Browse Prior Art Database

Pattern Recognition for Automatic Visual Inspection Disclosure Number: IPCOM000131562D
Original Publication Date: 1982-Dec-01
Included in the Prior Art Database: 2005-Nov-11
Document File: 10 page(s) / 39K

Publishing Venue

Software Patent Institute

Related People

Kin2-sun Fu: AUTHOR [+3]


Application of more sophisticated pattern recognition techniques to difficult automatic inspection problems -- IC chips, for example -- is now technologically feasible.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 10% of the total text.

Page 1 of 10


This record contains textual material that is copyright ©; 1982 by the Institute of Electrical and Electronics Engineers, Inc. All rights reserved. Contact the IEEE Computer Society (714-821-8380) for copies of the complete work that was the source of this textual material and for all use beyond that as a record from the SPI Database.

Pattern Recognition for Automatic Visual Inspection

Kin2-sun Fu.

Purdue Universitv

Application of more sophisticated pattern recognition techniques to difficult automatic inspection problems -- IC chips, for example -- is now technologically feasible.

The many methods proposed for designing a pattern recognition system can be grouped into three major categories: the template matching approach, the decision- theoretic or discriminant approach,2~'0 and the syntactic and structural approach.~'~'3 From a more general viewpoint, these approaches can be discussed in terms of pattern representation and decision making (see Figure 1). Primarily, the pattern representation subproblem involves selection of the representation, and the decision-making subproblem involves selection of a decision criterion or similarity measure. Other approaches include problem- solving models,'4 category theory,' 5 and relation theory.'fi

Template matching.

In the template-matching approach, a set of templates or prototypes, one for each pattern class, is stored in the machine. The input pattern is matched or compared with the template of each class, and the classification is based on a preselected matching criterion or similarity measure (e.g., correlation). In other words, if the input pattern matches the template of the ith pattern class better then it matches any other template, then the input pattern is classified as being from the ith pattern class. For machine simplicity, input patterns and the templates are usually represented in their raw data form, and the decision-making process simply matches the unknown input to each template.

The template-matching approach has been used in printed- character recognizers and bank check readers. The main difficulties with this approach lie in selecting a good template for each pattern class and in defining an appropriate matching criterion, especially when large variations and distortions are expected in the patterns under study. Recently, the use of flexible templatematching or rubber mask techniques has been proposed. "

Decision theoretic.

In the decision-theoretic approach, a pattern is represented by a set of N features or an N- dimensional feature vector, and the decision-making process is based on a similarity measure, which, in turn, is expressed in terms of a distance measure or a discriminant function. Statistical and fuzzy-set methods have been proposed'8 for considering noise and distortions. The characterization of each pattern class could be in terms of an N-dimensional class-conditi...