Lane Marking Detection Using Low-Channel Roadside LiDAR
Bowen Gong, Benhan Zhao, Yue Wang, Ciyun Lin, Hongchao Liu
Abstract
Lane marking is the criterion line for the roadside unit to extract lane-level and high-resolution microlevel traffic data (HRMTD), which is essential information in the vehicle-to-infrastructure (V2I) cooperative. For lane marking detection in complex environments, inconsistent lane width and indistinctive laser intensity are two significant challenges for roadside LiDAR. To address these issues, we proposed an accuracy and robustness lane marking detection method. With the low-channel roadside LiDAR, we divided LiDAR scanned area into grids to extract vehicle points and filter out noise points. Then, the tilt angle of the lane markings was estimated using the grids. Finally, the lane number and lane marking were extracted by vehicle passing area identification according to the vehicle points and grids distribution. Experimental results show that the average distance error (ADE) of lane marking detection is less than 0.08 m, and the average detection time is 111 s in different environments.