A stereo system for 3-D measurements in robot workspaces

The authors describe a stereo system for robot vision applications. The geometry of the system is designed to make use of a simple and efficient camera calibration procedure that requires only three world points of known coordinates. The system introduces new variables that relate the output digital image directly to the viewed 3-D object. In addition, it combines the advantages of the area based and feature-based matching approaches. Experimental results indicate that the method is capable of extracting 3-D information of real objects with satisfactory accuracy.<
>