Fully Automatic Camera Calibration Using Self- Identifying Calibration Targets

Determining camera calibration parameters is an essential step in most computer vision endeavors; it is a time-consuming task despite the availability of calibration algorithms and software. A set of point correspondences between points on the calibration target and the camera image(s) must be found, usually a manual or manually guided process. Two commonly used calibration tools are implementations of Zhang’s (OpenCV) and Tsai’s algorithms, however, these assume that the correspondences are already found. A system is presented which allows a camera to be calibrated merely by passing it in front of a panel of self-identifying patterns. This calibration scheme uses an array of ARTag fiducial markers which are detected with a high degree of confidence, each detected marker provides one or four correspondence points. The user prints out the ARTag array and moves the camera relative to the pattern, the set of correspondences is automatically determined for each camera frame, and input to the OpenCV calibration code. Experiments were performed calibrating several cameras in a short period of time with no manual intervention. This system was implemented in a program for co-planar calibration, results are shown from several calibration tests with different cameras. Experiments were performed comparing using either the four ARTag marker corners or a single marker center as correspondences, and the number of image frames necessary to calibrate a camera was explored. This ARTag based calibration system was compared to one using the OpenCV grid finder cvFindChessBoardCornerGuesses() function which also finds correspondences automatically. We show how our new ARTag based system more robustly finds the calibration pattern and how it provides more accurate intrinsic camera parameters.