Skip to Main content Skip to Navigation

Suivi visuel par filtrage particulaire. Application à l'interaction homme-robot

Abstract : Nowadays, the major challenge of Robotics is certainly the personal robot which is able to be of use to human. Many researches in this field are axed on the development of autonomous robots intended to evolve in large human environments. Obviously, this perspective shows the problem of the interaction and the relation between men and robots. Indeed, during the navigation, the robot must be able to detect humans presence in its vicinity and to take them into account, by explicit manner, in order to avoid them or to let them pass: the goal is to facilitate and to make safe their displacements. Moreover, it must have capacities of interaction such as gestures' recognition which allow Men to communicate with him. This thesis is focused more specifically on the detection and the tracking of people and also on the recognition of elementary gestures from video stream of a color camera embeded on the robot. Particle filter is well suited to this context. It allows to avoid any restrictive assumption about the probabilities distributions which are throw in the characterization of the problem. Moreover, this formalism enables a straight combination/fusion of several measurement cues. Despite of this observation, particle filter data fusion seems to us to be little exploited and often confined to a relatively restricted number of visual cues. We propose various filtering strategies where visual information such as shape, color and motion are taken into account in the importance function and the measurement model. We compare and evaluate these filtering strategies in order to show which combination of visual cues and particle filter algorithm are more suitable to the interaction modalities that we consider for our tour-robot which is supposed to hail visitors, to interact with them and to guide them. Our last contribution relates to the recognition of symbolic gestures which enable to communicate with the robot. An efficient particle filter strategy is proposed in order to trac k the hand and to recognize at the same time its configuration and gesture dynamic in video stream.
Document type :
Complete list of metadatas
Contributor : Emilie Marchand <>
Submitted on : Friday, March 30, 2007 - 4:21:53 PM
Last modification on : Tuesday, November 24, 2020 - 12:12:03 PM
Long-term archiving on: : Tuesday, April 6, 2010 - 11:19:03 PM


  • HAL Id : tel-00139428, version 1


Ludovic Brèthes. Suivi visuel par filtrage particulaire. Application à l'interaction homme-robot. Automatique / Robotique. Université Paul Sabatier - Toulouse III, 2005. Français. ⟨tel-00139428⟩



Record views


Files downloads