Observer Detector Picks Out Crime Scene Regulars
posted by Paul Fiddian | 14.10.2011
US researchers are advancing new security technology that's based on the idea of criminals coming back to the scene of their crimes.
The criminal-returning philosophy's been seen often enough on TV shows but it's a staple of real world crime, too and it's been identified as a major issue in military conflict zones. So much so, that military officials have expressed a need for a system that can pick out insurgents responsible for building and/or deploying Improvised Explosive Devices (IEDs), knowing that they tend to be present at the point these devices explode.
That need's been addressed by the QuOD (Questionable Observer Detector), which works in quite a different way to more traditional facial recognition systems. While they compare fresh images to those contained in a historic database, in the search for potential matches, the QuOD creates so-called ‘face tracks' for all individuals that appear in the scenes it captures. These face tracks are constantly cross-referenced, meaning that if an individual's flagged up, it's because he/she has been face-tracked by the system before.
Questionable Observer Detector
From there, the Questionable Observer Detector creates a kind of individual representation portfolio, which can be used to determine if one particular person is appearing in a lot of settings. And if he/she is, is it to such an extent that it can be deemed suspicious? The Questionable Observer Detector allows scenarios of this kind to be reached and, potentially, for those with criminal intent to be identified.
That said, the QuOD system's not yet been perfected, with several issues yet to be overcome. Many of these are related to the subjects themselves which, for optimum image quality purposes, ideally need to be perfectly lit and looking straight at the camera.
Crime Scene Detector
Of course, that won't always happen in reality but that's an area that the crime scene detector system's developers, based at the Department of Computer Science and Engineering at the University of Notre Dame in Indiana, are looking into.
Other potential problem areas include that of data, amounts of which could stack up quickly and place a heavy burden on the computers trying to process.
The Questionable Observer Detector development programme is spearheaded by a three-man team made up of Notre Dame's Patrick Flynn, Jeremiah Barr and Kevin Bowyer. Security Technology will look again at their work in future News coverage.