Calibration and orientation of cameras in computer vision. Workshop at the 17th congress of the ISPRS (International Society of Photogrammetry and Remote Sensing), Washington, DC, USA, July 1992 (Q5942572)

This book was conceived during the Workshop ``Calibration and Orientation of Cameras in Computer Vision'' at the 17th Congress of the ISPRS (International Society of Photogrammetry and Remote Sensing), in July 1992 in Washington, D.C. The goal of this workshop was to bring photogrammetry and computer vision experts together in order to exchange ideas, concepts and approaches in camera calibration and orientation. Chapter 2 discusses minimum solutions of various orientation tasks of photogrammetry and computer vision. This review is based on examining the latest developments and also by looking back at the early beginnings of orientation procedures of photogrammetry to demonstrate that there are already answers to a lot of the conceptual questions of today. The first part of the chapter is devoted to the historical roots of orientation approaches and to a comparison of two very popular camera models: The first is based on projective collineation and the second on perspective collineation. In the next part of this chapter a systematic overview on standard orientation tasks of photogrammetry and computer vision, from single image to multiple image orientation, is given. Orientation of cameras with minimum or redundant information is the first step in 3D-scene analysis. The difficulty of this task lies in the lack of generic and robust procedures for geometric reasoning, calibration and especially orientation. Chapter 3 collects available tools from statistics, especially for the diagnosis of data and design and for coping with outliers using robust estimation techniques. Chapter 4 reviews modern analytical camera calibration techniques employed by photogrammetrists and discusses the potential of these methods for video cameras (principally CCD cameras). Chapter 5 describes a method for calibrating cameras including radial lens distortion, by using known points such as those measured from a calibration fixture. The distortion terms are relative to the optical axis, which is included in the model so that it does not have to be orthogonal to the image sensor plane. A priori standard deviations can be used to apply weight to zero values for the distortion terms and to zero difference between the optical axis and the perpendicular to the sensor plane, so that the solution for these is well determined when there is insufficient information in the calibration data. Camera systems with adjustable lenses are inherently more useful than those with fixed lenses. Adjustable lenses enable us to produce better images by matching the camera's sensing characteristics to the conditions in a scene. They also allow us to make measurements by nothing how the scene's image changes as the lens settings are varied. The reason adjustable lenses are not more commonly used in machine vision is that they are difficult to model. In Chapter 6 the authors demonstrate how two commonly used camera models. The method has proved to be one of the most powerful calibration techniques. If used in the context of a general bundle solution it provides for object space coordinates or object features, camera exterior and interior orientation parameters, and models other systematic errors as well. Therefore, because of its flexibility, it may be used in stereo, multi-frame systems, egomotion computations, etc. Chapter 7 gives a brief introduction to the principle of self-calibration, emphasizes some of the problems which are associated with it, and demonstrates with practical data to what extent geometry and network design will influence the determinability of the self-calibration parameters. Finally, a system test will show the high accuracy performance of self-calibrating CCD camera systems.