The day when access badges and PIN codes will be outdated may be coming. Facial Recognition Technology is a hot topic these days. It is popping up online, in newspapers, and on TV. All you’ll need to bring with you will be your face. And when that day comes your face may prove to be your best friend, or your worst enemy.
Trying to check your text messages? Can we see you first?
Recently, Google and Samsung introduced Galaxy Nexus, the first Smartphone to feature Android 4.0 Ice cream Sandwich platform. One of the innovations this new mobile OS offers is facial recognition. Upon setup, the user looks into the screen so the phone can scan and store her image. That’s all it takes. On subsequent uses, she only has to look into it and if the real time scan matches the stored image, access is approved and the phone unlocks, all in less time than it takes to read this sentence.
Are sure you’re not on the Most Wanted List? Let’s see
The FBI is preparing to inaugurate a nationwide facial recognition service by January 2012. Authorized personnel of participating states will be able to upload a photo of a suspect into the bureau’s biometric identification system. If there are possible matches, a file of mug shots ranked by similarity to the facial features of the person in the the photo will be sent to the inquiring agency.
Are you on the VIP list? Let’s take a look.
In reality, there will be no VIP list, just faces. Facial recognition verification procedures are already established at business such as banks, retail stores and places of amusement where registered customer enjoys VIP privileges. When one applies for membership, her/his face is scanned into the system and on subsequent visits a real time scan must match up with the stored image in order to gain access.
These practices all work will for one-on-one identity checks but what about time critical locations such as airports, train stations and sports arenas where steady streams of people pass through gates and checkpoints. Scanning faces one at a time would result in a huge bottleneck and a loud outcry. Kintronics is looking at solutions that would allow traveler’s scanned faces to be compared to images of terrorists on a watch list, gamblers visiting casinos, to known card counters, and shoppers at shopping malls, to convicted shoplifters, all without slowing the pace of foot traffic.
One such software product works with real time video. First off since this solution makes use of biometric analysis of facial features there are certain requisites that apply:
· IP Cameras should be mounted at eye level of a person of average height.
· The face should be visible in as close a frontal position as possible with a pan or tilt of no more than +/-15 .
· It is critical that the eyes be visible.
· Only minor parts of the face may be covered.
The video streams to a Tracking Station where it is analyzed for face visibility, position, and rotation angle, relying heavily on eye position. Next the face is cropped from the frame, scaled, and rotated to a size of 128x128 pixels. The extracted face is then time stamped and given a camera identifier before being sent on to the Watch List station.
Here it is compared to a data base of similarly configured images of persons on targeted watch lists of terrorists, kidnappers, fugitives. If a match is made, an alert is issued and sent to manned alert monitors dedicated to receiving and displaying just such alerts, allowing security officials to focus on the identified person.
Video content analysis like any analytic process is developed and tested in controlled cooperative parameters. Unfortunately video surveillance is neither controlled nor cooperative. Environmental problems abound. For example:
· Cameras mounted improperly, e.g. on a ceiling mount, will give a bad pose angle
· Bad illumination will result in indistinguishable texture information in the facial area
· Thick lenses or sunglasses impede proper face recognition
· An unstructured location offers suspects an opportunity to hide.
Some of these difficulties can be eliminated, the simplest and most direct, being camera placement. Since a frontal image offers the best chance of recognition cameras can be installed at the end of a long hallway approach or at the bottom of an escalator. One clever trick might be mounting a camera close to an information screen such as arrivals/departures in an airport. Illumination problems indoors can be minimized by adjusting the artificial lighting or using dim outs but outdoor cameras where the light changes as the day goes on proves more of a challenge. But if a person wants to take an indirect route behind pillars or crowds, hide behind dark or reflecting glasses, or cover his face with his hands there are no corrective measures and the system will be thwarted.
So, is this a replacement for trained officials manning video consoles? Absolutely not! While the software will not fall prey to attention lapses brought on by viewing simultaneously streaming screens of video all day, its pinpoint precision can be compromised. Can it improve a comprehensive IP based video management system? Yes, in situations and locales such as those cited above, a video content analysis tool such as facial recognition can prove to be a valuable second set of eyes.
