The pattern of veins on the back of someone鈥檚 hand is as unique as their fingerprints and can be used to identify people even with a cheap commercial camera. The technique could be used in smart door locks or even to identify people from CCTV images.
Syed Shah at the University of New South Wales in Australia and his colleagues used images from 35 volunteers, taking 500 photos of the back of one of their hands each time with a commercial Intel RealSense D415 camera. This device is sensitive to infrared light and can detect how far away something is with stereo lenses.
The range data was used to isolate the part of the image showing the back of the hand from the background. The team only used infrared information from the images because it can differentiate the blood in veins from body tissue.
Advertisement
The pattern of veins in each image was extracted and further processed for clarity, then used to train a neural network to connect a pattern to a particular subject. Shah used the resulting model to identify the test subjects with an accuracy of 99.8 per cent.
The researchers also tested the software鈥檚 ability to detect vein pattern that weren鈥檛 included in the original set. Four new subjects were identified as unknown with 96 per cent accuracy.
Shah says the technique is more robust to attacks than existing biometric tests. Fingerprints can be collected from smooth surfaces and used to trick sensors, while face recognition can sometimes be fooled with a photo. The team also found that vein detection works reliably on people of all ethnicities, which has been an issue for some biometric tests.
Shah believes that the technology can be adapted to run on smartphones. Apple鈥檚 iPhone 12 already has a built-in lidar scanner that can detect range information, and many cameras are also able to detect at least part of the infrared range of light. He also believes that it could eventually be made to work with CCTV cameras.
IET Biometrics
