point cloud with respect to surface colour. Key to 
repeatability and quantifiable nature of these tests are a set of 
standardised colour reference surfaces. Such a set were 
provided by a GretagMacbeth ColorChecker chart (Figure 1) 
which is manufactured to provide a standard set of colour 
patches with known spectral reflectance for the photographic 
industry. For the purposes of these tests the chart was adhered to 
a rigid planar surface to provide a planar object suitable for 
scanning. 
1031