Method for automatically inspecting parts utilizing machine vision and

Abstract

A method for automatically inspecting parts utilizing machine vision and system utilizing same are disclosed wherein a large variety of parts are inspected with a single optical setup at a vision station. Cameras and their respective illuminating lamps are located at the vision station. The inspection task requires multiple scenes or views to inspect part features such as inner and outer diameters, part height and concentricity. Initially, the system is taught which features of the parts are to be inspected by generating reference data. The reference data includes predetermined feature-detecting algorithms, model data defining typical features and limit data which is related to the allowable deviation of the features. The parts to be inspected are conveyed past the vision station. The signals from each of the active cameras are digitized and then processed by its own group of pipelined processors. Each group of processors is programmed with the reference data to process the gray-scale digitized data to determine whether the features appearing in a particular view of the part are acceptable. A coordination processor coordinates the operation of the groups of processors and is responsive to the transport signal. A general purpose computer is responsive to the part sense signals and feature-acceptance signals from each of the groups of processors to provide reject signals for use by an actuator to remove unacceptable parts from the path of conveyance.