China’s 500 megapixel camera is capable of mega-facial-recognition
Credit to Author: Danny Bradbury| Date: Tue, 01 Oct 2019 10:31:34 +0000
Stop bragging about how many megapixels your snazzy new prosumer DSLR camera has – China has beaten you to it. Researchers there have just announced a 500mp camera. Rather than taking stunning vacation photos, though, one of the most likely uses for this wide-angle, beer crate-sized device is for identifying people dozens of meters away using facial recognition.
Fudan University worked with the Changchun Institute of Optics, Fine Mechanics and Physics of Chinese Academy of Sciences to develop the camera, which takes both pictures and video in unparalleled detail. ABC’s story suggests that this is five times the resolution of the human eye, but scientific imaging specialist Roger Clark says that the human eye has an effective resolution of around 576mp.
Whichever figure you believe, 500 megapixels (or 5 million pixels) is more than enough to pick out faces in a stadium or on a street corner with the camera’s built-in facial recognition techniques.
This should have your privacy alarm bells ringing, but that’s just one part of the story. There’s also the possibility of a link with China’s emerging social credit system (SCS). Designed for a full national rollout in China next year, it assigns points for activities deemed socially acceptable, like donating blood and doing volunteer work, while subtracting them for negative actions like jaywalking or not showing up for restaurant reservations.
Apparently, in some local prototypes, telcos show you a message when calling someone on the social credit system’s blacklist telling you that the person you’re calling is dishonest. We didn’t think that we’d find ourselves living in Black Mirror’s excellent Nosedive episode for a while yet, but oh well, here we are.
A camera like this would be a boon for such a system. You could theoretically spot someone jaywalking, littering on the street, or doing anything else that you deemed inappropriate, and decrease their score. Excuse me while I go and find an old lady to help across the road.
Commentators have remarked that actually processing these images would be problematic because of the difficulty in beaming massive files back to a data centre for processing. However, edge computing, in which people place the computing power next to the sensor at the edge of the network, is already gaining traction, and could conceivably support a high-volume processing activity like this.
It isn’t just China that’s interested in identifying people from far away. The Intelligence Advanced Research Projects Activity (IARPA), an organisation that encourages innovation to tackle tricky problems for the US intelligence community, recently put out a request for information under the heading Biometric Recognition and Identification at Altitude and Range (BRIAR). It asks for “research efforts and datasets that may be useful in planning a program focused on advancing the state-of-the-art of biometric recognition and identification at altitude and range”.
Spotting people from a long way away opens up more possibilities for recognition algorithms using other characteristics such as gait recognition (and yes, the Chinese are already doing this). What next?