Short description
(English)
|
The main objective of the proposed project is to develop methods for the automated enrichment of existing spatial databases in order to make it possible to map the urban ontological concepts developed in the course of COST Action 21to real-life urban spatial data. The proposed research is organized into five work packages (WP). WP 1 aims to develop, icollaboration with other projects of COST Action 21, an ontology of relevant urban concepts, including their scale-dependent representation. WP 2 will develop techniques for the automated recognition of urban patterns that represent simple object relationships as well as complex ontological concepts (e.g. city block, old town, residential neighborhood etc.) from spatial databases. WP 3 will deliver techniques for object matching between different levels of detail (LOD) of multi-resolution, multi-representation spatial databases (e.g. to associate a street of a cadastral database with a street in a coarser grained navigation database), and form ‘vertical’ object relations across scales. WP 4 will develop methods for data integration and propagation of semantics between different databases and object classes in order to form complex object relations. Finally, WP 5 is to furnish the enriched data to other COST 21 projects in order to evaluate their suitability to represent the ontological concepts developed in a variety of UCE applications in real-life spatial databases.
|
Partners and International Organizations
(English)
|
BE, CH, ES, FI, FR, GR, IT, NO, RO, UK
|
Abstract
(English)
|
Most of spatial databases that exist today have been designed to serve multiple purposes and hence concentrate on the 'least common denominator'. These general purpose spatial databases are rich in geometry, yet they are poor in semantics - in particular with regards to the representation of higher order semantic concepts that extend beyond the semantics of individual, discrete objects. While such higher level semantic concepts are not explicitly coded in current cartographic databases, they are nevertheless implicitly contained, owing to the fact that there often exists a relationship between the form (i.e. geometry) and function (i.e. semantics) of real-world phenomena, particularly in the built environment. Hence, it is possible - at least to some extent - to 'enrich' cartographic databases retrospectively, making implicitly contained higher level semantic concepts explicit through cartographic pattern recognition processes. The main goal of our project is therefore to develop automated methods to make this hidden information explicit. There are a number of solutions for the enrichment of cartographic/spatial databases, especially in the domain of automated cartographic generalisation. We argue, however, that the versatility and reusability of these solutions is often rather limited, since they were developed for specific databases and geospatial concepts, and encapsulated in algorithms. In our work, we have aimed to provide a more general approach by formalising the definition of semantic concepts through ontologies, and investigate how these formal definitions can be used to drive cartographic pattern recognition processes in order to enrich spatial databases. We argue that following this approach, enhanced understanding of generated patterns, easier adaptibility for different patterns, and enhanced interoperability can be provided. To this end, following issues have been adressed in our research: 1) Identification and formalisation of relevant urban concepts and their spatial properties. 2) Transformation from ontologies to algorithms that allow their automatic detection in existing spatial databases. 3) Design of intuitive human-computer interaction methods with the pattern recognition system: How can a human operator define concepts and how can he/she explore generated patterns/relations? 4) Evaluation of the enriched database, in order to demonstrate the utility of ontology-enriched databases. Objective 1 has been addressed by extracting knowledge from various sources about urban morphology, urban design, and city guides, and using this knowledge to define ontologies. Concerning objective 2, a methodology and framework for ontology-driven pattern recognition has been developed and published. It builds on a formalisation of the pattern recognition process by relating geographic concepts to cartographic measures and to other geographic concepts. The information provided by the formalisation process can then be used to carry out the actual spatial data enrichment process through logic deduction. Several approaches for conducting reasoning, considering uncertainties inherent in geographical phenomena, were considered: Deduction by description logic, Fuzzy Logic, Bayesian reasoning, and Support Vector Machines. These approaches have been evaluated with respect to fitness for use (i.e. reasoning power), and complexity of parameterising the pattern enrichment process for the domain expert. In conclusion, we propose translating the ontologies into Bayesian networks for carrying out the data enrichment process. Concerning objective 3, a simple control language has been developed that allows domain experts to define their own geographic concepts, yet is sufficiently formal to facilitate automatic translation to pattern recognition algorithms. Research on the subject will be pursued beyond the end of the SER funding period, with additional funding from a national mapping agency (NMA) in a major European country. Future research is devoted to firstly improving the current pattern recognition prototype. It will be transformed to a service-based framework, such that it can be more easily extended by more cartographic measures. Secondly, we wish to integrate more geographic concepts. And finally, the liaison to an NMA, representing end users, will enable us to perform more in-depth work on objective 4.
|
