Apple Patent Uses 3D Facial Recognition

 Apple

A new Apple patent application signals that iOS may be getting a facial recognition security update. The patent is titled “3-D Object Recognition” and is said to describe a new way to generate 3D facial models using 2D images. The patent was originally filed by a firm called Polar Rose from Sweden that Apple purchased in 2010.

A new Apple patent application describing face-recognition technology suggests an interesting security update for iOS. And that could be just the beginning of what the technology might enable.

The patent titled, “3D Object Recognition,” describes a novel way to generate 3D models using 2D images. It’s a follow-up to a patent Apple already owns, originally filed by Swedish firm Polar Rose, which Apple purchased in late 2010.

In essence, the technology would use multiple photos, or even video, to create a robust 3D representation of a user’s face. With this 3D representation locked in, it could then be compared, on the fly, to a 3D representation built in real time from a 2D image captured from a user’s phone.

The system has obvious applications in system security — namely, a home screen unlocking mechanism that’s not easily fooled. But the technology need not be limited to just face detection. It could also be used to identify an entire body, or employed in medical applications to identify specific organs, or even tumors.

Another Apple patent application dealing with facial recognition discussed a lower-power-intensive way to identify a device owner when he or she approaches the system. And, of course, Apple already employs face detection for identifying people in iPhoto images, but doesn’t use the feature for system security in iOS.

The patent states that there are inherent challenges in using current facial recognition technology in system security. Variations in ambient illumination and face positioning can make it difficult to reliably match source images to real-time camera images.

Apple’s proposed patent gets around the lighting roadblock by analyzing features like corners and other spatial points of reference rather than performing direct image correlation or comparisons of image parts. As for the positioning issue, it can be resolved by using multiple images for your security system’s source material. This way, even if you try to gain access to your phone by looking at its camera from a slightly askew angle, the system will have that angle in its reference library, and still let you in.


SOURCE  Wired

By 33rd Square

Subscribe to 33rd Square