Modeling and calibration of automated zoom lenses

摘要: Camera systems with automated zoom lenses are inherently more useful than those fixed-parameter lenses. Variable-parameter enable us to produce better images by matching the camera's sensing characteristics conditions in a scene. They also allow make measurements noting how scene's image changes as parameters varied. The reason variable-parameter not commonly used machine vision is that they difficult model for continuous ranges of lens settings. We show this thesis traditional modeling approaches cannot capture complex relationships between control and imaging processes. Furthermore, we demonstrate assumption idealized behavior models can lead significant performance problems color focus ranging. By using strategies were able reduce or eliminate these problems. The principal contribution our research methodology empirically producing accurate camera We developed comprehensive taxonomy property "image center." To effectiveness applied it an "adjustable," perspective-projection based on Tsai's fixed model. calibrated tested two different systems. In both cases operated across average error less 0.14 pixels predicted measured positions features plane. one system range aperture achieved similar results.