Close range 3-D position sensing using stereo vision and point-of-correspondence through mathematical morphology

Mohamed T. Boraie, University of Windsor


In this thesis, a procedure for calibrating 2 CCD cameras for the purpose of close-range position sensing will be introduced. The procedure results in an over-determined set of linear algebraic equations that can be utilized in determining the coordinate position of some identifiable feature in the scene relative to any set of orthogonal axes.The proposed approach relaxes the geometrical constraint required by most of the available procedures. The performance will then be evaluated and both accuracy and repeatability will be examined.For the purpose of determining the feature position in a multi-feature image, the point-of-correspondence problem will be discussed and a procedure for the point-of correspondence in stereo vision will be introduced. The procedure utilizes mathematical morphology as an image processing tool in finding edge-strength images and in determining the location of the point-of-correspondence.Finally, introducing lens distortion parameters to the calibration process will be presented and the effect of each term will then be illustrated.