A Novel Background Filtering Method With Automatic Parameter Adjustment for Real-Time Roadside-LiDAR Sensing System

Roadside-light detection and ranging (LiDAR) sensing system can provide the full trajectories of all-type road users around the deployed traffic facility, which is a new-generation traffic data to assist traffic safety and operation applications. Background filtering is a critical step of roadside-LiDAR data processing that significantly affects processing quality and efficiency. Existing background filtering methods heavily rely on statistical or empirical approaches for model parameter determination, so they normally work well for some scenarios but cannot accommodate others due to different traffic characteristics. In this article, a novel background filtering method is developed, whose model parameters can be automatically determined with the site's traffic-related measurements. The new method is designed to work on a ranging image data structure derived from the spherical features of the LiDAR sensor. The performance evaluations are conducted at three signalized intersections equipped with 32-line LiDAR sensor roadside-LiDAR under 10-Hz operational frequency, which demonstrated that the developed method can guarantee a high background filtering accuracy with more underlying foreground points detected while simultaneously achieving a significantly higher processing efficiency in comparison with existing methods.

Automated summary

The roadside-LiDAR sensing system provides valuable traffic data for safety and operation applications by capturing the trajectories of road users. Background filtering plays a crucial role in processing LiDAR data, but existing methods have limitations in adapting to different traffic scenarios. This article presents a novel background filtering method that can automatically determine model parameters based on traffic-related measurements, resulting in higher accuracy and efficiency compared to existing methods.