A 3-D vision algorithm for robot applications

A three-dimensional (3-D) vision algorithm using an off-shelf system with two cameras for robot applications is presented. The 3-D vision algorithm can be trained to recognize a 3-D object from a scene and input the center of location, the orientation, and the dimensions of the object to a robot control system. Taking advantage of the hard-wired fst blob analysis algorithm and the hard-wired fast correlation search algorithm from an off-shelf computer vision system, an object can be recognized, located, and measured in seconds. This model-based 3-D vision algorithm also features automatic calibration, adaptive histogram analysis to reduce light interference from other sources, and a menu-driven model database.<>