Pyramid calibration and correspondence analysis in binocular vision

The authors present a technique for 3-D camera calibration using a pyramid calibration frame. The focus is on camera calibration and correspondence analysis. A rule is presented to set up precorrespondence automatically before the calibration process begins. To strengthen practicability, a careful discussion on the change of the parameters of a calibration matrix when the camera moves is provided. Five constraints are integrated into a feature-point-based matching algorithm: epipolar, common view, uniqueness, compatibility, and model. The calibration process equips the stereo system with common view and model constraints. Uniqueness and compatibility constraints have their origins in Marr´s (1982) stereo theory. Based on the calibration process, a stereo algorithm synthesizes general constraints and is applied to a real situation. The experimental results have shown its practical use in a general situation