In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
484 
electoral or district division. At the moment there is no 
automatic conversion of addresses or ITN features into land 
uses or integration of derived information from remote sensing 
sources. There is a need for content specification of the data in 
terms of land cover and land use that is currently not in place. 
In order to have a clear internal structure of land cover and land 
use information there was therefore a need to develop an 
ontology that could establish links and associations. This 
ontology was developed in the Protégé environment. In the 
ontology a high level set of classes were created for both land 
cover and land use and relationships established with each 
other. It was important to clearly define the different nature of 
land use and land cover classes as they are normally mixed up 
in many of the existing classifications. For example, the NLUD 
land cover classification included classes such as agricultural 
land or features such as roads and pavements and the land use 
includes water and open land. Also CORINE land cover 
classification includes industrial, commercial and transport 
units as well as mine, dump and construction sites. So a clear 
definition of terms and relations is obviously needed. 
The integration of this ontology within the framework follows 
the suggestion of Fonseca et al (2000). They suggested an 
inheritance structure by which we define general classes that 
contain the structure of a generic type of object and then 
specialise these classes into subclasses. The subclasses would 
inherit all the properties of the super-class and add some more 
of its own. In this way the combined use of objects and 
ontologies could provide a rich model to represent geographic 
entities. It also allows us to provide a sensible relationship 
between land cover and land use in order to help define 
topographic objects. For example by knowing that the land 
cover for a “field” is “trees” and also knowing that the land use 
is “fruit production” then we can deduce that a more specific 
classification for the area of land is “orchard”. Similarly if we 
know that an object is classified as an orchard that it must have 
land cover of trees and its land use is “fruit production”. Also 
different users have different views on classes. With this 
framework they could use the main class as a common starting 
point and add their own subclasses that would inherit all the 
main properties of that class and server their particular purpose. 
Figure 1 reflects this principle. 
The link between the object oriented structure of Ordnance 
Survey data and a user classification is done through the 
intermediate ontology that can translate or incorporate new 
attributes or subclasses that then can be added to the existing 
features. Ultimately this framework would allow having 
automatic customised legends for bespoke maps. 
3. BOURNEMOUTH CASE STUDY 
3.1 Data and methods 
A preliminary trial was focused on the centre of the urban area 
of Bournemouth (south of England). The framework was used 
as guidance as there was still not an automatic transformation 
devised for the trial. The data used for both land cover and land 
use comprised: 
• MasterMap Topography layer 
• MasterMap Address Layer 2 
• Mastermap Integrated Transport Network (ITN) layer 
• Aerial imagery acquired in June 2006, 4 band spectral 
resolution and spatial resolution of 60 cm. 
The customised user classification included in this trial was the 
NLUD classification. 
3.1.1 Land Cover: For the Land Cover classification we 
used a combination of the layers described above. The ITN 
layer was used to identify artificial non-build up surfaces 
whereas the Topography layer did so for artificial build-up 
areas. For the classification of natural surfaces there was some 
information available on the Topography layer but the 
attribution was not rich enough to create a complete 
classification. So it was a classification from aerial photography 
that provided the most useful information. The classification 
consisted of a combination of texture, NDVI and height data 
that helped to identify high vegetation (trees) from low 
vegetation (shrubs and grass) and bare surfaces using a 
Maximum Likelihood classifier in Erdas Imagine. 
Object-Oriented Ontology Framework 
Figure 1: Object-Oriented Ontology Framework 
3.1.2 Land Use: For the Land Use classification Address 
layer 2 provided basic information about uses in different 
addresses. ITN layer provided information about transport uses. 
Topography layer and aerial photography were used to assess 
whether there were any morphological or visual signs that could 
help distinguish different areas within the urban space 
(residential, services, industrial, green open space, etc) 
3.2 Results 
The Bournemouth trial highlighted different issues for Land 
Cover and Land Use. 
A Land Cover classification that uses the general classes 
defined in the ontology could be quite straight forward using 
the data sources available. However this information could be 
insufficient depending on the level of granularity of other 
classifications. Also it has to be taken into account that the land 
cover classification of an urban environment is less complicated 
than that of a rural environment. The link between the ontology 
general classes and the NLUD classes worked well. However,