Multi-target camera tracking, hand-off and display LDRD 158819 final report
Modern security control rooms gather video and sensor feeds from tens to hundreds of cameras. Advanced camera analytics can detect motion from individual video streams and convert unexpected motion into alarms, but the interpretation of these alarms depends heavily upon human operators. Unfortunately, these operators can be overwhelmed when a large number of events happen simultaneously, or lulled into complacency due to frequent false alarms. This LDRD project has focused on improving video
... improving video surveillance based security systems by changing the fundamental focus from the cameras to the targets being tracked. If properly integrated, more cameras shouldn't lead to more alarms, more monitors, more operators, and increased response latency but instead should lead to better information and more rapid response times. For the course of the LDRD we have been developing algorithms that takes live video imagery from multiple video cameras, identifies individual moving targets from the background imagery, and then displays the results in a single 3D interactive video. In this document we summarize the work in developing this multi-camera, multi-target system, including lessons learned, tools developed, technologies explored, and a description of currently capability.