The rapid proliferation of surveillance cameras makes it increasingly difficult to have humans assist in monitoring. Reliable methods are needed for autonomous video analysis.
One of the most difficult goals of video analytics is to detect abnormalities or events that differ from what is considered usual, such as an abandoned package, a car traveling against traffic, or a fallen elderly person. While it is now possible to identify a simple abnormality using motion detection, that technology does not work in more complex scenarios, such as a car traveling against dense traffic.
Pierre-Marc Jodoin at the University of Sherbrooke (Sherbrooke, QC, Canada) and Janusz Konrad and Venkatesh Saligrama at Boston University (Boston, MA, USA) have developed a simple, memory-light approach. They are exploring implementing behavior subtraction in embedded architectures used in IP surveillance cameras. This would permit edge-based processing to reduce data flow in the network by communicating frames with unusual content only. They are also working on extending their method to multicamera configurations.