A discriminant method of single-optical-axis omnidirectional vision system

Abstract

Generally, due to the instrumental error of omnidirectional camera, it is very difficult to satisfy the standard single-optical-axis request. Thus, it is necessary to evaluate whether a single-optical-axis camera lens is aligned or has tangent distortion. In this paper, we propose a discriminant function of single-optical-axis omnidirectional vision system based on checkerboard, which is only related to the image points and does not involve any camera parameters. Firstly, under single-optical-axis omnidirectional camera, the geometric invariance among image points of collinear and equidistant space points is derived. Next, based on the derived geometric invariance in a single image, we construct the object function to discriminate the standard single-optical-axis omnidirectional camera. Finally, a discriminant method of single-optical-axis omnidirectional vision system is proposed based on the checkerboard. Experimental results on both simulated and real data have demonstrated the usefulness and effectiveness of our method.