Kyocera, FotoNation to equip autonomous cars with AI cameras

Companies in the automotive industry are constantly on the lookout for alliances.

For example, Kyocera and FotoNation announced Tuesday (May 17) a partnership agreement to develop intelligent automotive camera technology—deemed critical in the coming era of semi- and fully autonomous cars.

Kyocera has already dabbled in the auto sector with its rearview camera modules. FotoNation, with the lion's share of computational imaging solutions for mobile phones, entered the driver monitoring system market last year.

"The two companies' interests are aligned," Sumat Mehra, senior vice president of marketing and business development at FotoNation, told EE Times, "to expand each company's presence in the automotive market"—well beyond what they offer today.

Mehra said that FotoNation has amassed a collection of advanced computer vision IPs ranging from object detection algorithms to correct image distortions. These IPs, hardware accelerators and image processing units (IPUs) "can be embedded into various processors and SoCs," he explained. Other automotive vision solutions offered by FotoNation include driver identification, surround view, e-mirror, smart rearview cameras, and 360-degree occupant monitoring.

Neither Kyocera nor FotoNation, however, is revealing specifics on the technology they plan to develop.

Mehra declined to comment on FotoNation's processor choice, or where its IPs will be embedded. Nor did the company talk about what's inside Kyocera's automotive camera modules—beyond a sensor, processors and special filters.

FotoNation's driver monitoring system, unveiled last year, runs on Texas Instruments' TDA3x SoC processor. FotoNation's technology identifies the driver, detects driver drowsiness and distraction.

No more dumb cameras

Given that purpose-built automotive vision SoCs—such as those from Mobileye, NXP and Texas Instruments—are already reaching the commercial automotive market, what sort of new innovations will Kyocera-FotoNation's automotive camera technology bring?

Mehra stressed that the market for "intelligent" automotive cameras is "growing very fast." Advanced computer vision algorithms are replacing "dumb cameras which just display [on a screen inside a vehicle] what they see." Intelligent cameras can identify objects, separate a bicycle from a truck coming from behind, understand what's going on around a vehicle, and help the car make decisions, he explained.

Mehra boasted that even before Nvidia started talking about deep learning for automotive computer vision, "We've been using some deep learning and neural network-based systems to develop our own computer vision technologies. We have built a database of more than 20 million images for face detection, for example."

FotoNation, whose experience is rooted in detecting and tracking faces and eyes, has over the years expanded its technology base. The company now has biometrics technology (through the acquisition of Smart Sensors last year) that enables iris detection, image capture and recognition, and provides identity management.

Mehra predicted that the intelligent automotive camera technology would be broadly used in vehicles ranging from high-end models to mass-market cars. The new automotive technology to be jointly developed by Kyocera and FotoNation will be on the market by 2019 or 2020, he noted.

-Junko Yoshida,
EE Times