In: Stilla U, Rottensteiner F, Paparoditis N (Eds) CMRT09. IAPRS, Vol. XXXVIII, Part 3/W4 — Paris, France, 3-4 September, 2009
DETECTION OF BUILDINGS AT AIRPORT SITES
USING IMAGES & LIDAR DATA
AND A COMBINATION OF VARIOUS METHODS
Demir, N. 1 , Poli, D. 2 , Baltsavias, E. 1
1 - (nusret,manos@geod.baug.ethz.ch)
Institute of Geodesy and Photogrammetry, ETH Zurich, CH-8093, Zurich, Switzerland
2- (daniela.poli@jrc.ec.europa.eu)
European Commission - Joint Research Center, Ispra (VA), Italy
KEY WORDS: DTMs/DSMs, Lidar Data Processing, Multispectral Classification, Image Matching, Information Fusion, Object
Detection, Buildings
ABSTRACT:
In this work, we focus on the detection of buildings, by combining information from aerial images and Lidar data. We applied four
different methods on a dataset located at Zurich Airport, Switzerland. The first method is based on DSM/DTM comparison in
combination with NDVI analysis (Method 1). The second one is a supervised multispectral classification refined with a normalized
DSM (Method 2). The third approach uses voids in Lidar DTM and NDVI classification (Method 3), while the last method is based
on the analysis of the density of the raw Lidar DTM and DSM data (Method 4). An improvement has been achieved by fusing the
results of the different methods, taking into account their advantages and disadvantages. Edge information from images has also
been used for quality improvement of the detected buildings. The accuracy of the building detection was evaluated by comparing the
results with reference data, resulting in 94% detection and 7% omission errors for the building area.
1. INTRODUCTION
In this work, we focus on the building detection for airport sites.
The acquisition of a reliable geospatial reference database of
airports, and in particular the automatic extraction of buildings
and obstacles at airports, both have a critical role for aviation
safety. Often, 3D information of airports is not available, not
accurate enough, not complete, or not updated. Thus, methods
are needed for generation of accurate and complete 3D geodata
with high degree of automation. In particular, buildings and
trees are considered as obstacles, so they should be correctly
extracted. In this work, we focus on the detection of buildings,
as a first step for their 3D extraction. There are several methods
applied for this purpose, based on image and/or airborne Lidar
data. In our approach, buildings are detected in aerial images
and Lidar data through multiple methods using multispectral
image classification, DSM (Digital Surface Model) and DTM
(Digital Terrain Model) comparisons and density analysis of the
raw Lidar point cloud. The detection quality is improved by a
combination of the results of the individual methods. This paper
will give a brief overview of the related work on this subject.
Then, after the description of the test area at Zurich Airport,
Switzerland, the strategy and methodology will be presented
and the results will be reported, compared and commented. This
work is a part of the EU 6 th Framework project PEGASE
(Pegase, 2009).
2. PREVIOUS WORK
Aerial images and Lidar data are common sources for object
extraction. In digital photogrammetry, features of objects are
extracted using 3D information from image matching or
DSM/DTM data, spectral, textural and other information
sources. Pixel-based classification methods, either supervised or
unsupervised, are mostly used for land-cover and man-made
structure detections. For the classical methods e.g. minimum-
distance, parallelepiped and maximum likelihood, detailed
information can be found in (Lillesand and Kiefer, 1994).
In general, the major difficulty in using aerial images is the
complexity and variability of objects and their form, especially
in suburban and densely populated urban regions (Weidner and
Foerstner, 1995).
Regarding Lidar, building and tree extraction is basically a
filtering problem in the DSM (raw or interpolated) data. Some
algorithms use raw data (Sohn and Dowman, 2002; Roggero,
2001; Axelsson, 2001; Vosselman and Maas, 2001; Sithole,
2001; Pfeifer et al., 1998), while others use interpolated data
(Elmqvist et ah, 2001; Brovelli et ah, 2002; Wack and Wimmer,
2002). The use of raw or interpolated data can influence the
performance of the filtering. The algorithms differ also in the
number of points they use at a time. In addition, every filter
makes an assumption about the structure of bare-earth points in
a local neighbourhood. This assumption forms the concept of
the filter (Sithole and Vosselman, 2003). The region-based
methods use mostly segmentation techniques, like in Brovelli et
ah (2002), or using Hough transformation (Tarsha-Kurdi et ah,
2007). Some researchers use 2D maps as prior information for
building extraction (Brenner, 2000; Haala and Brenner., 1999;
Durupt and Taillandier, 2006; Schwalbe et ah, 2005).
Topographic maps provide outlines, classified polygons and
topologic and 2D semantic information (Elberink and
Vosselman, 2006).
In general, in order to overcome the limitations of image-based
and Lidar-based techniques, it is of advantage to use a
combination of these techniques. Sohn and Dowman (2007)
used IKONOS images to find building regions before extracting
them from Lidar data. Straub (2004) combines information
from infrared imagery and Lidar data to extract trees.
Rottensteiner et ah (2005) evaluate a method for building
detection by the Dempster-Shafer fusion of Lidar data and
multispectral images. They improved the overall correctness of
the results by fusing Lidar data with multispectral images.
