Generation of qualitative spatio-temporal representations from visual input

The simultaneous interpretation of object behaviour from real world image sequences is a highly desirable goal in machine vision. Although this is rather a sophisticated task, one method for reducing the complexity in stylized domains is to provide a context specific spatial model of that domain. Such a model of space is particularly useful when considering spatial event detection where the location of an object could indicated the behaviour of that object within the domain. To date, this approach has suffered the drawback of having to generate the spatial representation by hand for each new domain. An algorithm, complete with experimental results, is described for the automatic generation of a hierarchical region based on context specific model of space for strongly stylized domains from the observation of objects moving within that domain over extended periods.

The highest (hierarchical) level of region describes areas of behavioural significance or the paths followed by moving objects. An extension to the region generation algorithm allows these regions to tbe further sub-divided into equi-temporal regions (where it takes an object approximately the same time to traverse each sub-division) that can be used by an attention control mechanism to identify interacting objects.

By using a region based model, it becomes possible to convert the quantitative object locations into qualitative locations which then enables the use of the rich family of qualitative logics for real-world surveillance. To demonstrate the effectiveness of the spatio-temporal model combined with qualitative object representations, an event learning strategy is demonstrated that allows the automatic generation of contextually relevant event models, which are usually provided as part of the a priori system knowledge.