PATTERN RECOGNITION APPARATUS

• US 20120230592A1
• Filed: 02/09/2010
• Published: 09/13/2012
• Est. Priority Date: 02/10/2009
• Status: Active Grant

• Current Assignee: Osaka Prefcture University
• Original Assignee: Osaka Prefcture University
• Inventors: Iwamura, Masakazu, Kise, Koichi
• Granted Patent: US 8,422,793 B2
• Time in Patent Office: Days
• US Class Current: 382/201
• CPC Class Codes: G06F 16/5838 using colour
• CPC Class Codes: G06T 7/66 of image moments or centre of gravity
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/1478 of characters or characters lines
• CPC Class Codes: G06V 30/268 Lexical context

Abstract

A pattern recognition apparatus including: an extracting section for extracting, from a query image that is composed of at least one piece of pattern component and previously undergoes a geometric transformation, the pattern component; a feature acquiring section for acquiring a geometric invariant feature of the pattern component as a query feature, the query feature being represented by at least three feature points including first, second and third feature points, each feature point locating on the pattern component and being retrieved from the pattern component based on a predetermined rule; a comparing section for comparing the query feature with a plurality of reference features, each reference feature representing different reference patterns prepared as candidates for pattern recognition; and a pattern determination section for determining, as a recognition result, a specific reference pattern out of the candidates based on a similarity of features therebetween and, wherein: each reference feature is represented using feature points retrieved from each reference pattern based on the same rule as that of the query feature, and based on the predetermined rule, a position of the first feature point is specified out of points which locate on the pattern component and are invariant to the geometric transformation, a position of the second feature point is specified using a characteristic regarding a shape of the pattern component, the characteristic being invariant to the geometric transformation, and a position of the third feature point is specified from a predetermined value being invariant to the geometric transformation and from the specified positions of the first and second feature points.

Claims

1. A pattern recognition apparatus comprising:

an extracting section for extracting, from a query image that is composed of at least one piece of pattern component and previously undergoes a geometric transformation, the pattern component;

a feature acquiring section for acquiring a geometric invariant feature of the pattern component as a query feature, the query feature being represented by at least three feature points including first, second and third feature points, each feature point locating on the pattern component and being retrieved from the pattern component based on a predetermined rule;

a comparing section for comparing the query feature with a plurality of reference features, each reference feature representing different reference patterns prepared as candidates for pattern recognition; and

a pattern determination section for determining, as a recognition result, a specific reference pattern out of the candidates based on a similarity of features therebetween and, wherein;

each reference feature is represented using feature points retrieved from each reference pattern based on the same rule as that of the query feature, and

based on the predetermined rule, a position of the first feature point is specified out of points which locate on the pattern component and are invariant to the geometric transformation, a position of the second feature point is specified using a characteristic regarding a shape of the pattern component, the characteristic being invariant to the geometric transformation, and a position of the third feature point is specified from a predetermined value being invariant to the geometric transformation and from the specified positions of the first and second feature points.

2. A pattern recognition apparatus according to claim 1, wherein:

the position of the first feature point is specified out of pixels on a contour of the pattern components.

3. A pattern recognition apparatus according to claim 1, wherein:

the characteristic is characterized in that a center of gravity is invariant to affine transformation as a kind of the geometric transformation, and

the position of the second feature point is specified as a center of gravity of the pattern components, using the characteristic.

4. A pattern recognition apparatus according to claim 1, wherein:

the characteristic is characterized in that an area ratio is invariant to affine transformation as a kind of the geometric transformation, and

the position of the third feature point is specified from a contour of the pattern component and based on a predetermined value of an area ratio of the area of the pattern components to the area of a triangle defined by apexes corresponding to the first, second and third feature points, using the characteristic.

5. A pattern recognition apparatus according to claim 1, wherein:

the feature acquiring section acquires the geometric invariant feature using a geometric invariant coordinate system that is spanned by two linearly independent vectors, each vector connecting two of the acquired three feature points.

6. A pattern recognition apparatus according to claim 1, wherein:

the comparing section makes comparisons of the query feature with the reference features that are related to the corresponding reference patterns and stored in a hash table, wherein;

the hash table has a plurality of bins;

each reference feature is preliminarily classified and stored in one of the bins determined through a calculation of a predetermined hash function from the reference feature; and

the comparing section refers an appropriate bin for each comparison using an index that is obtained through a calculation of the hash function from the acquired query feature.

7. A pattern recognition apparatus according to claim 6, wherein:

each reference feature is stored in the bin with related to coordinate data of the acquired three feature points and to an identifier of the reference pattern corresponding to the reference feature; and

the pattern determination section estimates a pose of the query image based on matching of each coordinate data related to the query feature to that related to each reference feature stored in the referred bin and based on a majority rule on the matching.

8. A pattern recognition apparatus according to claim 1, wherein:

the pattern determination section has a discrete pattern table storing at least one group of discrete patterns, each discrete pattern corresponding to one of the reference patterns and the group of discrete patterns providing one recognition result;

refers the discrete pattern table to decide whether a correspondence between the determined specific reference pattern out of the candidates and one of the discrete patterns of the group exists or not; and

determines a recognition result as that being provided by the group of which the corresponding discrete pattern to the determined specific reference pattern belongs to the group when the correspondence exists and a correspondence of every other discrete pattern of the group is previously exists.

9. A pattern recognition apparatus according to claim 8, wherein:

the discrete pattern table stores a relative position of one discrete pattern to another discrete pattern in the group, and

the pattern determination section determines a recognition result when another determined specific reference pattern locates at a position specified by the stored relative position of the discrete pattern corresponding to one determined specific reference pattern.

10. A pattern recognition apparatus according to claim 1, wherein

the query image includes a word pattern composed of plural characters, and

the pattern recognition apparatus further includes:

a word candidate determining section for obtaining the shortest route in which each character recognized by the pattern determining section is traced once, and for determining an order and inverse order of the obtained route as candidates of the word;

a rotation angle determining section for obtaining a rotation angle of each character in a predetermined direction of the query image; and

a reading-order determining section in which a difference in the rotation angle between two characters adjacent to each other along the order or inverse order of the route is specified as a first evaluation index, the rotation angle that the first character should take is estimated based upon the direction toward a second character, which is adjacent to a first character that is defined to be either one end of each candidate, and a predetermined rule involved with a reading direction, and the difference between the estimated rotation angle and the rotation angle of the first character determined by the rotation angle determining section is specified as a second evaluation index, wherein the candidate that minimizes the first and second evaluation indexes is selected in order to determine a reading order of characters constituting the word.