Acte de conférence, colloque et atelier

Localizing People in Crosswalks using Visual Odometry: Preliminary Results

Lalonde, M., St-Charles, P.-L., Loupias, D., Chapdelaine, C. et Foucher, S. "Localizing People in Crosswalks using Visual Odometry: Preliminary Results" dans Proc of the 7th International Conference on Pattern Recognition Applications and Methods (ICPRAM 2018). Funchal, Madeira - Portugal, du 16 au 18 janvier 2018

ABSTRACT
This paper describes a prototype for the localization of pedestrians carrying a video camera. The application envisioned here is to analyze the trajectories of blind people going across long crosswalks when following an accessible pedestrian signal (APS), in the context of signal optimization. Instead of relying on an observer for manually logging the subjects’ position at regular time intervals with respect to the crosswalk, we propose to equip the subjects with a wearable camera: a visual odometry algorithm then recovers the trajectory and spatial analysis can then determine to which extent the subject remained within reasonable boundaries while performing the crossing. Preliminary tests in conditions similar to a street crossing show that our results qualitatively agree with the physical behavior of the subject.