A comprehensive representation of traffic scenes for autonomous driving

(De Gruyter) – In order to react to the world around them autonomous vehicles and sophisticated driver assistance systems must perceive their environment in 3D through lasers, cameras, and other technologies. Next generation automotive vision systems will use at least two million 3D points to understand the environment around them, which is a daunting amount of data to have to process.

To cope with such a large amount of data in real-time, the authors developed a medium level representation, named Stixel world. This representation condenses the relevant scene information by three orders of magnitude.

Since traffic scenes are dominated by planar horizontal and vertical surfaces the representation approximates the 3D scene by means of thin planar rectangles called Stixel.

The authors summarize the progress of their Stixel world, starting with a rather simple representation based on a flat world assumption.

A major break-through was achieved by introducing deep-learning that allowed to incorporate rich semantic information. In its most recent form, the Stixel world encodes geometric, semantic and motion cues and is capable of handling difficult to analyze terrain including even the steepest roads in San Francisco.

***

Edited for Content and Length by Dr. Matthew A. Hood.

The full article can be found at De Gruyter in the journal of at – Automatisierungtechnik.

AT – Automation Technology: Methods and Applications of Control, Regulation, and Information Technology

Impact Factor 2017 : 0.503

DOI: 10.1515/auto-2018-0029

 

Pin It on Pinterest