LiDAR-Based Traffic Data Extraction for Noise Mapping

Traffic noise pollution is a major environmental challenge, especially in urban areas, driven by transportation networks. Conventional noise mapping techniques rely on simulations based on limited traffic data or stationary monitoring stations, which are difficult to deploy and scale for large areas. Although traffic data can be extracted from video-based methods, these approaches often require complex calibration, which can introduce inaccuracies. A promising tool for smart city infrastructure, LiDAR captures 3D spatial data from which traffic information can be extracted through advanced processing techniques. This study explores the application of LiDAR for traffic noise mapping. A convolutional network processes point cloud data (PointPillars) to detect and classify vehicles directly from LiDAR data. Next, SORT algorithm tracks vehicle positions over time, enabling the estimation of trajectories and speeds. Traffic parameters are provided as inputs to a model that calculates street noise emissions based on vehicle dynamics. The sound power of the corresponding linear source is estimated and a noise map is generated. The accuracy of the proposed method is assessed through comparison with standard noise map simulation tools and measurements. This LiDAR-based approach has the potential to be scaled for large urban areas and enhance noise assessments for urban planning support.