A layered approach to robust lane detection at night

A layered approach is designed to address many of the real-world problems that an inexpensive lane detection system would encounter. A region of interest is first extracted from the image followed by an enhancement procedure to manipulate the shape of the lane markers. The extracted region is then converted to binary using an adaptive threshold. A model based line detection system hypothesizes lane position. Finally, an iterated matched filtering scheme estimates the final lane position. The developed system shows good performance when tested on real-world data that contains fluctuating illumination and a variety of traffic conditions.

I. Introduction

DRIVER safety has always been an area of interest to research since driver distractions account for numerous accidents on highways. In multiple studies performed by the National Highway Traffic Safety Administration (NHTSA), it was reported that over 20% of the accidents are due to driver distractions. Further analysis showed that approximately 60% of these accidents were caused by drivers who were either talking on the phone, adjusting GPS devices or tampering with the CD player [1], [2], [3]. The high fatality rate has prompted industry and academic institutes to focus on embedding smart systems in the automobile that can aid a driver during a commute.