REDUCING THE ERROR IN TERRESTRIAL LASER SCANNING BY OPTIMIZING THE
MEASUREMENT SET-UP
Sylvie Soudarissanane, Roderik Lindenbergh and Ben Gorte
Delft Institute of Earth Observation and Space Systems(DEOS)
Delft University of Technology
Kluyverweg 1, 2629 HS Delft, The Netherlands
(S.S.Soudarissanane, R.C.Lindenbergh, B.G.H.Gorte)@tudelft.nl
http://www.deos.tudelft.nl/
Commission WG V/3
KEY WORDS: Laser scanning, point cloud, error, noise level, accuracy, optimal stand-point
ABSTRACT:
High spatial resolution and fast capturing possibilities make 3D terrestrial laser scanners widely used in engineering applications and
cultural heritage recording. Phase based laser scanners can measure distances to object surfaces with a precision in the order of a few
millimeters at ranges between 1 and 80 m. However, the quality of a laser scanner end-product, like a 3D model, is influenced by
many different parameters, especially the relative object surface orientation and the local point cloud density. This paper introduces the
notion of point cloud quality. The obtained point cloud is first segmented using a planar feature extraction segmentation. Each segment
is subdivided into smaller patches of 20 x 20 cm. For each patch a patch quality parameter is determined, which incorporates the local
point density and local point quality. By averaging the patch quality over the complete point cloud, a point cloud quality is derived.
This paper demonstrates this approach in practice by comparing two scans of the same test room obtained from different stand-points.
As a result, it is shown and analyzed that by simply moving the scanner by two meters, the quality of the point cloud can be improved
by 25 %.
1 INTRODUCTION
The terrestrial laser scanning technology is increasingly being
used for representing and analyzing 3D objects in a wide range
of engineering applications. One of the main applications of the
terrestrial laser scanner is the visualization, modeling and moni
toring of man made structures like buildings. Especially survey
ing applications require on one hand a quickly obtainable, high
resolution point cloud but also need observations with a known
and well described quality. The phase based measurement tech
nique, where the phase of a multi-modulated wave determines the
distance to an object, is used in recent years mainly because of its
high speed. A complete quality description of individual scan
points is still under active research. However, major error com
ponents have already been identified. In this paper, it is shown
how an analysis of the individual point quality and local point
density can be exploited to improve the measurement set up.
A laser scan provides a spherical representation of the surround
ings with the center of the scanner as the origin of a local coordi
nate system. It uses the reflection of the laser beam on the object
surface to acquire a range measurement as well as an intensity
value of the reflected light. The accuracy of the range measure
ment is dependant on four main parameters:
• Scanner mechanism precision, e.g. mirror center offset, rota
tion mechanism abberations (Lichti, 2007, Li and Mitchell, 1995,
Zhuang and Roth, 1995).
• Properties of the scanned surface, e.g. roughness, reflectivity,
color (Bucksch et al., 2007, Kfemen et al., 2006, Clark and Rob
son, 2004).
• Conditions of the experiment environment, e.g. ambient light,
humidity, temperature (Pfeifer et al., 2007, Lichti and Gordon,
2004, Bòhler et al., 2003).
• Scanning geometry, e.g. incidence angle on the surface, range
differences (Bòhler et al., 2003, Cheok et al., 2002, Soudaris
sanane et al., 2007).
To obtain a 3D point cloud, the scene is scanned from different
positions around the considered object. The scanning geometry
plays an important role in the quality of the resulting point cloud.
Errors due to the scanning geometry are relatively well-described.
The ideal set-up for scanning a surface of an object is to position
the laser scanner in such a way that the laser beam is near perpen
dicular to the surface. Due to scanning conditions, such an ideal
set-up is in practice not possible. The different incidence angles
and ranges of the laser beam on the surface result in 3D points of
varying quality. Here we define the incidence angle as the angle
between that surface normal that is pointing in the surface, and
the incoming laser beam direction. The following two correlated
components variate with respect to the scanning geometry:
• Range quality. The study of the quality of range measurements
as a function of the scan angle has proven that in general the lower
the incidence angle and the lower the range, the higher the accu
racy of the range distance measurement. (Soudarissanane et al.,
2007)
• Point cloud density. The density of the point cloud decreases
with increasing incidence angles and range (Lindenbergh et al.,
2005).
Usually the scene is scanned from several positions around the
area of interest. The position or stand-point of the scanner that
gives the best accuracy, in terms of Least Mean Square Error (Te-
unissen, 1991), is generally not known. Using the optimal stand
point of the laser scanner on a scene will improve the quality of
individual point measurements and the overall local redundancy
of the measurements. This paper deals with the design of a mea
surement setup by showing how the stand-point of the laser scan
ner influences the point cloud quality.
2 METHODOLOGY
The quality analysis of the point cloud is described using the er
ror propagation techniques. The point cloud is first segmented