Vehicle Occupancy Detection for HOV/HOT Lanes Enforcement

摘要: High-Occupancy Vehicle (HOV) and HighOccupancy Toll (HOT) lanes have gained interest in recent years since they provide innovative solutions to roadway congestion traffic safety urban areas. Enforcement is one of the key components HOV/HOT lane operations order maintain efficiency operational integrity facilities. The incorporation video imagery monitor has driven development automatic occupancy detection systems oriented enforcement. However, seeing through vehicle glass a challenging task, independently type sensor use. In this paper we present front-seat as part low-computational computer vision enforcement system. Two main stages composes proposed solution: (i) A novel front windshield image segmentation, using Local Appearance Model (LAM) approach supported on multi-dimensional local features (HOGs); (ii) machine learning classifier detect representations occupant region-of-interest. Experimental evaluation was conducted dataset frontal images obtained real highway scenario. comparative against stateof-art performed results reported confirm good performance system robustness segmentation technique.