Monte Carlo-based Bi-directional pedestrian counting with compound-eye sensor

Abstract

In this paper, we propose a method for estimating the number of pedestrians walking in opposite directions, as in the case of a corridor. The proposed method utilizes a ceiling-mounted compound-eye sensor composed of two binary sensors with partially overlapping sensing regions. The output of the compound-eye sensor is sent to a monitoring server, where a number of Monte Carlo simulations are conducted. The simulation scenario with the smallest difference from the output history of the compound-eye sensor is selected as an estimate of the number of pedestrians. Simulations show that in crowded situations the relative error in the proposed method decreases by up to 45% compared to an existing method.