Geometric calibration of a camera or a stereoscopic vision sensor

Imagerie

Jean-José ORTEU - École des Mines d'Albi

Professeur des Ecoles des Mines. Directeur adjoint du centre de recherche CROMeP de l’École des Mines d'Albi. Directeur adjoint du GDR CNRS « Mesures de champs et Identification en Mécanique des Solides » Responsable du groupe de recherche « Mesure, Contrôle et Surveillance » de l'IGM. Domaines de recherche : métrologie (3D) par vision artificielle, photomécanique, surveillance de procédés, CND par vision. Domaines d'enseignement : automatique, signaux et systèmes, calcul numérique, photomécanique.

Site Web de l'auteur : http://www.orteu.fr/

The geometric calibration of a camera consists in determining the existing mathematical relation between the coordinates of the 3D points of the scene observed and the 2D coordinates of their projection in the image (image-points). This step of calibration represents the initial point for several applications of the artificial vision, such as for instance: recognition and objects localization, control procedure for the dimensional measurement of cast parts, reconstruction of the environment for the navigation of a mobile robot, etc. Calibrating a camera is about choosing a model of camera a priori and then determining the parameters of that model. We will describe the main models of camera used as well as the main methods suggested in order to determine the parameters of the model chosen. To get tridimensional information, it is necessary to combine two cameras in order to constitute a stereoscopic vision sensor. The calibration of such a sensor is a specific problem and is also described.

Geometric models of camera, camera calibration, geometric distortions, stereoscopic vision.

Basics in linear algebra.

Upon completion of module, students should be able to basically explain how to mathematically model a camera and how to experimentally determine the parameters of the model chosen.

• Course

• Modelling and calibration of a camera

• Modelling and calibration of a stereoscopic vision system

PRN - Le Mans Université

Licence GNU