Detecting and tracking vehicles, pedestrians, and bicyclists at intersections with a stationary LiDAR
Language: English Publication details: Linköping : Statens väg- och transportforskningsinstitut, 2018Description: 5 sSubject(s): Online resources: Summary: The recent progress in the application of sensing technologies has encouraged the development of new methods for measuring the motion of road users. The exciting prospect of autonomous vehicles, long anticipated and recently determined as conceivable, has directed many of the attempts toward high quality detection and tracking of objects in close vicinity of moving vehicles. Another potential development avenue for research is a system for area-wide traffic measurement from a fixed roadside position. The precise and accurate detection and tracking of road users is a key process in a wide range of traffic and safety applications. Video cameras have become the most widely used sensing technology for detecting multiple road users and estimating their trajectories from a roadside position. The advantages of video technology include its portability, low cost, and easy installation. Mature machine vision techniques and strong research to develop algorithms for extracting data from video streams will provide additional advantages. However, the reliability of these methods is limited in adverse weather and light conditions and for heterogeneous dense traffic. Light Detection and Ranging (LiDAR) technology is a promising alternative or supplement to video technology. Due to its high cost, LiDAR technology has not received much attention until recently. The price of these sensors is expected to decrease with their proliferation in vehicles and incorporation in other applications. The presented study evaluates the feasibility of using a LiDAR-based station for detecting, identifying, and tracking vehicles and vulnerable road users (VRUs), including pedestrians and bicyclists, at various ranges from the sensor. Better knowledge of LiDARâs capabilities and limitations would be useful when considering joint use of LiDAR and video sensors. The attractiveness of LiDAR sensors will grow with the expected reduction in their cost. The presented study used the HDL-64E LiDAR sensor. A relatively compact sensor pod, the HDL-64E provides 64 laser beams for 360 o scanning of the surrounding environment. A single HDL-64E reading includes the distance and the intensity of the returned light. An algorithm was developed for processing the LiDAR readings including two major components: background identification and object tracking.The recent progress in the application of sensing technologies has encouraged the development of new methods for measuring the motion of road users. The exciting prospect of autonomous vehicles, long anticipated and recently determined as conceivable, has directed many of the attempts toward high quality detection and tracking of objects in close vicinity of moving vehicles. Another potential development avenue for research is a system for area-wide traffic measurement from a fixed roadside position. The precise and accurate detection and tracking of road users is a key process in a wide range of traffic and safety applications. Video cameras have become the most widely used sensing technology for detecting multiple road users and estimating their trajectories from a roadside position. The advantages of video technology include its portability, low cost, and easy installation. Mature machine vision techniques and strong research to develop algorithms for extracting data from video streams will provide additional advantages. However, the reliability of these methods is limited in adverse weather and light conditions and for heterogeneous dense traffic. Light Detection and Ranging (LiDAR) technology is a promising alternative or supplement to video technology. Due to its high cost, LiDAR technology has not received much attention until recently. The price of these sensors is expected to decrease with their proliferation in vehicles and incorporation in other applications. The presented study evaluates the feasibility of using a LiDAR-based station for detecting, identifying, and tracking vehicles and vulnerable road users (VRUs), including pedestrians and bicyclists, at various ranges from the sensor. Better knowledge of LiDARâs capabilities and limitations would be useful when considering joint use of LiDAR and video sensors. The attractiveness of LiDAR sensors will grow with the expected reduction in their cost. The presented study used the HDL-64E LiDAR sensor. A relatively compact sensor pod, the HDL-64E provides 64 laser beams for 360 o scanning of the surrounding environment. A single HDL-64E reading includes the distance and the intensity of the returned light. An algorithm was developed for processing the LiDAR readings including two major components: background identification and object tracking.