Ambarella, Lumentum detail why 3D sensing is best for contactless biometric access control
Ambarella, Lumentum, and ON Semiconductor have offered up details about the 3D sensing technologies that will enable more accurate biometric identification. The companies believe their jointly developed product will lead to a new wave of smart access control and video security products with facial biometrics and liveness detection. Equally important, these products will be accurate for ethnically diverse populations and promise to maintain user privacy, executives said.
The companies are targeting a market that is being invigorated by new requirements for access security driven by COVID-19 and a need for contactless access. The market for contactless tech is expected to exceed estimates of 135 million units by 2024, as doorknobs and fingerprint scanners will be replaced by new biometric access systems.
Growth of the market will depend on the availability of products with advanced facial recognition techniques and technology. Executives from the companies detailed in a recent webinar how a combination of advanced processors are being combined into just such a product that leverages 3D biometric facial recognition rather than the more widespread 2D techniques.
“Why do we need 3D? 2D biometric facial security is done with high resolution cameras, often with infrared LED illumination, or simply with the ambient light. Features on the face are used as identifiers,” explained Dr. Andre Wong, VP Product Line Management, Lumentum. He noted that the relative distance between the eyes, the length of the bridge of the nose, and the shape of facial features are all used and compared with a reference image to unlock the system.
The issue is that “There is a lot of room for error. If the lighting isn’t right, the sensor may not be able to recognize you, or someone put the printed image of your face in front of the camera. They may be able to spoof your identity as well,” he noted.
Another key issue: 2D facial biometric security is “notorious for having gender and ethnic biases,” Wong explained. “In addition, using an image of the face also presents potential privacy concerns, since the image can also be used to link to other images of the user placed or posted elsewhere.” Some data suggests this method is roughly 80 percent accurate, he said.
Three-dimensional biometric security commonly uses infrared light to project a dense map of dots onto a face to create a “depth map” that’s unseen by the naked eye. This depth map is what is used in comparison to the subject’s reference image and is over 99 percent accurate according to Wong.
The 3D approach itself is not entirely unique. What these companies are claiming is that the combination of respective technologies will make 3D biometrics more affordable, more accurate, and therefore more ubiquitous.
Janus reference design to enable contactless access systems
Ambarella’s CV25 CVflow AI vision SoC, Lumentum’s VCSEL structured light technology, and the AR0237 RGB-IR CMOS image sensor from ON Semiconductor are all included in the Janus reference platform, which is now available. The companies previously announced their partnership in January.
Ambarella provides depth processing, anti-spoofing algorithms, 3D facial recognition, and video encoding on a single chip, which it says reduces system complexity and improves performance. Lumentum’s provides “vertical-cavity surface-emitting lasers” (VCSELs) to reliably project infrared and RGB light onto faces. RGB and IR depth sensing are performed by the ON Semiconductor sensor, rather than existing systems which use separate cameras.
“By using personal identification in the access control device, we reduce public touch points, thereby improving hygiene and safety,” said Chris Day, vice president of Marketing & Business Development for Ambarella.
He explained how the CV25 SoC is able to work in all kinds of lighting environments – changing lighting conditions can normally present issues for older generations of facial recognition systems, but newer systems like the Janus reference can account for these conditions using more advanced HDR (high dynamic range) algorithms. At the same time, the platform is able perform these functions even when subjects are wearing face masks – a critical requirement in light of the COVID-19 virus.
3D | access control | accuracy | Ambarella | biometric identification | biometric liveness detection | biometric sensors | biometrics | contactless | facial recognition | Janus | Lumentum | ON Semiconductor