lap
XI,
ino
ind
George Vosselman
SLOPE BASED FILTERING OF LASER ALTIMETRY DATA
George VOSSELMAN
Department of Geodesy, Faculty of Civil Engineering and Geosciences
Delft University of Technology, The Netherlands
g.vosselman € geo.tudelft.nl
Working Group III/2
KEY WORDS: Laser, DEM, Filtering
ABSTRACT
Laser altimetry is becoming the prime method for large scale acquisition of height data. Although laser altimetry is full
integrated into processes for the production of digital elevation models in different countries, the derivation of DEM's
from the raw laser altimetry measurements still causes problems. In particular the laser pulses reflected on the ground
surface need to be distinguished from those reflecting on buildings and vegetation. In this paper a new method is
proposed for filtering laser data. This method is closely related to the erosion operator used for mathematical grey scale
morphology. Based on height differences in a representative training dataset, filter functions are derived that either
preserve important terrain characteristics or minimise the number of classification errors. In experiments it is shown that
the latter filter causes smaller errors in the resulting digital elevation models. In general the performance of the filters
deteriorates with a decreasing point density.
1 INTRODUCTION
Laser altimetry is becoming the prime method for large scale acquisition of height data. Several countries are currently
using laser altimetry for creating or updating very detailed regional or nation-wide digital elevation models. In the
Netherlands, for example, a nation-wide digital elevation model is being created with a density of 1 point per 16 m2
[Huising and Gomes Pereira, 1998]. Even though production lines are set up, the derivation of digital elevation models
from the measurements by airborne laser scanners is not without problems. The two major problems are the elimination
of systematic errors and the selection of ground points [Huising and Gomes Pereira, 1998].
The presence of errors in laser altimetry data often becomes evident when combining data from adjacent strips. In the
overlap between the strips systematic differences between the heights of the points and the location of height jumps can
be observed. A topic of research is to model these errors and to eliminate them in a so-called strip adjustment, much like
eliminating lens distortion in a self-calibrating block adjustment [Kilian et al., 1996].
The other problem is the selection of the ground points. The pulses emitted by the laser scanner can be reflected by
buildings, trees, cars, electricity wires, and many other objects on top of or above the ground surface. The height of
these objects should not be included in the conventional digital elevation models. Therefore, filter methods are used to
discriminate the ground points from other points. In this paper, a new approach to filtering is presented.
Lindenberger [1993] describes how mathematical morphology can be used for filtering data recorded by a laser profiler.
A first estimation for the ground surface is obtained by an opening on the recorded data with a horizontal structure
element. All points within some distance of the estimated ground surface are classified as ground points. An auto-
regressive process is used to improve the results obtained by the opening. The application of the auto-regressive process
requires an ordering of the laser points. Therefore it is suited for processing data obtained by laser profilers. Data
obtained by laser scanners is, however, scattered in the XY-plane, for which there is no logical one-dimensional
ordering. In contrast to the auto-regressive process, the opening and selection of points within some distance of the
resulting surface can easily be extended to two-dimensionally scattered data.
Kilian et al. [1996] note that the size of the structure element used for the opening is a critical parameter for which there
is no single optimal value. They suggest to use a series of openings with different structure element sizes. For each
point, the maximum size at which this point is within some distance of the opened surface is assigned as a weight to this
point. These weights are used in a final smoothing step to estimate the ground surface.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 935