Fuzzy cross-correlation algorithm for classifying high tolerance engine components

In this paper the structure of a vision-based object recognition system for classifying high tolerance engine parts is presented. This system could be divided into four basic levels: (1) data acquisition operation; (2) feature extraction algorithm; (3) classifier design; and (4) decision-making strategy. A mechano-optical arrangement in conjunction with a frame grabber and a software program are used to generate and capture the images of the components. Another software program based on the circular Hough transform algorithm is applied to the images for extracting the centre coordinates of the side profiles of the image as the features of the component. Then, a bell-shape membership function is generated using the extracted data. A unique feature matrix for each component could be defined from the corresponding membership functions. For object classification, a fuzzy cross-correlation algorithm is used for comparing the testing feature matrix with the reference feature matrix and obtained an aggregative relationship between the components.