Asservissement visuel basé mélanges de gaussiennes photométriques

Abstract

Photometric visual servoing does not require any feature detections, matching or tracking process. However, the convergence to the solution pose is highly related to the distance between the desired and the intial camera pose. This work proposes a new dense visual feature to enlarge the convergence domain : the photometric Gaussian mixtures. During the servoing, additionaly to the camera velocities, the Gaussian spreads are also optimized. The camera is able to converge to a desired pose even from a far initial one and still has the convergence accuracy obtained with the photometric. Simulation that overcomes the state1. Une version anglophone de cet article a été publiée dans les actes de la conférence IROS (IEEE/RSJ International Conference on Intelligent Robots and Systems 2015 à Hambourg) [6]. of-the-art and real experiments highlight the success of our

8 Figures and Tables

Cite this paper

@inproceedings{Crombez2016AsservissementVB, title={Asservissement visuel basé mélanges de gaussiennes photométriques}, author={Nathan Crombez and Guillaume Caron and El Mustapha Mouaddib and Jules Verne}, year={2016} }