An algorithm to compute similarity between danger objects based on ontology for danger-aware systems

Context awareness is one of the most important aspects in the current ubiquitous computing research field. Danger-aware system is a kind of context awareness system, which detects the dangerous situations and provides services responsive to the dangerous situation. However, there is a shortcoming in the current danger/context awareness systems that the reasoning mechanisms are not flexible. Without enough predefined contexts, the system cannot work well, e.g., if a detected context is similar to a predefined context, but not exactly same as the predefined one, the system cannot recognize it. There is a similarity computing method based on ontology, which detects the dangerous objects similar to the predefined ones. However the method only considers the taxonomy similarity but does not consider the feature similarity. Without considering the feature similarity, the similarity will not so precise even some objects are in the same category, i.e. the taxonomy similarity is high, but exactly they are not so similar since they have different features. To solve the above problems, we propose a two ontology based similarity computing algorithm, which considers both the taxonomy similarity and the feature similarity. Firstly, two kinds of ontology are built, i.e. an object ontology and a dangerous property ontology. Then, the taxonomy similarity between two objects is computed based on the object ontology, and the feature similarity is computed based on the dangerous property ontology. The similarity between a detected object and a predefined object is calculated by combining the two similarities together. Finally, the degree of danger of the detected object is calculated based on the similarity, and the corresponding services are provided to users according to the degree of danger. The algorithm is implemented in an indoor ubiquitous test bed. And an experiment was performed to evaluate the system. Results of the experiment show the algorithm is effective.