Vehicle occlusion model for traffic monitoring
John E. Laird, Glenn Geers, Yang Wang, Chun Tung Chou · 2010
This paper focuses on modelling sensor placement for traffic monitoring using vision based sensors. A significant problem with using such sensors is that vehicles can be merged in an amorphous group making detection difficult. Sensor placement has a direct impact on the efficiency of traffic monitoring. We simulate various sensor placements and measure the apparent occlusion between vehicles. Parametric distributions are utilised for modelling vehicle height, length, width and gap length when simulating traffic mixes. These models are used to predict the probability of on-axis and off-axis occlusion of vehicles as perceived from different sensor locations. These occlusion models show that poor sensor placement can have a direct impact on the detection of vehicles.