Vehicle Recognition from Lidar Data

This paper focuses on the potential of using airborne laser scanning technology for transportation applications, especially for identifying moving objects on roads. An adaptive thresholding algorithm is used to segment the LiDAR point cloud, which is followed by a selection process to extract the vehicles. The LiDAR data are capable of measuring the vertical profile of a vehicle, and hence provide a base for distinguishing major vehicle types. Various techniques, such as the use of statistical, neural, and rulebased classifiers, were used to recognize the vehicle classes. The classification is based on features derived from a principal component transformation. Thereafter the extracted vehicles were classified into main categories, such as passenger cars, multipurpose vehicles, and trucks. The feasibility of the developed method to effectively extract vehicles from LiDAR data has been demonstrated on several datasets. The proposed technique makes LiDAR suitable for new transportation applications, such as collecting data for traffic flow monitoring and management, including data on the vehicle count, traffic density, and velocity.