Recognition device, recognition method, and computer program product

• US 10,438,097 B2
• Filed: 09/20/2017
• Issued: 10/08/2019
• Est. Priority Date: 05/11/2015
• Status: Active Grant

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation), Toshiba Digital Solutions Corporation (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Suzuki, Tomohisa
• Primary Examiner(s): Allison, Andrae S
• Application Number:
  US15/709,779
• Publication Number:
  US 20180012112A1
• Time in Patent Office: 748 Days
• Field of Search: None
• CPC Class Codes: G06F 18/2415 based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/1985 Syntactic analysis, e.g. using a grammatical approach syntactic image representation G06V30/18171
• CPC Class Codes: G06V 30/268 Lexical context

Abstract

According to an embodiment, a recognition device includes a candidate detection unit, a recognition unit, a matching unit, and a prohibition processing unit. The candidate detection unit detects, from an input image, character candidates each being a set of pixels estimated to include a character. The recognition unit recognizes each of the character candidates and generates one or more recognition candidates each being a character of a candidate as a recognition result. The matching unit matches each of the one or more recognition candidates with a knowledge dictionary in which a recognition target character string is modeled, and generates matching results obtained by matching a character string estimated to be included in the input image with the knowledge dictionary. The prohibition processing unit deletes, from the matching results, a matching result obtained by matching a character string including a prohibition target character string with the knowledge dictionary.

Claims

1. A recognition device comprising:

a memory; and

a hardware processor configured to;

detect, from an input image, character candidates, wherein each of the character candidates is a set of pixels estimated to include a character;

recognize each of the character candidates and generate one or more recognition candidates, wherein each of the one or more recognition candidates is a character of a candidate as a recognition result;

match each of the one or more recognition candidates with a knowledge dictionary in which a recognition target character string is modeled, and generate matching results obtained by matching a character string estimated to be included in the input image with the knowledge dictionary; and

match each of the matching results with a prohibition dictionary in which a prohibition target character string is modeled, and delete, from the matching results, a matching result obtained by matching a character string including the prohibition target character string with the prohibition dictionary.

2. The device according to claim 1, wherein the hardware processor is further configured to:

assign a score to each of the matching results representing likelihood of a corresponding character string, and

select, from the matching results, a matching result based on the score assigned to the matching result and extract a character string specified by the selected matching result.

3. The device according to claim 1, wherein the knowledge dictionary is a finite automaton in which a recognition target character string is modeled.

4. The device according to claim 3, wherein

the knowledge dictionary is a first deterministic finite automaton, and

in the matching of each of the one or more recognition candidates with the knowledge dictionary, the hardware processor is configured to:

write, in a matching result of a first character candidate, a number indicating a state of the first deterministic finite automaton reached through matching, and

follow, in matching a second character candidate subsequent to the first character candidate with the first deterministic finite automaton, an edge corresponding to a state transition based on a recognition candidate of the second character candidate from the state indicated by the number written in the matching result of the first character candidate to match the second character candidate.

5. The device according to claim 1, wherein in the matching of each of the matching results with the prohibition dictionary, the hardware processor is configured to match the character string specified by the matching result with the prohibition dictionary.

6. The device according to claim 5, wherein the prohibition dictionary is a finite automaton in which a prohibition target character string is modeled.

7. The device according to claim 6, wherein

the prohibition dictionary is a second deterministic finite automaton, and

in the matching of each of the matching results with the prohibition dictionary, the hardware processor is configured to:

write, in the first matching result, a number indicating a state of the second deterministic finite automaton reached through matching of a character string specified by a first matching result, and

follow, in matching a character string specified by a second matching result obtained by matching a second character candidate subsequent to the first matching result with the second deterministic finite automaton, an edge corresponding to a state transition based on a recognition candidate of the second character candidate from the state indicated by the number written in the first matching result to match the character string specified by the second matching result.

8. The device according to claim 7, wherein the second deterministic finite automaton is configured to transition to an accepting state when a character string including a prohibition target character string is received.

9. A recognition method comprising:

detecting, from an input image, character candidates that each comprise a set of pixels estimated to include a character;

recognizing each of the character candidates and generating one or more recognition candidates that each comprise a character of a candidate as a recognition result;

matching each of the one or more recognition candidates with a knowledge dictionary in which a recognition target character string is modeled, and generating matching results obtained by matching a character string estimated to be included in the input image with the knowledge dictionary; and

matching each of the matching results with a prohibition dictionary in which a prohibition target character string is modeled, and deleting, from the matching results, a matching result obtained by matching a character string including the prohibition target character string with the prohibition dictionary.

10. A computer program product comprising a non-transitory computer-readable medium including programmed instructions for causing a computer to function as a recognition device, the instructions causing the computer to execute:

detecting, from an input image, character candidates each of which is a set of pixels estimated to include a character;

recognizing each of the character candidates and generating one or more recognition candidates each of which is a character of a candidate as a recognition result;

matching each of the one or more recognition candidates with a knowledge dictionary in which a recognition target character string is modeled, and generating matching results obtained by matching a character string estimated to be included in the input image with the knowledge dictionary; and

matching each of the matching results with a prohibition dictionary in which a prohibition target character string is modeled, and deleting, from the matching results, a matching result obtained by matching a character string including the prohibition target character string with the prohibition dictionary.