Panoramic sensor calibration using computational projective geometry

This paper presents a new method of camera calibration for a panoramic stereovision sensor based on a cylindrical projective geometry. The multidirectional sensor allows us to get peripheral information about unknown environments, the architecture simplifies considerably computation. The device needs no calculation of epipolar lines. The calibration method derives directly from the benefits of the chosen architecture of the sensor and is not based on computation of vanishing points like former methods, which can only be an approximation and introduces uncertainty in the calibration process. Our calibration is based on the theory of projective vectors but uses simple geometry based on colinearity and orthogonality of vectors existing in the scene.