1 2004 
SS 
n 
S 
DISCREPANCIES BETWEEN OVERLAPPING LASER SCANNER STRIPS - 
SIMULTANEOUS FITTING OF AERIAL LASER SCANNER STRIPS 
H. Kager 
Vienna University of Technology,Institute of Photogrammetry and Remote Sensing, 
Gusshausstr. 27-29 / 122, A-1040 Wien, hk@ipf.tuwien.ac.at 
Commission II, WG II[/1 
KEY WORDS: Adjustment, Calibration, Orientation, DEM/DTM, Laser scanning, Gps/Ins, Georeferencing, Reliability 
ABSTRACT: 
This paper deals about discrepancies between overlapping laser scanner strips. We assume these discrepancies stemming from non- 
sufficient system calibration. These discrepancies - in height as well as in planimetry - are unsatisfactory phenomena for end-users 
of the ground data. Nevertheless, these gaps can be eliminated to a great portion doing a simultaneous 3D adjustment by least 
squares. An adjustment strategy is proposed for doing that: correcting those exterior orientation elements as recorded by dGPS and 
IMU, as well as interior orientation elements concerning the Scanner-dGPS-IMU system. The method (functional model) chosen is 
to apply correction polynomials in the time domain to all degrees of freedom as determined by the dGPS-IMU components and to 
the relative orientation parameters between those scanner-system components. All these parameters may be chosen block- or strip- 
variant and are determined simultaneously with hybrid adjustment by least squares. "Preventive regularisation" is used to catch un- 
or weakly determinable parameters. Automated determination (measurement) of tie features (instead of tie points) is described. 
Since in the point-clouds no corresponding points can be found, tying features - as planes and straight lines as their intersection - are 
used. Noise (e.g.) from the cover of vegetation has to be considered in this context; so, homologous point-clusters with low noise 
and few exceptions with respect to an adjusting plane have to be searched for. Speaking in the terms of standard photogrammetry, 
"homologous planes" replace "homologous points" as tying features in block adjustment of strips as unit; "strips" replace "photos" or 
"models". Nevertheless, an originally photogrammetric adjustment programme could be successfully extended to perform the task. 
Once, this programme had also been extended to handle scanner images, introducing time dependent parameters. The history of 
evolution of the mathematical model reveals the strong relationship between laser scanning and photogrammetry and geodesy. The 
distribution of control features (instead of control points) is discussed. Colour-coded difference-DEMs are used to judge the 
improvement of interior and exterior orientation. 
1. INTRODUCTION 
For transforming laser scanner strips into the national ground- 
1.1 General 
Laser scanners are mounted in aircrafts for collecting 3D-data 
of the surface of the earth. Proceeding the flight path, the laser 
beam sent downwards is deflected rhythmically aside and scans 
the ground surface in a meandric or parallel pattern with a high 
pulse rate. Most such devices use the technique of run-time 
measurement: the distance to a ground point then is a function 
of the time gap between the pulse was sent and received. 
The direction of the laser beam is given by some deflecting 
device like a rotating or oscillating mirror and some trigger 
causing discrete pulses. So, the device records polar co- 
ordinates of ground points in its own local co-ordinate system. 
The origin of this device co-ordinate system follows the flight 
path and its movement can be measured with dGPS (differential 
Global Positioning System) very precisely using the phase 
comparison method. Since coupled to the aircraft, the attitude 
of the device changes also during the flight and can be recorded 
with INS (Inertial Navigation System) — more exactly spoken, 
with an IMU (Inertial Measurement Unit). 
The components GPS, IMU and laser scanner have to be 
synchronised; moreover, their relative - but constant - 
displacements have to be determined (calibration of 
eccentricities). 
555 
survey co-ordinate system using dGPS and INS, we principally 
need only one ground reference station with known ground- 
survey coordinates. Moreover, we need also the form of the 
geoid. But, in practice, we should not be satisfied with that 
minimal solution because: 
e The form of the geoid is not sufficiently ( up to some few 
cm ) known in many regions. 
e The on-the-fly-initialisation for solving the GPS phase 
ambiguities nowadays is possible for fast moving objects 
like aircrafts with a r.m.s.e. of about 10cm; this might 
result in errors of some dm. Usually, neighbouring 
precision of dGPS is better by one order of magnitude. The 
errors increase with the strip length. (Cramer, 2000) 
e The attitudes as delivered from IMUs in use are prone to 
errors of about 0.01gon resulting in 16cm on the ground 
assuming 1000m relative flying height. Errors of IMU 
attitude also introduce some torsion of the laser scanner 
strips inducing errors in ground coordinates. Alike, IMU 
attitudes have a high neighbouring precision based on the 
gyros used; nevertheless, they show drifting phenomena. 
The resulting error effects might reach again some dm in 
the positions of ground points. (Cramer, 2000) 
e System failure or system instabilities shall be mentioned 
also: e.g. the change of the set of available GPS satellites