Pattern classification method and apparatus thereof, and computer readable storage medium

• US 20020143719A1
• Filed: 05/16/2002
• Published: 10/03/2002
• Est. Priority Date: 09/25/2000
• Status: Active Grant

• Current Assignee: Olympus Optical Corporation Limited (Olympus Corporation)
• Original Assignee: Olympus Optical Corporation Limited (Olympus Corporation)
• Inventors: Yoshihara, Takafumi
• Granted Patent: US 6,931,386 B2
• Time in Patent Office: Days
• US Class Current: 706/20
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06F 18/22 Matching criteria, e.g. proximity measures
• CPC Class Codes: G06T 2207/30148 Semiconductor; IC; Wafer
• CPC Class Codes: G06T 7/0004 Industrial image inspection

Abstract

A pattern classification method extracts a quantity of pattern characteristics from a pattern classification target, registers/sets in a pattern dictionary classification category information corresponding to the quantity of pattern characteristics, performs pattern classification based on the quantity of pattern characteristics and a Mahalanobis dictionary registered in the pattern dictionary, and creates the Mahalanobis dictionary by effecting optimization processing of the pattern dictionary based on the classification category name and the quantity of pattern characteristics.

Claims

1. A pattern classification method comprising:

extracting a quantity of pattern characteristics from a pattern classification target;

registering and setting classification category information corresponding to said quantity of pattern characteristics in a pattern dictionary; and

performing pattern classification based on said quantity of pattern characteristics and a Mahalanobis dictionary registered in said pattern dictionary,

wherein said Mahalanobis dictionary is created by performing optimization processing of said pattern dictionary based on said classification name and said quantity of pattern characteristics stored in said pattern dictionary.

2. The pattern classification method according to claim 1, wherein said Mahalanobis dictionary includes an optimization candidate table having a utilized characteristic quantity weighting matrix and a predicted classification ratio or utilized characteristic quantity information and a predicted classification ratio as elements.

3. A pattern classification apparatus comprising:

a pattern dictionary which stores a quantity of pattern characteristics extracted from a pattern classification target and classification category information corresponding to said quantity of pattern characteristic;

means for performing pattern classification based on said quantity of pattern characteristics and a Mahalanobis dictionary registered in said pattern dictionary; and

pattern dictionary optimizing means for creating said Mahalanobis dictionary by performing optimization processing based on said classification category name and said quantity of pattern characteristics stored in said pattern dictionary.

6. The pattern classification apparatus according to any one of claims 3, 4 and 5, wherein said Mahalanobis dictionary includes an optimization candidate table.

4. A pattern classification apparatus comprising:

pattern fetching means for fetching a pattern classification target;

pattern area cutting means for cutting out a specific pattern area from said pattern classification target supplied from said pattern fetching means;

classification characteristic quantity extracting means for calculating a quantity of pattern characteristics useful for pattern classification from said specific pattern area supplied from said pattern area cutting means;

dictionary registering means for registering and setting in a pattern dictionary said quantity of pattern characteristics calculated by said classification characteristic quantity extracting means and classification category information corresponding to said quantity of pattern characteristics;

characteristic quantity classifying means for performing pattern classification based on said quantity of pattern characteristics calculated by said classification characteristic quantity extracting means and a Mahalanobis dictionary registered in said pattern dictionary; and

pattern dictionary optimizing means for creating said Mahalanobis dictionary by performing optimization processing based on said classification category name and said pattern characteristic quantity stored in said pattern dictionary.

5. The pattern classification apparatus according to claim 4, wherein said pattern dictionary optimizing means has a function to execute processing for searching a pattern characteristic quantity weighting matrix which weights said quantity of pattern characteristics utilized for calculation of said Mahalanobis distance by using said quantity of pattern characteristics and said classification category name of a registered pattern in said pattern dictionary.

7. A computer readable storage medium storing a program of a pattern classification method comprising:

registering and setting in a pattern dictionary a quantity of pattern characteristics extracted from a pattern classification target and classification category information corresponding to said quantity of pattern characteristics;

creating a Mahalanobis dictionary by performing optimization processing based on said classification category name and said quantity of pattern characteristics stored in said pattern dictionary; and

performing pattern classification based on said optimized quantity of pattern characteristics and said Mahalanobis dictionary registered in said pattern dictionary.