Industrial inspection employing a three dimensional vision system and a neural network classifier

An automatic inspection system for manufactured parts employing a 3D machine vision system and associated software for part identification and dimensional inspection is described. The machine vision module collects a range image of accurate data from the part surface employing a structured light approach. In order to measure specific surface parameters, the entire part data set is decomposed into its constituent surface patches. A neural network classifier is employed to recognise each part from its range data set and also to classify a specific surface patch, on a part, from the overall set of part surface patches. The output of the neural network classifier is presented to a database of part information which is created off-line. The performance of the system has been tested by experimenting on real range data.