Adaptive Polar-Grid Gaussian-Mixture Model for Foreground Segmentation Using Roadside LiDAR

Roadside LiDAR has become an important sensor for the detection of objects in cities, such as vehicles and pedestrians, which is due to its advantages all-weather operation high-ranging accuracy. In order serve intelligent transportation system, efficient accurate segmentation pedestrians needed coverage area LiDAR. this study, a roadside was fixed on brackets both sides road obtain point-cloud information urban surrounding environment. A method that based scanning proposed. First, polar grid coordinates constructed count rotations original angle distance point cloud, background image dynamically updated over time. By aiming at complex environment interference trees light poles background, adaptive polar-grid Gaussian-mixture model (APG-GMM) uses proposed improve accuracy foreground segmentation. density-adaptive DBSCAN target-clustering algorithm proposed, well dynamic neighborhood radius, solve problem low clustering caused by uneven density clouds are collected LiDAR, divide into pedestrians. Finally, tested intersections roads with dense traffic flows. The experimental results show can segment cluster while reducing number calculations time complexity.