Regine Brügelmann
AUTOMATIC BREAKLINE DETECTION FROM AIRBORNE LASER RANGE DATA
Regine BRÜGELMANN
Ministry of Transport, Public Works and Water Management, The Netherlands
Survey Department
Section of Remote Sensing and Photogrammetry
r.brugelmann @mdi.rws.minvenw.nl
Working Group III/5
KEY WORDS: DTM/DEM/DSM, Surface Reconstruction, Feature Extraction, Automation, Edge Extraction.
ABSTRACT
Airborne laserscanning is a relatively new and powerful technique for the acquisition of digital elevation models. Mostly
they are represented by regular grids. For some purposes, however, the description of the terrain with such a regular
grid is insufficient. Knowledge over the precise position of breaklines is required. Image processing algorithms can be
used for automatically deriving these breaklines from laserdata In this contribution, the suitability of a method for the
extraction of breaklines from dense laserdata has been investigated. It is based on hypothesis testing. All pixels with
a significant homogeneity measure with respect to noise are denoted as potential edge pixels. The quadratic variation,
used as homogeneity measure, indicates the extent of curvature. Hypothesis testing yields broad breakline-regions with
significant curvature compared to noise. These regions are reduced to one pixel wide breaklines by means of nonmaxima-
suppression taking into account maximal curvature direction. For the assessment of the performance of the algorithm
the automatically extracted breaklines are compared with photogrammetrically measured breaklines. It is shown that
automatic breakline extraction from airborne laserdata in principle is feasible.
1 INTRODUCTION
Airborne laserscanning is a relatively new and powerful technique for dense digital 3D-data acquisition. Beside applica-
tions such as derivation of building models, monitoring of power lines and extraction of forest stand attributes such as tree
heights, the main application of airborne laserscanning still is the determination of digital elevation models.
Compared with traditional methods such as photogrammetry and tachimetry, laser altimetry, with its increasing opera-
tionality, has become a cost and time effective alternative for the acquisition of digital elevation models. That's why this
new technique nowadays is used to build elevation models for even whole countries such as for The Netherlands. The
entire surface of The Netherlands is planned to be covered with laserdata with a density of one point per 16 m“ in 2000.
From the originally irregularly distributed measured points a regular grid is derived at the Survey Department of the Min-
istry of Transport, Public Works and Water Management and delivered to customers such as the local water management
organisations.
For some purposes, however, a regular grid on its own is insufficient for an accurate description of the terrain. This
applies e.g. to road planning and road design or to computations in the field of river management, e.g. the determination
of waterheights and streamvelocities in case of flooding. In these cases additional terrain shape information is required.
Breaklines provide this information. Describing the upper and lower edges of geomorphological features such as dikes
(see fig. 1), breaklines contain the most important morphological information of a digital elevation model. As 3D-vectors
they represent local maxima of surface curvature.
Furthermore, the use of breaklines as constraints for DEM calculation considerably contributes to morphological correct
DEM computation (Petzold et al., 1999). Apart from this, the automatic derivation of breakline information forms a crucial
contribution to the task of data reduction. The huge amount of data acquired by laserscanners can be expected to cause
difficulties in further processing, e.g. with computations concerning managing of whole rivers fluvial zones. This can
be overcome by the conversion of raster or triangle representation to vector structures which forms a sort of information
condensation (Weidner, 1994). The implicit, in a grid or TIN (triangulated irregular network) represented information, is
transformed into explicit surface descriptions embodied by vector structures such as break-, ridge- or valleylines.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 109