Geometric calibration of cameras applying least-squares method

Abstract

To make artificial vision or computer vision possible, it is necessary to solve a series of problems of reasonable complexity. The first problem is the calibration of the cameras which will be used by the artificial vision system. The calibration process requires the, use of well-known patterns with known ℜ³ coordinates, once the basic problem consists of associating a point in ℜ³, tridimensional space, with another point in ℜ², image plane or plane of camera retina, involving optical, electronic, electronic-numerical and mechanical phenomena that contribute with mistakes to the system. To minimize them it is necessary that the artificial vision system be redundant having more equations than variables. The least-squares method was applied to solve this system. A methodology appropriate to the acquisition of the necessary calibration patterns is also presented.