Abstract:

In recent dozen years,computer vision becomes more popular,in which omnidirectional camera has a larger field of view (FOV) and widely been used in many fields,such as:robot navigation,visual surveillance,virtual reality, threedimensional reconstruction, and so on.Camera calibration is an essential step to obtain threedimensional geometric information from a twodimensional image.Camera calibration results directly decide the results of 3-D reconstruction and other computer vision application effect.Hence,the study of such camera calibration method has important theoretical significance and practical applications.The paper classifies these methods from 2000 to 2012,by means of the difference calibration blocks,into five categories:calibration based on line;calibration based on two-dimensional calibration block;calibration based on three-dimensional point;calibration based on balls;selfcalibration. And these methods’advantages and disadvantages are analyzed.
    

Key words: spherical model;catadioptric;camera calibration