A Real-time Lane Candidates Detection Method Based On Arm Embedded Platform

DOI

10.2991/aeecs-18.2018.42How to use a DOI?

Keywords

Automatic driving, Advanced Driver Assistance System, Lane detection, EDLines, . Vanishing point

Abstract

The real-time lane detection and tracking on mobile embedded platform is the key module of lane departure warning and lane keeping systems. In this paper, a framework for lane candidates detection based on improved EDLines and vanishing point estimation was proposed. EDLines directly detects one pixel width and smooth continuous line, which is free of parameters and post processing, research shows that it is more robust compared to canny descriptor. Robust vanishing point estimation can not only be applied for lane candidates filtering, but also can improve the stability of lane detection. Quantity evaluation on 20000 video frames achieves 95.3% accuracy for vanishing point estimate and 95.1% higher recall for lane candidates. Taking these two key points into account, a real-time lane detection method is proposed in these articles. In experiments under daytime and nighttime, highway and urban roads, the method is robust and stable, which is 500 fps fast in current general I7 CPU platform for 720p images, and can be applied to embedded platforms.

Copyright

© 2018, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Song Li
AU  - Jingping Jiang
PY  - 2018/03
DA  - 2018/03
TI  - A Real-time Lane Candidates Detection Method Based On Arm Embedded Platform
BT  - Proceedings of the 2018 2nd International Conference on Advances in Energy, Environment and Chemical Science (AEECS 2018)
PB  - Atlantis Press
SP  - 243
EP  - 252
SN  - 2352-5401
UR  - https://doi.org/10.2991/aeecs-18.2018.42
DO  - 10.2991/aeecs-18.2018.42
ID  - Li2018/03
ER  -