Subaru's EyeSight tackled pre-collision assistance with a deliberately legible sensor choice: dual cameras mounted near the rear-view mirror, rather than a black-box radar-only setup.

The 2017 Subaru-authored paper describes improvements to the Pre-Collision Braking System, including faster reaction to crossing pedestrians and better nighttime pedestrian recognition by reducing recognition loss from darkness and clothing color.

The genius was product architecture as much as detection. A stereo pair made depth visible from ordinary images, so the system could reason about objects, distance and closing risk in one camera-based pipeline.

Use two cameras as car eyes

Stereo cameras gave the car distance information without relying on a separate ranging sensor.

The system could improve through targeted recognition changes, such as night pedestrian detection.

The windshield-mounted package fit normal cars, making the safety feature easier to scale.

A familiar sensor can beat a heavier stack when the product problem is perception plus timing, not raw sensor prestige.

The J-STAGE paper records the system as Subaru-developed, published in 2017, and focused on reducing traffic accidents through refinements to EyeSight's pre-collision braking.

FOLLOW THE EVIDENCE

The sources

  1. Development of Improvements to Driver Assistance System “EyeSight” for Reduction of Traffic Accidents api.openalex.org
  2. Development of Improvements to Driver Assistance System “EyeSight” for Reduction of Traffic Accidents jstage.jst.go.jp