Subaru launches Levorg model with stereoscopic camera ADAS

Author: EIS Release Date: Aug 31, 2020


The vehicle has been launched in Japan this week with an enhanced version of EyeSight, the vehicle manufacturer’s vision-based ADAS.

The Levorg’s EyeSight is powered by Xilinx’s automotive-qualified (XA) 16nm Zynq UltraScale+ multiprocessor SoC (MPSoC). According to the company, it delivers the high-performance, ultra-low latency, and functional safety (ASIL) capabilities to accurately depict and react to dynamic driving scenarios.

Tetsuo Fujinuki, chief technology officer at Subaru, explained the stereoscopic vision system. “[The image processing] scans everything captured by stereo cameras and creates high-precision, 3D point clouds, enabling us to offer advanced features such as pre-collision braking at an intersection and assisting with hands-off driving in traffic congestion on a highway”.

The stereoscopic camera vision system increases the field of view (FoV) in the ADAS which improves situational awareness in road safety. For example, explained Wayne Lyons, director of automotive and strategic customer marketing at Xilinx a conventional ADAS can detect a cardboard box in the road, but situational awareness enables it to determine whether it is an empty box, which will cause no damage or one with that has, say a fridge in it, and needs to be avoided.

The two cameras enhance the FoV but also increase the processing power required for stereoscopic images which are processed into 3D point clouds for features such as adaptive cruise control, lane-keeping assist and pre-collision braking, which require high speed algorithms for their execution.

The MPSoC improves the speed of running the algorithms, said Lyons, in advanced functions such as unintentional acceleration and runaway in response to driver error.

The Levorg also has an enhanced driver monitoring system (DMS) which can detect if the driver is distracted, and provides functions for driver convenience, such as active lane change assistance and auto-deceleration based on a curve prediction when going around corners.