A Linear Self-Calibration Method Based on Active Vision System

Abstract

This paper presented an improved self-calibrate method based on active vision system. The camera was controlled to move directly in a plane and rotated around the X-axis in the camera coordinates in five times or more, then we calibrated the camera by the epipoles in the images. This method improved the accuracy and robust of the epipoles, compared with other method. The theory based on Kruppa equation of self-calibration of a moving camera was put forward by Maybank and Faugeras(1). Then, a lot of self-calibration methods were put forward. All these methods need to solve non-linear equations which are hard to work out. Ma gave a linear self-calibration method based on active vision system(2). The camera moved twice in the two plumb planes, and then the FOE (Focus of Expansion) was used to create linear equations and solve the internal camera parameters. The method carried out self- calibration by assuming that the skew factors are equal to zero. Those methods mentioned above don't work in the 5 factors model, and this paper introduced a method which can solve the internal camera parameters in 5 factors model without the plane information of the 3D space.