Abstract: Most of the existing algorithms for camera self-calibration have following problems: non-linear optimization of Kruppa equations, and not robust In order to solve these problems, a multi-step self-calibration method was proposed. This method begins with getting the fundamental matrix between images by 8-point algorithm. I}he candidate matching images arc acquired by taking two pictures of one scene using different focal length. We established the constraint correspondence between feature points of two images. The coordinate of principle point was calculated by using least square method. Finally the genetic algorithm was adopted to optimize the scale factor of the Kruppa equations and accomplishing the task of camera calibration. Experiment results indicate that the proposed method can effectively improve the accuracy and robustness of self-calibration method based on Kruppa equations.

Key words: Self-calibration,Kruppa equations,Genctic algorithm,Fundamcntal matrix