Mathematical models for camera calibration and application to the PANCAM

Recent years have seen an international growing interest in lunar exploration. In 2019, the European Space Agency (ESA) opened a campaign asking for innovative proposals aimed at the exploration, documentation and 3D mapping of lunar volcanic cavities. One of the selected projects was the DAEDALUS Sphere (Descent And Exploration in Deep Autonomy of Lava Underground Structures): a spherical robot which hosts four bifocal panoramic lenses, the so-called BLPs, and is able to perform an immersive stereoscopic map of lunar caves. Due to the strong optical anamorphism inherent to the PANCAM lens, which captures the panoramic field, it is necessary to utilize image dewarping algorithms. A preliminary yet crucial step in the dewarping process is camera calibration, which consists in computing the parameters of a camera. In this thesis we analyse mathematical models for camera calibration. We start with Zhang’s algorithm, which follows the simple pinhole projection model and then introduces lens distortion. In order to better describe very-wide angle cameras, we present the model and toolbox proposed by Scaramuzza. To test both the model and the software, we use the PANCAM to acquire some pictures of a checkerboard, which is the calibration pattern that allows to obtain the parameters of the camera. In particular, we try to understand if Scaramuzza’s model accurately describes the hyper-hemispheric part of the PANCAM lens.