MINES ParisTech CAS - Centre automatique et systèmes

Curvilinear velocity estimation using low-quality stereo-vision systems and a gyrometer

Authors: N. Zarrouati, M. Hillon and N. Petit, American Control Conference 2012, pp. 4108 - 4115, 27-29 June 2012, Montreal, Canada.
We propose in this paper a method to estimate the velocity of a rigid body, using a novel stereo-vision principle. It is presented and applied in a laboratory test case which is representative of low-cost navigation for ground vehicles. The method exploits the dynamics of a scalar field obtained by weighting and averaging the brightness perceived by two embedded neighboring cameras. To be more specific, the cameras are complemented with a gyrometer to retrieve the curvilinear velocity of the moving rigid body. The proposed method is first tested on synthetic data, then on real data, and shows robustness to poor quality of image data. Significant levels of noise and blur are tested; in addition, this method does not require high resolution images, as opposed to any existing methods based on triangulation and tracking of keypoints.
Download PDF
BibTeX:
@Proceedings{,
author = {N. Zarrouati, M. Hillon and N. Petit},
editor = {},
title = {Curvilinear velocity estimation using low-quality stereo-vision systems and a gyrometer},
booktitle = {American Control Conference 2012},
volume = {},
publisher = {},
address = {Montreal},
pages = {4108 - 4115},
year = {2012},
abstract = {We propose in this paper a method to estimate the velocity of a rigid body, using a novel stereo-vision principle. It is presented and applied in a laboratory test case which is representative of low-cost navigation for ground vehicles. The method exploits the dynamics of a scalar field obtained by weighting and averaging the brightness perceived by two embedded neighboring cameras. To be more specific, the cameras are complemented with a gyrometer to retrieve the curvilinear velocity of the moving rigid body. The proposed method is first tested on synthetic data, then on real data, and shows robustness to poor quality of image data. Significant levels of noise and blur are tested; in addition, this method does not require high resolution images, as opposed to any existing methods based on triangulation and tracking of keypoints.},
keywords = {Adaptive optics, Cameras, Equations, Mathematical model, Observers, Optical imaging, Trajectory}}