Laser-based vehicles tracking and classification using occlusion reasoning and confidence estimation

In this paper, we present a robust approach for the detection, tracking and classification of multiple vehicles using a vehicle mounted laser scanner working independently in highways an urban centers. Our classification is based on different criteria: geometrical configuration, occlusion reasoning, sensor specifications and tracking information. The estimated confidence level is thus computed accounting the classification, the geometrical configuration and the tracking duration. Our system has been validated under various conditions (highways, urban centers) with three different laser scanners and proved is robustness on real data and with real time constraints.