Vision-based Lane Detection and Tracking for Driver Assistance Systems: a Survey

Lane detection and tracking has been an active research area in the past twenty years mainly for the driver assistance application. Due to the large variations of traffic scenes and illumination conditions, this problem causes the usage of diverse approaches and sensing modalities. In this paper, we review the vision-based lane detection and tracking methods complemented with other sensor information when necessary. Approaches that adopt conventional computer vision techniques are reviewed and compared according to the separate functional modules in a generic framework. The recently developed machine learning especially deep learning based methods in the limited literature are analysed and also discussed, demonstrating high potential in the current and future challenging lane perception scenarios. While impressive achievements have been demonstrated under limited scenarios, new ideas and approaches are still desired such that the next generation robust and efficient system can be built in the service of autonomous vehicles.