B3. Istanbul 2004
THE EVALUATION OF THE INTERNAL QUALITY OF LASER SCANNING STRIPS
USING THE INTERACTIVE ORIENTATION METHOD AND POINT CLOUDS
Petri Rónnholm ^ *
* Helsinki University of Technology, Institute of Photogrammetry and Remote Sensing, 02015 HUT, Finland -
petri.ronnholm@hut.fi
Commission III, WG III/3
KEY WORDS: Laser scanning, Adjustment, Quality, Inspection, Photogrammetry, Orientation, and Visualization
ABSTRACT:
The quality of laser scanning point clouds has become a topical research issue. The quality has been determined by the sum of
several error sources caused by various factors affecting accuracy. In this paper, it is proposed that overlapping laser strips are
favourable for inspecting the quality of the point clouds. The internal quality of five almost completely overlapping strips from
TopoSys Falcon was investigated using the interactive orientation method. The orientation was solved in several small test sites
located in different parts of the complete overlapping area. Each relative orientation between two laser point clouds revealed possible
height or planimetric shifts at the examination area. When this procedure was repeated in various locations within laser scanning
strips, internal deviations of laser data strips became visible. The comparison was done relatively. Therefore, no ground control
points were used. As a result, the repeatability in heights was excellent, whereas the planimetric repeatability, however, included
more systematic and non-systematic errors. Interestingly, the flight direction was the main error source, and visible in the observed
bias and random errors.
1. INTRODUCTION
The development of airborne laser scanning has been rapid
within last ten years. The LIDAR can capture 3-D point samples
from our environment. The strength of the LIDAR is good over-
sampling of the target — not necessarily the individual
measurements. The laser frequency and point density tend to
increase, when new generations of laser scanners are
introduced. In laser scanning technology, the focus area has
been the performance of laser scanning, including the
implementation of laser emitter and receiver, data handling and
direct orientation with GPS and INS.
Besides the technical development, another important issue is to
develop reliable and accurate methods to verify the quality of
laser scanning data. Several sub-factors can affect the quality.
According Baltsavias (1999), for example, time offsets, failed
system calibration, errors of GPS and INS, flying height, scan
angle, coordinate transformations, laser power, beam
divergence, atmospheric transmission, weather conditions,
target reflectivity, detector sensitivity and density of point cloud
can dilute the quality of the final laser point cloud.
Recently, promising results to inspect and improve the quality
of laser scanning data have been obtained using the adjustment
of overlapping strips (e.g. Kilian et al., 1996; Burman, 2000;
Crombaghs et al., 2000; Kager & Krauss, 1999; Maas, 2002).
The error sources of laser data due to flying or measurement
parameters, integration of the instruments, GPS, INS, laser
systems and processing errors has been reported by Schenk
(2001), Vosselmann (2002), and Burman (2000). Many of these
errors can be corrected using shift and drift parameters
(Burman, 2002). Ahokas et al. (2004) have studied the
repeatability of laser scanning strips, which is important to
verify in order to judge the usability of the data. However, the
focus of prior work in adjusting the overlapping strips has been
on fitting smooth surfaces.
The objective of this paper was to study the repeatability of
laser scanning strips, using the interactive orientation method
and five completely overlapping laser strips. The interactive
orientation method is based on visual interpretation of the data
obtained by superimposing 3-D laser point clouds on 2-D
images (Rónnholm et al, 2003). The method allows direct
relative orientation between laser scanning data and digital
images or between another laser point clouds. The strength of
the interactive orientation method over computational methods
is that the human intelligence can understand, interpret and fit
the entity quite easily even when working with difficult source
data, such as airborne laser scanning data. The advantage of the
proposed approach is that both the elevation and planimetric
errors can be defined and the complexity of the object studied
with overlapping strips can be high.
The comparison between laser scanning strips was done directly
in several small test sites of the entire overlapping strip area,
and without using ground control points. The orientation
method can be classified rather as an area-based matching than
to any point-to-point method, although sometime small details
of the tie features can be in key role for the orientation. The
original point clouds were used and no filtering or classification
is involved. The tie areas were inspected from several different
viewing angles — in central perspective. Therefore, both the
vertical and horizontal structure of the targets was available for
orientations. The interpretability was usually improved using
color-coding according the height value or distance from the
inspecting location.
One relative orientation. between two subsets from different
laser scanning strips reveals possible height or planimetric shifts
between data sets at the examination area. If this procedure is