- This paper deals with semantic interpretation of instance [1]. We deal with a particular situation of the subway audiovisual data. It investigates how explicit interpretation of monitoring system in this paper and show how the selected uncertainty can help to bridge the semantic gap. We present a scenario can be modelled in the given framework. fcraasmesetwuodrykof-FuazpzypOlieWdLin-thaecucrornetnetxtunocfertthaeinrtyecreenptreEsuenrotapteiaonn To demonstrate the main features of the formalism, we project CARETAKER focusing on automated situation aware- defined the following task connected to subway monitoring: ness, diagnosis and decision support. A particular user-oriented There are 4 cameras installed in a station - 2 in the corridor task is modelled in FuzzyOWL. The paper briefly summarizes (different directions), and one for each platform. There is a basic features of the formalism, discusses pros and cons and microphone array in the main corridor. The cameras report points out difficulties and problems connected to its employment. an unusual (for non-peak hours) crowd of people in the main corridor. One of the cameras there shows that most of people Index Terms- uncertainty representation, FuzzyOWL, audio- are not standing in a reading distance from the travel info visual system. sign. Although quite a long time elapsed from the departure of the last train, there is a fuss (strong noise) detected by the I. INTRODUCTION microphones. The system should fire alarm for the operator.
Uncertainty representation in audiovisual domain is one
of the topics explored within a recent European project –
the Network of Excellence K-Space (Knowledge Space of
semantic inference for automatic annotation and retrieval of
multimedia content). We participate in the preparation of a
survey [
That is why we decided to demonstrate one of the formalisms – Fuzzy OWL – in real conditions and to show its features in a case study from a real project. The selected testbed is provided by the current European project CARETAKER (Content Analysis and REtrieval Technologies to Apply Knowledge Extraction to massive Recording) in which the first author participates. CARETAKER focuses on the extraction of a structured knowledge from large multimedia collections recorded over networks of camera and microphones deployed in real sites. The produced audio-visual streams, in addition to surveillance and safety issues, could represent a useful source of information if stored and automatically analyzed, in urban planning and resource optimization, environment planning and disabled/elderly person monitoring for
CARETAKER
As the name suggests, Fuzzy OWL combines OWL with
the fuzzy logic [
The second approach [
As the non-fuzzy T-box and R-box of ontology can be developed by standard techniques, we decided to use just the first mentioned approach for demonstration purposes (although the second approach is more expressive). Our use-case does not need the representation of uncertainty on T-box or R-box levels.
To evaluate Fuzzy OWL representation of uncertainty, we
developed simple domain ontology in Prote´ge´.2 It is designed
as a spatio-temporal ontology based on DOLCE [
The described ontology (see Figure 1) consists of a T-box terminology and a partial A-box containing information about “static” individuals like microphones or physical sectors of subway stations. The second part of the A-box is defined 1http://www.image.ece.ntua.gr/˜nsimou/ 2http://protege.stanford.edu as a result of real world analysis of sensory data. Resulting features contain typically vague or uncertain information. Microphone can report noise in the corridor at level 0.8 between No_noise and a chosen top level High_noise. One can append this information in Fuzzy OWL form to the Abox (as a fuzzy instance of the relation has_noise_level). Similarly, we could include fuzzy A-box axiom of relation is_crowded_in_time between a given sector and the current time instant which can be computed from the number of people in particular physical sectors.
Various aspects of Fuzzy OWL have been taken into account in our work. We paid attention especially to the complexity of reasoning, the way of dealing with imperfect knowledge, availability of tools for reasoning, the support for ontology development, the complexity of the modelling from the user point of view, and the possibility to present frequent patterns in an intuitive form.
In summary, reasoning in Fuzzy OWL is decidable. The formalism handles imperfect knowledge in both – ABox and TBox. There is a prototype of a reasoning tool for Fuzzy OWL available for download. It does not use the standard XML-based OWL syntax. Ontologies (T-boxes) can be developed using standard tools (e. g., Prote´ge´) if one decides to include fuzzy information on the A-box level only. A general conclusion of the work can be that the fuzzy approach is appropriate for the textual specification of the vague and possibly ambiguous description of the use-cases given in the natural language. On the other hand, it is rather tricky to model the dependencies expressed by conditional probabilities.
Our future work will focus on the development of an intuitive user environment enabling application-specific visualisation of the domain knowledge. Future directions of our research will also lead to a detailed specification of the background knowledge for the subway monitoring task. The scenario-based approach will allow us to create user-friendly interfaces, to enable end-users to introduce context information about a new scene, add new scenarios adapted to a specific environment, and define the specification compliant with the given ontology.
This work was supported by the European Commission under the 6th Framework Programme, projects CARETAKER (contract no.: FP6-027231) and K-Space (contract no.: FP6027026).