To address the issues of misdetection and missed detection in nighttime lane marking detection due to factors such as occlusion, insufficient lighting, glare, and interference from objects similar to lane markings, this paper proposes a nighttime structured road lane marking detection method based on LiDAR and camera fusion. LiDAR is used in the first stage to detect road boundaries and define the lane detection region based on the extracted boundary points. In the second stage, a camera performs lane detection within the LiDAR-defined region using an improved UFLD method. The first-stage method achieves an accuracy of 93.2% on a randomly selected 300 frames from the KITTI dataset. The second-stage method achieves F1 scores of 73.8% on the CULane dataset and accuracies of 95.26% and 77.61% on the TuSimple and self-built datasets, respectively. The real-world deployment of the fusion method shows a 1.05% improvement in accuracy compared to the non-fused algorithm. The experimental results demonstrate that the proposed method has good performance in lane marking detection on structured roads at night.
Du et al. (Mon,) studied this question.