A rapid and automatic method for camera calibration based on LCD circle pattern

Camera calibration based on plane pattern is widely used for its convenience, but current methods to detect patterns and point correspondences are mainly manual or interactive, and the pattern commonly used is a printed chessboard image sticking on plane plate. The paper proposes a novel method to calibrate the camera accurately in a rapid and automatic way. It uses a circle pattern displayed on a LCD screen to avoid the error caused by non-planar calibration object. A moment-based algorithm is used to detect the circle projection center automatically and an algorithm based on topological structure is proposed to obtain an initial point correspondence set. Finally, a subset of the initial point correspondences is used to estimate homography by RANSAC and wrongly detected blob centers are eliminated. Our experiment shows that our method is convenient and rapid to implement without any manual clicking on the images and the calibration results are accurate with a re-projection error of less than 0.6 pixels.