Road Boundary-Enhanced Automatic Background Filtering for Roadside Lidar Sensors

Abstract

The roadside-deployed lidar sensor provides a solution to obtain the real-time, high-resolution micro traffic data (HRMTD) of all road users in the mixed traffic situation (connected vehicles and unconnected vehicles both exist on the roads). Background filtering is a necessary and important step for the HRMTD collection to serve the connected vehicle. Without excluding the background points, the accuracy of Euclidean-based object clustering and tracking algorithms can be reduced. The widely used method-3D-density-statistic filtering (3D-DSF) for roadside lidar background filtering can effectively exclude the background points for free-flow conditions. However, the performance of 3D-DSF can be greatly influenced by congested traffic conditions. This article presents a revised 3D-DSF algorithm to automatically extract background points by involving the road-boundary information. This new method is named <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">road-boundary-enhanced, 3D-density statistic filtering</i> ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3D-DSFRB</i> ). This algorithm involves the boundary of the historical trajectories of road users as the region of interest (ROI) to enhance the accuracy of background filtering. A revised grid-based method was developed for road-boundary ID. The 3D-DSF was only applied for the area outside of the ROI. Within the ROI, only ground surface was excluded. Case studies were conducted to evaluate the effectiveness of the 3D-DSFRB algorithm. The results showed that the 3D-DSFRB can filter background points for both free-flow conditions and congested traffic conditions. The time cost of the 3D-DSFRB was also reduced compared to the 3D-DSF. Compared to the state of the art, the 3D-DSFRB improved the accuracy of background filtering.