Catadioptric Stereo with Planar Mirrors: Multiple-view Geometry and Camera Localization [chapter]

Gian Luca Mariottini, Stefano Scheggi, Fabio Morbidi, Domenico Prattichizzo
2010 Lecture notes in control and information sciences  
Planar catadioptric stereo (PCS) vision sensors consist of a pinhole camera and two or more planar mirrors. PCS systems have recently received an increasing attention in computer vision and have a promising applicability in robotics, since the use of mirrors allows to obtain a stereo view without the need of exact multi-camera synchronization and stereo calibration. The paper presents a rigorous analytical treatment of the imaging geometry of PCS sensors and introduce new multiple view
more » ... s that are instrumental in addressing the camera localization problem. Original results on mirror calibration are also provided. Extensive simulation and real-data experiments conducted with an eye-in-hand robot illustrate the theory and show the effectiveness of the proposed designs. Stereoscopic vision has been playing a key role in many fields, such as, e.g, 3-D imaging for entertainment and mapping, medicine, industrial robotics, augmented reality [13, 11] . Standard stereo sensors are composed of two coupled pinhole cameras and are generally expensive and difficult to calibrate. In addition, the limited field of view considerably limits their application range. To alleviate this problem, several works in the robotics and computer vision literature have recently proposed the use of catadioptric vision sensors [18, 4, 21, 15] , which, combining
doi:10.1007/978-1-84996-089-2_1 fatcat:uezhaq3qvbdmrk767xy6556n5y