ISPRS Commission III, Vol.34, Part 3A ,,Photogrammetric Computer Vision", Graz, 2002 
  
GROUND SURFACE ESTIMATION FROM AIRBORNE LASER SCANNER DATA 
USING ACTIVE SHAPE MODELS 
M. Elmqvist 
Department of Laser Systems, Swedish Defence Research Agency,Linkóping 
KEY WORDS: Ground Surface Estimation, Laser Scanning, Digital Terrain Model, Active Shape Models, Deformable Models 
ABSTRACT 
Various filtering techniques exist to obtain the ground from laser radar data to use when building digital terrain models. This paper 
develops an active shape model approach to estimate the ground surface from laser radar data. The active shape model acts like a 
rubber cloth with elasticity and rigidity. With constraint forces the model is formed to an estimate of the ground surface. The model 
is glued against the measured points from underneath, forming the envelope of the point cloud. Even in a thick forest as much as 25 
per cent of the data points represent the ground. The stiffness of the shape model stretches it out to a continuous surface in between 
the ground points. The algorithm implemented in this paper is suited to use on very dense data sets, it has been designed for data sets 
of more than 10 points per square meter. We propose suggestions of changes to the algorithm to adjust it to work on more sparse data 
sets. 
1. INTRODUCTION 
At FOI at the Division of sensor technology there is a 
program for synthetic environments and sensor simulation 
(Ahlberg et al. 2001). The aim of the program is to develop 
methods for automatic construction of terrain models based 
on laser radar data. The cause is to meet the need for high 
resolution terrain data for mission planning, command and 
control and accurate sensor simulation in tactical simulations. 
Apart from the ground estimation algorithm developed in this 
paper, algorithms for single tree extraction (Persson, 2001) 
and algorithms for estimation of buildings have been 
developed within the program. 
Both airborne laser radars and ground based laser radars are 
used for the data gathering. The data used as input for the 
ground estimation algorithm is from an airborne system 
provided by TopEye AB. The system contains a vertical 
scanning direct detection laser radar operating at a 
wavelength of 1.06um. The pulse rate is between 2 and 7 
kHz and the emitted energy is about 0.1 mJ per pulse. The 
operational altitude is approximately 60-900m. The TopEye 
system is able to produce point position, intensity of 
reflection as well as multiple return or double echo data. The 
laser data used in our work was acquired at missions in 1998, 
1999 and 2000. We required dense data sets and hence the 
missions were flown at slow speed, i.e. 10-25 m/s, and at 
rather low altitudes, 120-375m. Some areas were also flown 
in two directions perpendicular to each other. The resulting 
data sets have a density that varies between 2 - 16 points per 
square meter. 
1.1 Related work 
At the Institute of photogrammetry and remote sensing at 
Vienna University of technology, a method based on iterative 
linear prediction has been developed and tested (Pfeifer et al. 
1998). The test has been performed on data from wooded 
areas. This method works by iteratively interpolating the 
A - 114 
ground surface. First, a rough surface approximation is 
computed using the same weight for all points. The obtained 
surface will run in an averaging way between ground points 
and non-ground points. Next residuals are computed for each 
point with respect to this surface. The majority of the ground 
points will get a negative residual, whereas the majority of 
non-ground points, e.g. vegetation, will get small negative or 
positive residuals. Finally, new weights are computed for 
each point using the residuals. To compute the new weights, 
a special adaptive weight function is used. In Figure 1 the 
principal form of this function is illustrated. 
  
v 
| 
0 
Figure 1. Weight function used to associate new 
weights to measured points using residuals. Points with 
large negative residuals get large weights and points 
with medium residuals will get smaller weights. Points 
with large positive residuals get no weight at all and 
hence become “eliminated” (Pfeifer et al. 1998). 
In each of the succeeding iterations, a new surface is 
interpolated using the original points and the previously 
derived weights. In this way, the iterative surface 
interpolation will converge towards a final solution. Break 
lines, like cliffs or edges of an embankment, always become