Multi-sensors people detection system for heavy machines

Manh Tuan Bui (1) , Vincent Frémont (1) , Djamal Boukerroui (1) , Pierrick Letort (2)

1 Heudiasyc - Heuristique et Diagnostic des Systèmes Complexes [Compiègne]
2 CETIM - CEntre Technique des Industries Mécaniques

In this paper, we propose a multi-sensors system to detect people in the context of construction sites using heavy machines. The system includes a LIght Detection And Ranging (Lidar) sensor and a fisheye camera. We present an effective method to determine regions of interest (ROI) on fisheye images using Lidar data, which can be used with different sensors configurations. A Deformable Part Model (DPM) approach is adapted and used as the main people detector on image. We also present a specific dataset built using the multi-sensor system mounted on a heavy machine for evaluation.

Mots clés: Heavy machines; sensor fusion; pedestrian detection; deformable part model; fisheye; histogram of oriented gradients

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]; Robotique [cs.RO]; Apprentissage [cs.LG]

Manh Tuan Bui, Vincent Frémont, Djamal Boukerroui, Pierrick Letort. Multi-sensors people detection system for heavy machines. 17th International IEEE Conference on Intelligent Transportation Systems (ITSC), Oct 2014, Qingdao, China. pp.867-872. ⟨hal-01083998⟩