My primary interest in developing a query model for continuous data derived from the fact that there exist no pre-defined units in continuous data - neither for indexing, nor for querying. This is particularly true in the case of streams of video data. Automatic analysis techniques for video based on low level features such as colour, shape, texture and motion have been predominant issues in signal processing and computer vision communities. Several successful techniques have been developed for video data segmentation, summarization, indexing and retrieval based on visual contents. Automatic in-depth analysis of high-level semantic contents in video data, however, is beyond the scope of many of the techniques developed for low-level feature analysis. Consequently, these techniques have been proved insufficient for semantic retrieval of video data. In addition, due to the continuous nature of video data, traditional database query models are incapable of answering most interval queries. Semantic-based retrieval of video data, therefore, requires a new approach to query modelling.

One of my research goals has been to address variety of problems in semantic retrieval of video data, in particular on problems concerning the retrieval unit of video data. Our focus has been mainly on developing a rich query model for enabling users to specify several types of declarative interval queries against video data. Several interval queries are required to be considered and their semantics are required to be formally defined.