News

"Excuse me, mate, my phone thinks you're a villain."

Motorola thinks it has found a serious application for a mobile phone fitted with a camera - picking out criminals in a crowd.

The software, FaceIt, is known to some users of IBM ThinkPad notebooks as a way of ensuring that unauthorised people don't use their PC. It's been ported to Java, and will run on the new range of Motorola (and other) Java-enabled mobile phones, in a joint effort by Visionics, Wirehound, and Motorola.

At Sun's JavaOne conference in San Francisco last week, all three companies issued a press release saying that "Law enforcement officials may one day use their mobile phones to help identify criminals."

The application, developed specifically for a law enforcement agency, uses Visionics' FaceIt Argus as the delivery platform for facial recognition capabilities, linked to Wirehound's Birddog software on any Java 2-enabled mobile phone with a colour display.

It's not clear why the colour display is relevant, unless the real work of recognising faces is going to be done by a policeman on the beat, examining an image sent down from HQ by a central database.

"The FaceIt Argus system automatically finds faces in a field of view and searches them against a mug shot database," said the joint announcement. "Upon finding a match, the Birddog component generates a wireless alert to the phones used by mobile law enforcement officials, who are then able to verify the identity of the subject."

The phones can store multiple images and are alerted when a new image arrives. Non-matched images are automatically discarded from system, says the announcement.

This software is not brand new. It is more commonly found on PCs - it is bundled with clip-on digital web-cams sold for the IBM Thinkpad range. The cameras use the UltraPort extension to the USB which IBM builds into the lids of the Thinkpad PCs, and when invoked, work with a screen saver. Unless the right face appears in the field of view, the screen saver can't be ended.

It remains to be seen just how useful such a package can be to law enforcement in the real world.

It's quite feasible to design a system which recognises a static face about two feet away from the camera, and looking straight into the lens - using all the power of a PC processor. It's quite another to pick out faces which will be at varying distances, in various profiles, moving and animated, by pointing a phone-based webcam at a crowd.

The chance is that the phone won't be able to match the faces, or it would generate multiple false positive reports - not to mention the awkwardness of pointing a phone randomly at people in a crowd for a busy police officer. So what is probably happening, is that the software identifies patterns which look like faces, and transmits just those portions of the whole image back to HQ, where human beings will pick out likely suspects, and transmit them back to the phone display for the officer on patrol to check against the crowd. However, it does provide a nice illustration of what the Java platform is capable of doing, which was the point of the exercise, according to VNU Net