Istanbul 2004
A ROBUST METHOD USED WITH ORTHOGONAL POLYNOMIALS AND ROAD
NETWORK FOR AUTOMATIC TERRAIN SURFACE EXTRACTION FROM LIDAR
DATA IN URBAN AREAS
Nizar Abo Akel* *, Ofer Zilberstein" , Yerach Doytsher?
* Department of Transportation and Geo-Information Engineering
Faculty of Civil and Environmental Engineering,
Technion — Israel institute of technology
Technion City, Haifa 32000, Israel
(anizar, doytsher)@tx.technion.ac.il
? Telecom & Systems Group, Ness A.T. Ltd, Tel-Aviv 61581, Israel
oferzilberstein@hotmail.com
Commission WG_III_3
KEY WORDS: LIDAR, DTM, ALGORITHMS, EXTRACTION, METHOD, DATA, URBAN, DSM.
ABSTRACT:
The data provided by LIDAR systems are presented in a dense, accurate three-dimensional form without point classification, such as
buildings, trees roads, terrain, etc. The principal objective of this research is automatic terrain extraction, without prior knowledge of
the DTM. The key issue is first extracting the road network, and then use it as input information to the reconstruction process of the
DTM. This approach is naturally applicable to urban areas.
Various methods have been developed for terrain extraction. Most frequently used is the robust method. The essence of the
robust method is computing the height of measured points by means of an interpolation function based on neighbouring points, and
comparing the resulting height to the measured height. Points describing buildings will be characterized by a large positive
difference versus surface points characterized either by a large negative difference or a small positive difference. To compute the
new height for a certain point, a bi-dimensional interpolation function is used, whose coefficients are extracted by adjustment
process in which all points within a certain computed radios around the point take part. For a function of higher order, a singularity
in the coefficient matrix is possible with the resulting solution (if any) is unstable and influenced by measurement errors. Moreover,
the robust method has been developed specifically for forested areas; its application to urban areas leads to erroneous classification
of buildings as land and vice verse, due to the large area of buildings, particularly with complex shapes.
The objective of the present research is to improve the current method, in order to obtain better results. To achieve it, we are
using orthogonal polynomials, which permit the usage of interpolation functions of a polynomial kind, without restricting the
polynomial degree. An attempt was made to develop a new method base on the roads net, for computing an approximate model of
the terrain. For improving the extraction process, the Roads method, and the Robust method have been merged. The algorithm for
terrain extraction has been applied to four urban areas which include various kinds of problematic objects such as bridges and large
buildings containing open roofs etc. Based on the qualitative aspect, the results are promising.
1. Introduction
LASER SCANNING
LIDAR has become a reliable technique for data collection
from the earth surface. The LIDAR system integrates three
basic data collection tools: a laser scanner, Global Position
System (GPS), and an Inertial Measuring Unit (IMU) (see
figure 1.1). The laser scanner sends pulses toward the round, Lm Es
which are then reflected back to the sensor after hitting an t ;
object. The time it takes for the pulse to travel from the sensor LÉ oe
to the object and back to the sensor, together with the position : wr
and the orientation of the sensor, makes it possible to calculate
the position and the elevation of the ground points. TTT
Using the LIDAR technology we obtain dense (0.5-3 point i SR
per meter) and accurate (5-20 cm) Digital Surface Model
(DSM), but with no point classification that describe various
objects, such as buildings, roads, trees, terrain, etc.(see figure The most frequently used is the Robust method (Kraus and
1.2). Pfeifer 2001, Briese et al. 2002).
Varies algorithms for automatically extraction DTM is Moreover, the Robust method has been developed
available in the literature, Masaharu & Ohtsubo (2002) specifically for forest area; its application to urban areas causes
OBL
Figure 1.1 - Laser system
extracted DTM from complex terrain by dividing the area into
small tiles and selected the lowest point to construct the initial
approximation for the DTM (the method is applicable for flat
areas only).
mistaken classification of buildings as land and vice verse, due
to the large area of buildings, in particular if their shape is
complex.
