Key points are not available for this paper at this time.
The high-resolution traffic data (HRTD) of all roadway users is important to connected-vehicle systems, traffic safety analysis, performance evaluation and fuel efficiency study. The roadside LiDAR (light detection and ranging) sensors can provide HRTD by collecting real-time 3D point clouds of surrounding objects, which is significant to connected-vehicle applications with mixed traffic flow - connected and unconnected road users. The background filtering is a necessary step to improve the accuracy and efficiency of HRTD extraction from raw LiDAR data. At the same time, the lane space information can further enhance the accuracy and reliability of data extraction. This paper presents an algorithm to extract background automatically from the LiDAR data based on the density of points in 3D space; and a multi-classified density-based spatial clustering method (MC-DBSCAN) is developed to identify road lanes automatically. The boundaries of road lanes are automatically located with aggregated trajectories of vehicles. The algorithms were applied for roadside LiDAR data preprocessing at different sites. Data and results from one site are presented in this paper.
Wu et al. (Sun,) studied this question.