Istanbul 2004
n Y, 2003.
aw LIDAR
rking Week
"eb. 2004).
J. and Clark-
ach to Thick
ion of Road
and Existing
ission3/wg3/
] Jamet; O.,
action. In:
ic Extraction
rial Imagery.
ote Sensing
action from
ator Fusion.
mote Sens-
ni/reini.html
I.-G., 2000.
ds from 1m-
; Southwest
jon, Austin,
ottensteiner,
canner Data
' Workshop,
Kubic, K.,
| and multi-
ydney, Aus-
ares: A Dif-
h European
. 630-641.
f Planimet-
"RSIS, Vol.
> Extraction
te Imagery.
age Project
)P0344678.
ich, Queens-
|
A QUALITY ASSESSMENT OF REPEATED AIRBORNE LASER SCANNER
OBSERVATIONS
E. Ahokas*, H. Kaartinen, J. Hyyppä
Finnish Geodetic Institute, Geodeetinrinne 2, 02430 Masala, Finland
Eero.Ahokas@fgi.fi, Harrı.Kaartinen@fgi.fi, Juha.Hyyppa(@fgi.fi
KEY WORDS: Comparison, LIDAR, DTM, accuracy, quality, planimetric error, repeatability
ABSTRACT:
This paper describes the height and planimetric errors of repeated ALS (airborne laser scanning) strips with a deeper focus on
building extraction. Measurements with Toposys Falcon airborne laser scanner were arranged in May 2003 in Espoo, Southern
Finland. A 5 km” test area, consisting of urban settlements and forests, were collected from the altitude of 400 m resulting in
measurement density of about 10 points per square metre. One 4 km long and about 100 m wide strip was collected five times
allowing the analysis of the repeatability of the laser scanning- One strip was used as a reference and inter-strip comparisons were
made. Point wise comparison methods were also used to characterize the differences. Additionally, target models were compared
against each other. Real Time Kinematic (RTK) GPS and also tachymeter measurements were used as ground reference. Extraction
of building vectors from laser scanner data was performed using interactive methods implemented in the TerraScan software. The
accuracy of the vectorization is also reported. Mean height errors for elevation points were —2 to 1 cm and standard deviations were
mainly £3-4 cm. In planimetry, mean errors of the centre points of the buildings were less than 30 cm for the first and also for the last
pulse data when compared with the buildings on the map. The standard deviations varied between «11-28 cm (first pulse) and +14-
18 cm (last pulse) for extracted buildings using repeated observations. Mean errors were between 3-8 cm and standard deviations £3-
6 cm using last pulse data of repeated observations and extracted ridge information. Extracted buildings were systematically larger
from first pulse data than from last pulse data.
1. INTRODUCTION
Airborne laser scanning (ALS) produces 3D information about
the object, giving both the terrain elevations and 3D target
models. The original output of the ALS is a point cloud
containing x, y and z coordinates and intensity values of the
points. The main applications of airborne laser scanning are
digital elevation models, but it can be used successfully for e.g.
3D city modelling, power line corridor mapping, forest
mapping, urban planning, water resource management and
railway surveying. More and more applications will appear as
the ALS equipment and analysis methods improve all the time.
Quality of airborne laser scanning has been studied for many
years (e.g. Crombaghs et al, 2002; Maas, 2002). Rónnholm
(2004) has studied the repeatability of laser scanning strips by
analyzing the shifts of local laser point clouds with the
interactive orientation method (Rónnholm et al., 2003). The
measurement accuracy of the laser points for different sensors is
reported in (Baltsavias, 1999). Building modelling accuracy in
position and in elevation has been reported by Steinle et al.,
(2000). Maas et al. (1999) used raw laser altimetry data and
two methods, the intersection of planar faces and the analysis of
invariant moments methods, to model buildings. The invariant
moments method yielded a precision of 0.1-0.2 m for the
building dimensions.
2. MATERIAL AND METHODS
2.1 Airborne laser scanner data
Toposys Falcon airborne laser scanner measurements were
carried out in May 14”, 2003 in Espoonlahti area, Southern
Finland. A 5 km? test area, consisting of urban settlements and
forests, were collected from the altitude of 400 m resulting in
nominal measurement density of about 10 points per square
metre. One 4 km long and about 100 m wide strip was collected
five times. First and last pulse data with intensity values were
recorded. Technical details about the Toposys Falcon laser
scanner system can be found in (www.toposys.com) and are
summarized in Table 1. A constant elevation adjustment for the
whole strip was done by Toposys. In their method the value of
the correction was determined at the overlaps of the flight strips
in the profiles that were running across the flight strips. The Z
correction was —0.03 m for strips ID number 2, 3 and 4 and —
0.01 m for strips 5 and 6. Strips 3 and 5 were flown to the
southeast direction and strips 2, 4 and 6 to the northwest
direction.
Sensor type Pulsed fibre scanner
Range 1600 m
Distance resolution 0.02 m
Scan width 14° (47°)
Scan rate 653 Hz
Laser pulse rate 83 000 Hz
Pulse length 5 ns
Laser wavelength 1560 nm
Data recording First and last echo, intensity
Table 1. Technical parameters of the Toposys Falcon airborne
laser scanner.