Background Noise Filtering and Clustering With 3D LiDAR Deployed in Roadside of Urban Environments

Abstract

Traffic information collection is an important foundation for intelligent transportation systems. In this paper, 3D Light Detection And Ranging (LiDAR) is deployed in the roadside of urban environments to collect vehicle and pedestrian information. A background filtering algorithm, including a mean background modeling to build a background map and a background difference method to filter static background noise points, is proposed for roadside fixed LiDAR facilities. Background points are filtered through the difference between data frames and a multi-level background map, and then there are still a small number of noise points. Aiming to reduce the noise points, a hierarchical maximum density clustering of applications with noise (HMDCAN) algorithm, utilizing both density clustering and hierarchical clustering, is proposed to effectively achieve both noise point filtering and target recognition. We verify our methods in a facility with a 16-channel LiDAR in which background filtering and target recognition are tested with different scenarios, and the accuracy rate is over 97%.