Capturing Optimal Image Networks for Planar Camera Calibration

Abstract

This paper details a novel approach to specifying the optimal pose of planar targets in camera calibration that both reduces the number of images required, and improves the parameter estimates. This is accomplished within a semi supervised trategy where virtual images of planar calibration targets are generated and displayed. These virtual targets are then replicated by the user to generate an image network with optimal geometry for the recovery of the camera parameters. Optimal planar pose is specified by enforcing maximum independence within the calibration constraints offered by each image within the network. This solution space is further refined to ensure that the generated target pose is suitable for easy acquisition and subsequent feature extraction processes. The results on simulated and real data demonstrate that proper consideration of image network geometry directly leads to more accurate camera parameter estimates.