Video analysis presents challenges and promise for analysts attempting to turn footage from countless hours of raw video feed into timely, actionable intelligence.
Since its inaugural data mining report, as required under the Federal Agency Data Mining Reporting Act of 2007, the Directorate of National Intelligence has detailed efforts to develop software capable of sorting and analyzing video data. The latest report describes an initiative known as the Automated Low-level Analysis and Description of Diverse Intelligence Video (ALADDIN Video). Its current effort was preceded by research termed Video Analysis and Content Extraction (VACE), which sought “to automate what is now a very tedious, generally human-powered process of reviewing video for content that is potentially of intelligence value.” The effort focused on developing technologies related to object and event detection, scene classification and content browsing, as well as the discovery of anomalous behavior – “for example, a person is observed entering a restricted area or leaves a bag in a public space.” VACE research concluded in September 2009 and was followed by work on the ALADDIN Video system.
The ALADDIN Video effort builds off the technologies explored under VACE. Like VACE, ALADDIN Video seeks “to enable an analyst to query large video data sets to quickly and reliably locate those video clips that show a specific type of event.” Finally, the Defense Advanced Research Projects Agency (DARPA) has worked to produce similar technological capabilities, though ostensibly, those efforts are aimed primarily at battlefield deployment. (Lest we forget, the modern-day internet was also the progeny of a DARPA brainchild, originally intended to facilitate military communications.)
Taken together, video analysis is clearly an emerging field with significant data mining potential. We also know of various efforts to root out terrorists using “malintent theory” – programs that attempt to identify threats based on telltale facial expressions, body movements or other behavioral indicators. The Department of Homeland Security is researching a system focused on providing “wide area surveillance and change detection capabilities to protect the Nation’s critical infrastructure,” and technologies that can identify individuals using iris or gait recognition are also in development. All these efforts, along with the general trend toward the increasing presence of closed-circuit cameras in public spaces, means more and more video footage, and theoretically a more comprehensive homeland security infrastructure. It also means more complex data sets that could potentially paint a more complete picture of a known or suspected terrorist threat, but which will inevitably require data mining systems to make sense of it all.
