Robust U-Net-based Road Lane Markings Detection for Autonomous Driving

Engineering notes

The effectiveness of the system was validated by testing on CARLA simulator, an open-source simulator for research on autonomous driving.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

The rapid development of artificial intelligence leads to many studies on autonomous robot and self-driving vehicles, in which, autonomous driving plays one of the most important roles in supporting a robot or a car to be able to observe, move, and avoid obstacles. In this paper, a new method is proposed to detect road lane markings for supporting surveillance and autonomous driving. Images captured from a front-view camera are fed forward into a semantic segmentation network to extract features for detecting road lane markings, the network is constructed based on U-Net architecture, a convolutional neural network developed for biomedical image segmentation, then Hough Transform method is implemented in the system to determine lines in the segmentation network outcomes. In addition, Hough Transform yields plenty of lines from segmented images, thus K-means Clustering algorithm is also investigated to compute and point out the fittest line with each road lane marking. The effectiveness of the system was validated by testing on CARLA simulator, an open-source simulator for research on autonomous driving. Experiments proved that the proposed method can work with favorable results.

Links and sources

• Official / arXiv page

Comments