ıl 2004 
IDENTIFYING BUILDING TYPES AND BUILDING CLUSTERS USING 3D-LASER 
SCANNING AND GIS-DATA 
Hauke Neidhart, Monika Sester 
Institute of Cartography and Geoinformatics (IKG), University of Hannover, 
hauke.neidhart@ikg.uni-hannover.de 
KEY WORDS: Laser scanning, DEM/DTM, Building, Extraction, GIS, Classification, Combination 
ABSTRACT: 
Power authorities are highly interested in figures that indicate the energy requirements and especially the heat requirements on a 
local, regional and country-wide level. Such numbers are needed for their planning of new sites of power plants or for planning 
alternative energy modes. 
Existing methods for estimating those requirements heavily rely on local sampling methods as well as on the use of statistical 
estimates and models. The traditional way of acquiring area wide data is to use statistics and punctually acquired data and 
extrapolate it to wider areas. E.g. several districts of a city are investigated based on aerial photos and classified into different 
building and settlement typologies; the cities, in turn are classified according to certain types, which in the end will lead to a country 
wide statistics. 
In order to determine more accurate base information, in this project we are using laser scanning as a basic data acquisition method 
to determine building volumes, i.e. the volumes to be heated. This is due to the fact that laser scanning potentially allows for an area- 
wide data capture, and also has a high potential of automated data analysis and interpretation. The heat demand of an individual 
building depends primarily on its age and its type. Therefore, in order to assign head demands to individual buildings measured from 
laser scanning, the building type first has to be inferred from the available geometric characteristics. 
The paper will present the results of the automatic extraction of building volumes, and concentrates on the identification of the given 
building and settlement types that can be used to link the building volumes with specific heat coefficients. The results achieved with 
our approach will be compared with results derived in the traditional way. 
1. INTRODUCTION AND OVERVIEW seems not to be necessary, as we are interested mainly in the 
volume and not in the exact shape. Other geometrical features 
Our work is part of a project on pluralistic heat supply such as roof area and slope might be interesting at a later stage, 
(“Pluralistische Wärmeversorgung”) which is funded by the for example to derive assumptions about the year of 
AGFW (Arbeitsgemeinschaft Fernwärme, e.V.). AGFW is an construction. 
organization of energy and service providers which are engaged 
in local and district heating. DSMs from laser scanning are well suited to derive building 
volumes as they generally preserve jump edges quite well and 
One of the goals of this project is to detect locations where are easier to use in automated methods as compared e.g. to 
local and district heating can compete with traditional heating aerial images. There are many different algorithms to derive a 
by electricity, gas or oil. For that purpose, model calculations DTM from a given DSM. For example, Masaharu and Ohtsubo 
are performed which in turn need highly detailed information (2002) divide the area into small tiles and select the lowest 
on the heating demand. However, existing information is often points. In a further step, it is verified if these represent the 
out of date or not available on an area-wide basis. terrain. Then the initial DTM is created. At the moment this 
method can only be used in flat terrain. Briese et al. (2002) use 
Thus, we aim to derive this information from different data robust methods to classify the original points into terrain and 
sources. Since the heating demand of buildings is correlated off-terrain points. 
with the building volume, our first goal was to extract building 
volumes (Neidhart & Brenner, 2003). In a second step, these In order to obtain a precise definition of the terrain, additional 
can be combined with additional information such as specific data sources can be integrated. For our studies, we explored the 
heat coefficients which depend on the building type and year of use of ATKIS and ALK datasets. ATKIS is the Authoritative 
construction. Topographie Cartographic Information System in Germany 
(ATKIS, 2003). It contains information on settlements, roads, 
There has been a huge amount of research in the field of railways, vegetation, waterways, and more. However, for our 
automatic extraction and reconstruction of man-made objects, studies we only used the roads layer. ALK is the digital 
including buildings, see e.g. (Baltsavias et al. 2001). For cadastral map containing information on parcels and buildings 
example, Weidner (1997) uses laser scan data to extract (AdV, 2003). Again, we used only a small part of the available 
buildings. Using a segmentation of a normalized digital surface information, namely the ground plans of buildings. 
model (DSM) the locations of buildings are detected. From this, 
the ground plans are reconstructed. Brenner (2000) describes After the extraction of the building volumes from the laser 
the reconstruction of 3D-buildings from laser scan data and scanning data in the first step, the buildings have to be related 
ground plans, leading to detailed roof topologies. However, in to heat demand. Heat requirement not only depends on the 
our case a detailed reconstruction of the building’s geometry building volume alone, but also on other properties, especially 
715