Generating Dynamic Scene Descriptions
This chapter addresses the problem of extracting descriptions of object behavior from image sequences. Vision systems are now capable of delivering trajectory-based descriptions of moving objects in a scene but little work has been done on the spatiotemporal reasoning needed for the computation of behavioral descriptions. This level of understanding allows us to compute meaningful descriptions of what is happening in a scene. We have developed analogical, cellular representations of space and time together with deictic descriptions of the behavior of moving objects to support the generation of this kind of dynamic scene description. In addition, we propose that an active, purposive framework is required for advanced vision as we need to be selective and deliver a situated (here and now) analysis of behavior in the scene. The problem of dynamic surveillance of traffic scenes is used for illustration.