LaneNet: Real-Time Lane Detection Networks for Autonomous Driving

Ze Wang,Weiqiang Ren,Qiang Qiu

DOI: https://doi.org/10.48550/arXiv.1807.01726

2018-07-05

Abstract:Lane detection is to detect lanes on the road and provide the accurate location and shape of each lane. It severs as one of the key techniques to enable modern assisted and autonomous driving systems. However, several unique properties of lanes challenge the detection methods. The lack of distinctive features makes lane detection algorithms tend to be confused by other objects with similar local appearance. Moreover, the inconsistent number of lanes on a road as well as diverse lane line patterns, e.g. solid, broken, single, double, merging, and splitting lines further hamper the performance. In this paper, we propose a deep neural network based method, named LaneNet, to break down the lane detection into two stages: lane edge proposal and lane line localization. Stage one uses a lane edge proposal network for pixel-wise lane edge classification, and the lane line localization network in stage two then detects lane lines based on lane edge proposals. Please note that the goal of our LaneNet is built to detect lane line only, which introduces more difficulties on suppressing the false detections on the similar lane marks on the road like arrows and characters. Despite all the difficulties, our lane detection is shown to be robust to both highway and urban road scenarios method without relying on any assumptions on the lane number or the lane line patterns. The high running speed and low computational cost endow our LaneNet the capability of being deployed on vehicle-based systems. Experiments validate that our LaneNet consistently delivers outstanding performances on real world traffic scenarios.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The paper attempts to address the challenges of lane detection in autonomous driving systems. Specifically, the paper identifies several unique problems faced by lane detection:

1. **Lack of significant features**: The appearance of lanes is usually very simple, without complex or significant features, which makes lane detection algorithms easily confused by other objects with similar local appearances.
2. **Diversity of lane patterns**: The number of lanes on the road is inconsistent, and the styles of lane lines are diverse (such as solid lines, dashed lines, single lines, double lines, merging lines, and diverging lines), which further hinders detection performance.
3. **Limitations of existing methods**: Methods based on handcrafted features can only solve problems in limited scenarios, and most existing methods rely on strict assumptions, such as lanes being parallel or nearly parallel, which do not always hold in urban environments.

To address these issues, the paper proposes a deep neural network-based method called LaneNet. LaneNet decomposes the lane detection task into two stages: lane edge proposal and lane line localization. Through this approach, LaneNet can achieve robust lane detection in various scenarios on highways and urban roads without relying on assumptions about the number of lanes or the styles of lane lines. Additionally, LaneNet has high running speed and low computational cost, making it suitable for deployment in onboard systems.