Rank One launches periocular biometric recognition algorithm to identify people wearing masks

With facial recognition suddenly challenged by the popularity of masks, Rank One has released a next-generation periocular recognition algorithm to provide contactless biometric identification.
The ROC SDK Periocular algorithm performs person detection and identification using only images of the eye and eyebrow regions of the face with artificial intelligence and machine learning, according to the announcement. The algorithm works with any camera, and the company says it delivers the same best-in-class hardware efficiency as the ROC facial recognition SDK, with high accuracy.
The company points out that with masks occluding half of the face, the algorithms claimed to perform well when identifying people with masks have not been validated in NIST FRVT benchmarks, and are inconsistent with most previous algorithm design requirements.
The algorithm will be made available to all Rank One license-holders under active maintenance, and will include it in the standard ROC SDK at no extra cost. Special API features enable integrators to use both face and periocular algorithms together or separately in a stand-alone implementation.
By analyzing data only between the cheekbones and eyebrows, Rank One says the new algorithm can also provide an equal user experience for people wearing niqabs and other garments.
A blog post on the new technology provides some background on periocular biometric identification, and depicts the system being used to identify a person wearing a transparent face shield which occludes the lower portion of the face. Rank One also provides accuracy and efficiency details in the post, showing a false non-match rate (FNMR) of 0.015 at a false match rate (FMR) of 0.0001 with a frontal constrained image in company testing, compared to an FNMR of 0.004 for its full-face recognition in the same circumstances.
Rank One is also intending to issue a mask detection algorithm.
Article Topics
algorithms | artificial intelligence | biometric identification | biometrics | machine learning | mask detection | periocular biometrics | research and development | ROC
Comments