Une approche collaborative multi-vues pour la détection de piétons dans les systèmes de transport intelligents
As pedestrians are the most vulnerable road users, the design of accurate pedestrian detection methods is a priority task for safe intelligent transport systems.
However, traditional monocular pedestrian detection methods are limited, particularly in the case of occlusions. Therefore, a collaborative perception scheme in which vehicles no longer limit their input data to their immediate on-board sensors and instead exploit data from remote sensors is needed to obtain a more complete perception of the environment. In this work, we first propose a new public database (I2V-MVPD) that combines synchronised images from both a mobile camera in a car and a static camera in the road infrastructure. We also propose a new multi-view pedestrian detection approach based on collaborative intelligence between the vehicles and the infrastructure. Our results show a significant improvement in detection performance compared with monocular detection. 
A. Ben Khalifa, I. Alouani, M. Ali Mahjoub, A. Rivenq,
Future Generation Computer Systems, 113, 506 (2020)
