MPI-Video Surveillance and Monitoring:
Executive Summary

Video surveillance and monitoring systems (VSAM) must assimilate information from multiple sensors located on different platforms. This information must be combined with a variety of other, potentially disparate, information modalities. These include: terrain and other maps, information from various symbolic sources, a priori geometric models, human expert knowledge, spatio-temporal context, and user input. Uncertain and imperfect information and data must be fused to construct reliable conclusions about the dynamic entities of interest. In addition, the results of this data integration and analysis must be presented to a human user in an intuitive and efficient fashion.

In this project we are extending our active research in Multiple-Perspective Interactive Video (MPI-Video), and other areas, to develop technology crucial to VSAM data processing activities. We call this MPI-VSAM. Specifically, we have developed a novel architecture for the environment model (EM) in MPI-Video systems. This architecture integrates the information assimilation and data base functions of an MPI-Video system and encapsulates many algorithms for dynamic systems, including those useful for understanding algorithms (see our parking lot surveillance example). The EM represents a Gestalt of all the information obtained relating to the environment, and thus reflects a synergistic integration of a variety of information sources to construct a comprehensive, dynamically evolving model of the world. This model provides strong contextual guidance for the activity understanding algorithms integrated as part of this research. Some early results of systems using the novel architecture and data obtained from our new MPI-Video test bed are available. We feel that it is only through such integrated, contextually oriented information assimilation that truly revolutionary advances in image understanding are possible.

Web page design by Jeffrey E. Boyd