Abstract
This paper reviews some of the history of automated visual surveillance, from the second and third generation VMD days of the early 1990s, to the current state of the art. It discusses the inherent limitations that resulted in an nearly negligible "increase" in performance throughout the 1990s and still exist in commercially available systems. Then we review an approach that overcomes these limitations - active visual surveillance with geo-spatial rules. Active visual surveillance uses data from computer controlled Pan/Tilt/Zoom (PTZ) units combined with state of the art video detection and tracking to, in a cost effective manner, provide active assessment of potential targets. This active assessment allows an increase in the number of pixels on target and provides a secondary viewpoint for data fusion, while still allowing coverage of a very large surveillance area. This active approach and multi-sensor fusion, not a new concept, was developed as part of the DARPA Video Surveillance and Monitoring (VSAM) program in the late 90's. While we have continued to expand upon it since that time, there has been no commercial video surveillance, before Guardian Solutions, that provided these important abilities. The core ideas in this paper address limitations of the original VSAM designs, briefly introducing our enhancements including geo-spatial rules for wide area multi-sensor fusion, and key design issues to allow us to support wireless networks.