Automated casing event detection in persistent video surveillance

An increase volume of surveillance video is being collected, by various organizations, which has led to a need for automated video systems in order to reduce reviewing time. Using persistent video gathered from an aircraft overhead, as is done with unmanned aerial systems in Iraq and Afghanistan, we get a birds-eye view of vehicular activity. From these activities we can use a model to detect suspicious surveillance activity (casing). This paper builds a model to detect casing events and tests it using Global Positioning System (GPS) tracks generated from vehicles driving in an urban area to show the effectiveness of the model. The results show that several vehicles can be monitored at once in real-time. Additionally, the model detects when vehicles are casing buildings and which buildings they are targeting.