Robot manipulation of 3D cylindrical objects with a robot-mounted 2D vision camera

Abstract

Position measurements of objects (or parts) have always been one of the most important requirements in industrial robot applications. Machine vision has been one of the promising solutions on detecting objects for enhanced robot intelligence. This paper addresses the common problems and solutions related to developing vision-guided industrial robot applications with the use of two-dimensional (2D) vision data for detecting three-dimensional (3D) cylindrical objects. The discussions include the robot reference frames and their transformations, 2D camera calibration and view positions, vision measurements, and robot programming for vision-guided robot motions and operations. Results of a 2D vision-guided FAUNC robot project are used to demonstrate solutions of the method.