Side-Fire Lidar-Based Vehicle Classification

Abstract

Vehicle classification data are used for numerous transportation applications. Most of the classification data come from permanent in-pavement sensors or temporary sensors mounted on the pavement. This study developed a lidar-based classification system that used data from sensors mounted in a side-fire configuration next to the road. The classification system first distinguished between vehicle returns and nonvehicle returns, and then clustered the vehicle returns into individual vehicles. An algorithm examined each vehicle cluster to check for any evidence of partial occlusion from another vehicle. Several measurements were taken from each nonoccluded cluster to classify the vehicle into one of six classes: motorcycle, passenger vehicle, passenger vehicle pulling a trailer, single-unit truck, single-unit truck pulling a trailer, and multiunit truck. The algorithm was evaluated at six locations under various traffic conditions. When compared with concurrent video ground truth data for more than 27,000 vehicles on a per vehicle basis, 11% of the vehicles were suspected of being partially occluded. The algorithm correctly classified more than 99.5% of the remaining nonoccluded vehicles. This research uncovered emerging challenges that likely applied to most classification systems: differentiating commuter cars from motorcycles. Occlusions were inevitable in this proof of concept study because the lidar sensors were mounted roughly 6 ft above the road, well below the tops of many vehicles. Ultimately a combination of a higher vantage point and shape information will greatly reduce the effects of occlusions.