Calibration of a stereoscopic system without traditional distortion models

Abstract

In the field of photography measurement, parameterized models are usually established to account for lens distortions, such as the radial, tangential, affine, and nonorthogonality deformations. However, all of these models are the approximations of the realistic model of lenses, instead since some distortions cannot be parameterized accurately. This restricts the improvement of measuring accuracy. Moreover, the nonlinear minimization, which has been widely used with the pin-hole model and lens distortion model coupled, always converges to local solution because of the correlation of the parameters in the two models. Several researchers have proposed generic nonparametric idea, which can be applied equally well to all types of cameras, but the accuracy cannot meet the requirement of close-range photogrammetry. So an optical calibration method based on nonparametric ideas is proposed to find the mapping between incoming scene rays and image points, and subpixel image processing was used toposition the image points. It is applicable to a central (single viewpoint) camera equipped with any lenses. This method is applied to the stereoscopic system, and the results show a good measuring accuracy. © The Authors.