Recognition dictionary creation apparatus and method for creating recognition dictionary by the same

• US 9,454,708 B2
• Filed: 05/23/2014
• Issued: 09/27/2016
• Est. Priority Date: 05/28/2013
• Status: Active Grant

• Current Assignee: Toshiba Tec Corporation (Toshiba Corporation)
• Original Assignee: Toshiba Tec Corporation (Toshiba Corporation)
• Inventors: Miyakoshi, Hidehiko
• Primary Examiner(s): BELLA, MATTHEW C
• Assistant Examiner(s): Torres, Jose M
• Application Number:
  US14/285,688
• Publication Number:
  US 20140355838A1
• Time in Patent Office: 858 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
• CPC Class Codes: G06F 18/41 Interactive pattern learning with a human teacher
• CPC Class Codes: G06T 15/50 Lighting effects
• CPC Class Codes: G06V 10/7788 the supervisor being a human, e.g. interactive learning with a human teacher
• CPC Class Codes: G06V 20/68 Food, e.g. fruit or vegetables

Abstract

A recognition dictionary creation apparatus photographs a commodity by an image capturing section to capture the image of the commodity; measures a distance from the image capturing section to the commodity photographed by the image capturing section; generates a quasi-commodity-image in a state in which the commodity is imaginarily moved away from the image capturing section to a given registration distance longer than the measurement distance based on the image of the commodity photographed by the image capturing section; extracts an appearance feature amount of the commodity from the generated quasi-commodity-image; and registers, in a recognition dictionary in association with the registration distance data, the extracted appearance feature amount as the feature amount data of the photographed commodity at the registration distance to create the recognition dictionary.

Claims

1. A recognition dictionary creation apparatus, comprising:

an image capturing section configured to photograph a commodity to capture the image of the commodity;

a measurement module configured to measure a distance from the image capturing section to the commodity photographed by the image capturing section;

a specifying module configured to specify the commodity;

a storage module configured to store a correction coefficient for representing an attenuation trend of light reflected by a surface of the commodity having a standard figure when photographing the commodity at each separation distance from the image capturing section;

a generation module configured to generate a quasi-commodity-image in a state in which the commodity is imaginarily moved away from the image capturing section to a given registration distance longer than the distance measured by the measurement module based on the image of the commodity photographed by the image capturing section;

a correction module configured to correct the quasi-commodity-image generated by the generation module with the correction coefficient for the registration distance;

an after-generation extraction module configured to extract an appearance feature amount of the commodity from the quasi-commodity-image generated by the generation module, wherein the after-generation extraction module extracts the appearance feature amount of the commodity from the quasi-commodity-image corrected by the correction module; and

an after-generation registration module configured to register, in a recognition dictionary in association with the registration distance, the appearance feature amount extracted by the after-generation extraction module as the feature amount data of the commodity specified by the specifying module at the registration distance.

2. The recognition dictionary creation apparatus according to claim 1, further comprising:

an acquisition module configured to acquire the appearance shape of the commodity photographed by the image capturing section;

wherein

the storage module stores, for each standard figure, a correction coefficient for representing the attenuation trend of each light reflected by the surfaces of objects having different standard figures when photographing the objects at each separation distance from the image capturing section, and

the correction module corrects the quasi-commodity-image with a correction coefficient of a standard figure similar to the appearance shape acquired by the acquisition module.

3. The recognition dictionary creation apparatus according to claim 1, further comprising:

a before-generation extraction module configured to extract an appearance feature amount of the commodity from the image of the commodity photographed by the image capturing section; and

a before-generation registration module configured to register the appearance feature amount extracted by the before-generation extraction module as the feature amount data of the commodity specified by the specifying module at the distance measured by the measurement module in the recognition dictionary in association with the distance data.

4. The recognition dictionary creation apparatus according to claim 1, further comprising:

a determination module configured to determine whether or not the entire shape of the commodity is imaged in the image captured by the image capturing section; and

a warning module configured to issue a warning if the entire shape of the commodity is not completely imaged.

5. A method for creating a recognition dictionary by a recognition dictionary creation apparatus having an image capturing section, including:

photographing a commodity by the image capturing section to capture the image of the commodity;

measuring a distance from the commodity to the image capturing section;

specifying the commodity;

storing a correction coefficient for representing an attenuation trend of light reflected by a surface of the commodity having a standard figure when photographing the commodity at each separation distance from the image capturing section;

generating a quasi-commodity-image in a state in which the commodity is imaginarily moved away from the image capturing section to a given registration distance longer than the measured distance based on the image of the commodity photographed by the image capturing section;

correcting the generated quasi-commodity-image with the correction coefficient for the registration distance;

extracting an appearance feature amount of the commodity from the corrected quasi-commodity-image; and

registering, in a recognition dictionary in association with the registration distance, the extracted appearance feature amount as the feature amount data of the specified commodity at the registration distance to create the recognition dictionary.