3D images to help crack down on criminals

A recent breakthrough in face-recognition technology may save police officers hundreds of hours of painstaking work trying to identify culprits from video footage. 3D computer imaging is key to the new device. Scientists at the University of Hong Kong have developed new software that can reconstruct a person's full facial image from half-profile or profile images. The reconstructed facial image can then be searched in a database and identified by face recognition software with a high similarity score, regardless of different facial expressions. Dr. Chow Kam Pui, Associate Director of the Centre for Information Security and Cryptography at the Hong Kong University and one of the scientists working on the project, explains that the system can rotate a side-on facial image to face frontwards. It detects and extracts the features of a face from video footage and reconstructs the face on a three-dimensional model in computers. The reconstructed face can then be recognized by normal facial recognition systems. "Basically, the technology we use in this project is to actually take a snapshot of a photo from the face which probably won't be a frontal face image, and then we map it to our three-dimensional model. "After mapping into a 3-dimensional model, then we can freely rotate the face to any angles we like usually to a frontal position, so that it can be used for a normal facial recognition system." he says. Dr Chow Kam Pui adds that the accuracy of identifying a person by the old technology is often limited by face orientation. He believes that the new technology could improve this accuracy. "In some of our experiments, we find that when the face is rotated up to 80 degrees, in most of the commercial products we tested, the recognition rate is actually zero. Using our processing allows us to map the image to a 3-dimensional model and then rotate it to a frontal face. In the experiment we find our recognition is over 70 percent." he says. Ronald Chung, General Manager of CyberView Inc. Limited, a company which provides software to security outfits and one of the sponsors to the project, says he is very impressed by the research as the new technology bridges the gap between face recognition and a surveillance system. Chung adds that there are many useful applications for the new technology, including ATM machines. "There are plenty of application scenarios, for instance, for the ATM solution, well actually most of the ATM have installed with surveillance camera to capture the customer faces so that they can do some correlation to the customer and also for the forensic purposes and for investigative purpose. The face reconstruction technology can help the ATM camera to reconstruct a frontal face so that the face recognition technology can be applied more successfully in those application scenarios."—AP