This dissertation presents several aspects of the use of vision for dynamic systems and control. Vision-based feedback systems have been a central problem in computer vision and control for over two decades. However, it is only in recent years that the variations of images over time have begun to be used for control. Motivated from a visual servoing task of an omnidirectional vehicle, the problem of iterative visual servoing with an uncalibrated camera is studied. Then, lens distortion modeling and correction is addressed, where a series of experimental results are given that can serve as a general guidance for evaluation of the lens distortion correction alone. Next, recent work on perspective dynamic systems (PDS), which provides a general framework to discuss vision problems, such as motion and depth estimation, using concepts and methods from controls is considered. Focused on the estimation problem of a PDS, a linear approximation-based nonlinear observer is proposed. The final section of work introduces the idea of iterative learning control of a PDS system and presents preliminary results on this problem.
展开▼