HistoGrid: Robust LiDAR-Based Traffic Monitoring

Smart traffic monitoring is becoming increasingly important to manage the ever-growing amount of road users, to mitigate traffic congestion and improve road safety. For this purpose, a robust infrastructure-based perception system is required, which uses road-side sensors mounted for example on traffic light poles to detect and track all relevant objects in the environment. To ensure robustness, along with cameras also LiDAR sensors are of interest, due to their excellent performance even at night-time. However, reliable object detection from LiDAR data is still an open challenge and many approaches cannot provide the required quality for 24/7 operation. To overcome this gap, we propose in this work HistoGrid, a new approach that can robustly separate all static elements in the environment from the relevant (dynamic) objects. This is underlined by the results from our comprehensive studies in simulation as well as using a real testbed of an intersection.