Abstract
Conference Title: 2016 IEEE National Aerospace and Electronics Conference (NAECON) and Ohio Innovation Summit (OIS) Conference Start Date: 2016, July 25 Conference End Date: 2016, July 29 Conference Location: Dayton, OH, USA Calibration is a fundamental task in computer vision and photogrammetry. One of the most important intrinsic camera parameters is the principal point, which is the intersection between the optical axis and the image plane. The proposed method, which uses simple properties of vanishing points, provides a new technique for accurate identification of the principal point. It is done independently of all the other camera parameters. Checkerboard image corner points are located as saddle points, and the Hough transform is applied to remove spurious points and group them into rows and columns. The vanishing point for the columns lie on a horizon line. A group of images, created by rotating the checkerboard while holding the camera stationary, is processed in this manner to create multiple vanishing points on the same horizon line. A perpendicular line to the horizon is then projected back through the principal point. Repeating this for several image groups, corresponding to different camera orientations, allows for accurately locating the principal point as the intersection of the perpendiculars. Our approach does not need any prior information about the cameras being used, and does not require any manual user interaction. Experiments to evaluate the performance of this approach on real test images indicates that the uncertainty for the principal point location overlaps and is smaller than the region found by Bouget's toolbox.