AUTOMATED PEDESTRIAN DATA COLLECTION FOR THE ESTIMATION OF DISCRETE-CHOICE-BASED PEDESTRIAN BEHAVIOR MODEL

Abstract

Microscopic simulations based on discrete-choice-type pedestrian walking behavior model have potential of contributing to detailed spatial planning and designs. However, in order to estimate unknown parameters it needs large amount of pedestrian behavior data. This paper develops the efficient data collection method based on digital image processing techniques. In particular, we compare two methods: the standard background subtraction method and the newly proposed particle filte method. By conducting comparative studies with respect to model calibration using the data collected through a fiel experiment, we fin that the proposed method would have high accuracy enough to be practically applicable.