clock menu more-arrow no yes

Filed under:

This chip is bringing intelligent computer vision to the world of thermal cameras

Movidius and FLIR have come together to make a thermal camera core

If you buy something from a Verge link, Vox Media may earn a commission. See our ethics statement.

For several years car companies like Audi and BMW have incorporated thermal cameras powered by FLIR Systems, giving drivers the ability to enhance their night vision. Yesterday FLIR announced the next generation of its thermal imaging technology, the Boson camera core. It uses the Myriad 2 chip from Movidius, the vision-processing unit which powered the autonomous features and computer vision inside DJI's latest drone, the Phantom 4. This could allow the next generation of cars to have night vision that works alongside autonomous driving features, helping the vehicles make smart decisions about when to avoid obstacles.

"Compared the FLIR’s previous generation Tau 2, our revolutionary new Boson core is about a tenth the size, a fifth the weight, and at least half the power. These differences will have important implications in applications such as drones by extending battery life. flight times and range. It also means we might see much smaller form factor drones adopt high-performance thermal imaging capability that has previously been unobtainable," said Jeff Frank, senior VP of Global Product Strategy at FLIR. "Powerful on-board analytics that are enabled by the Movidius Myriad 2 can extend the useful information generated by this core from standard digital video to things like object detection and tracking, human or animal sensing, and dramatic improvements in image fidelity."

By putting a system on a sensor, the Movidius chip opens up new opportunities for FLIR's cameras in applications which demand extremely rapid response times. "With onboard intelligence, much of the analytics are crunched on the sensor itself, and only small amounts of meta data is shared via the network," says Movidius CEO says Remi El-Ouazzane. "In addition to these bandwidth improvements, running the neural nets at the sensor source is much faster, which is important for low-latency real-time applications such as automotive, drones or security cameras."

So far we've seen big advances in drones and thermal cameras with the addition of the new Movidius chip. It will be very interesting to see what happens once it starts rolling out on Android phones.