ABSTRACT
Many robotic applications benefit from a wide field of view. Wide fields of view facilitate tasks such as navigation and obstacle detection. A number of solutions have been proposed in the literature to enable the development of computer vision systems with panoramic fields of view. In most cases extended fields of view imply resolution tradeoffs. Most of these solutions also imply that the mathematical models of the systems are more complex and highly non-linear. Another important aspect for robotic applications is the ability to control and change the image acquisition parameters. In this paper we discuss aspects related to two of the commonly used solutions: active vision systems and catadioptric vision systems. We characterize several technical aspects relevant to robotic applications and discuss aspects related to their control and calibration.
