The invention concerns methods for calibrating at least two video cameras (1, 2) for a stereoscopic device (3). The inventive method is essentially characterized in that it consists in : providing, on the lane portion (4), nine marks of hue other than that of the lane (4), sequenced on a first set of three concurrent virtual straight lines in a first point and distributed in specific manner on a second set of concurrent virtual straight lines in a second point ; forming, with each of the cameras (1, 2) an image of said lane portion (4) ; defining, in each of said two images, one characteristic point of each mark image ; determining, by means of said characteristic points, six concurrent straight line images respectively in two concurrent points ; and processing the video signals delivered by each video camera such that said signals are representative of two images suitable for forming a stereoscopic video image. The invention is particularly useful for determining the occupancy condition of a lane portion and for detecting incidents on said lane portion.

The perspective-N-point problem is a well known issue in computer vision. It consists in the determination of the distance between the camera and a set of points well known in an object coordinate space. This problem has been extensively treated in the literature and is still opened. Many solutions already exist. All these approaches consider only common planar camera. We propose, with a new formulation, to extend this problem to non linear imaging sensors: catadioptric panoramic sensors. The proposed approach permits to get a strictly analytical solution to the perspective-N-point problem usable with this kind of sensors.

This paper deals with the calibration of a cylindric omnidirectional imaging system, based on a rotating 2,048 pixels linear camera which provides high definition panoramas. The two-step algorithm relies on line-segment projections as calibration features, which are sinusoidal curves. We proposed a cylindrical line detection, based on the dual principle of the Hough transform. Moreover, the use of Plucker coordinates introduces some new characteristics in the calibration process. This kind of formalism allow a linearization of the cylindrical projection, which is non-linear in the usual way. Results obtained from this first step are used to evaluate one of the intrinsics, the other one being determined by a linear criterion minimization in the dual space, i.e. the sines magnitudes space.

La geometrie epipolaire des paires de capteurs omnidirectionnels est souvent difficile a exprimer analytiquement. Nous proposons un algorithme pour estimer numeriquement les courbes epipolaires des paires de capteurs omnidirectionnels. Cet algorithme n'est toutefois pas limite a ce type de capteur et fonctionne, par exemple, avec une combinaison d'un capteur panoramique et d'une camera classique. Bien que la charge de calcul requise soit lourde, cet algorithme a l'avantage de fonctionner avec tous les types de capteurs, si la paire de capteurs est completement calibree (tous parametres determines). En particulier l'algorithme est applicable pour les capteurs catadioptriques ne respectant pas la contrainte du foyer de projection unique.

This paper deals with the calibration of a cylindrical omnidirectional imaging system, based on a rotating 2048 pixels linear camera which provides high definition panoramas. The two-step algorithm relies on line segment projections as calibration features, which are sinusoidal curves. A cylindrical image line detection algorithm is proposed, based on the dual principle of the Hough transform. Moreover, the use of Plucker coordinates introduces some new characteristics in the calibration process. This kind of formalism allows a linearization of the cylindrical projection, which is non-linear in the usual way. Results obtained from this first step are used to evaluate one of the intrinsic parameters, the other one being determined by a linear criterion minimization in the dual space, ie the sine magnitude space.

and key words The epipolar geometry of couples of omnidirectional sensors is often difficult to express analytically. We propose an algorithm to estimate numerically epipolar curves from omnidirectional pairs of stereovision. This algorithm is not limited to this type of sensors and works, for example, with a combination of a panoramic sensor and a traditional camera. Although the load of calculation necessary for this algorithm is heavy, it works with every kind of sensor (provided that the stereovision pair is completely calibrated) especially with sensor that do not respect the single viewpoint constraint. Epipolar geometry, Stereovision, Omnidirectional sensors.

and key words This paper deals with the calibration of a cylindrical omnidirectional imaging system, based on a rotating 2048 pixels linear camera which provides high definition panoramas. The two-step algorithm relies on line segment projections as calibration features, which are sinusoidal curves. A cylindrical image line detection algorithm is proposed, based on the dual principle of the Hough transform. Moreover, the use of Plucker coordinates introduces some new characteristics in the calibration process. This kind of formalism allows a linearization of the cylindrical projection, which is non-linear in the usual way. Results obtained from this first step are used to evaluate one of the intrinsic parameters, the other one being determined by a linear criterion minimization in the dual space, ie the sine magnitude space. Non-Linear Calibration, Omnidirectional Imaging System, Plucker, Coordinates, Hough Transform.

This paper deals with the calibration of a cylindric omnidirectional imaging system, based on a rotating 2048 pixels linear camera, which provides high definition panoramas. The two-steps algorithm relies on line segments projections, which are sinusoidal curves. Moreover, Plucker formalism makes it possible to linearize the full cylindrical model. Calibration data are provided by a sinusoid detection step, based on a "dual" Hough transform. Results obtained from the first process are used to evaluate one of the intrinsic, the other one being determined by a linear minimization criterion in the dual space.

The problem of estimating the distance between a unique image sensor and a well known pattern composed by N points (The Perspective-N-Points problem) has been extensively treated and many solutions exist. By the use of a new set of equations, we get, whereas most of existing solutions, a strictly analytical solution. This solution first request few and constant CPU time which allow us to use it in real time applications and second is usable with nonlinear sensor (particularly omnidirectional sensor).

The F180 RoboCup league relies on a single camera mounted on top of the field. It is of great importance to use an adapted calibration method to locate robots. Most of the methods used are developped for specific application where 3D is required. This paper presents a new calibration method specially developped for the F180 league geometry, allowing the determination of the camera pose parameters and the correction of the parallax in the image due to different heights of observed robots. This method needs one calibration plane that also could be used for correcting optical distortions introduced by the lens.

A new efficient matching algorithm dedicated to catadioptric sensors is proposed in this paper. The presented approach is designed to overcome the varying resolution of the mirror. The aim of this work is to provide a matcher that gives reliable results similar to the ones obtained by classical operators on planar projection images. The matching is based on a dynamical size windows extraction, computed from the viewing angular aperture of the neighborhood around the points of interest. An angular scaling of this angular aperture provides a certain number of different neighborhood resolution around the same considered point. A combinatory cost method is introduced in order to determine the best match between the different angular neighborhood patches of two interest points. Results are presented on sparse matched corner points, that can be used to estimate the epipolar geometry of the scene in order to provide a dense 3D map of the observed environment.

Camera calibration is a very important issue in computer vision each time extracting metrics from images is needed. The F180 camera league offers an interesting problem to solve. Camera calibration is needed to locate robots on the field with a very high precision. This paper presents a method specially created to easely calibrate a camera for the F180 league. The method is easy to use and implement, even for people not familiar with computer vision. It gives very acurate and efficient results.

The stereoscopic sensor presented in the last chapter is different from other systems due to the use of linear CCDs. The main problem is to find matching techniques to obtain a 3D reconstruction of the observed scenes directly from two linear images.

Nous presentons ici la geometrie d'un capteur panoramique tournant, ainsi que la generalisation du concept de contrainte epipolaire sur des images de type cylindrique. Differents procedes de minimisation sont utilises pour retrouver certains parametres experimentaux. Cette contrainte permet ainsi d'optimiser de l'etape de recherche de points apparies afin de reconstituer la configuration tridimensionnelle de la scene observee a partir d'images cylindriques haute definition.