Thermal imaging has been around for decades in various industrial and military applications, but could it be the extra sense that makes autonomous mobility machines safer than human-driven vehicles? AdaSky’s Sharon Fiss, Director of Sales Engineering, makes a convincing case that detecting thermal signatures and fusing with other sensor data will provide for a super-human vision and perception.
In the above video filmed on the streets of Las Vegas at CES2019, Fiss details some of the specifications for AdaSky’s FIR (Far Infrared) sensor with dedicated Image Signal Processor), which include:
- Temperature resolution of .05° Celsius, allowing it to not only tell between living and non-living objects but to distinguish between the type of living objects (human, animal).
- Field of view between 17° to 30°, which allows detection of animals and pedestrians at up to 120 meters and 200 meters, respectively.
- VGA quality images refreshed at 60 frames per second providing a complement to camera sensors.
Fiss points out that another benefit of infrared sensing is that it is impervious to headlights and direct sun that wreaks havoc with cameras and human eyes. In their recent white paper (PDF), AdaSky explains that they have implemented a hardware solution to prevent the “sunburn” effect which can be an issue with shutterless FIR. It is a passive solution, so it doesn’t interfere with or from other sensors like might happen with active sensors (e.g. radar, lidar).
Beyond sensing, AdaSky claims that their algorithms, running on their dedicated ISP, reduce power consumption relative to discrete solutions, allow for lower intensity processors for fusing the FIR data with other sensor data and speeds an OEM’s time-to-market. AdaSky indicates that by using new materials, combined with advanced detection fabrication techniques from the semiconductor industry, thermal imagers have improved quality in a smaller size with affordable pricing.
The size advantage was on display at CES2019, as Magnet Marelli hid AdaSky’s thermal imaging technology into a headlight assembly, which is fitting for a sensor which sees the invisible.