Image recognition system, image recognition method, and non-transitory computer readable medium storing image recognition program

• US 9,600,745 B2
• Filed: 11/16/2011
• Issued: 03/21/2017
• Est. Priority Date: 03/17/2011
• Status: Active Grant

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Sakurai, Kazuyuki
• Primary Examiner(s): TSAI, TSUNG YIN
• Application Number:
  US14/005,228
• Publication Number:
  US 20140010410A1
• Time in Patent Office: 1,952 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06V 10/50 by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands

Abstract

An image recognition system includes a partial area determination unit for determining a plurality of learning partial areas in a learning image including an object to be recognized, a partial area set generation unit for generating a learning partial area set including the learning partial area and a plurality of peripheral areas included in a predetermined range with reference to the learning partial area, and a learning unit for selecting an area including an image suitable to be determined as an object to be identified included the object to be recognized from a plurality of areas included in the learning partial area set, to learn a classifier so as to determine a likelihood that the image included in the area is the object to be identified to be higher based on a feature amount related to the selected area.

Claims

1. An image recognition system, comprising:

an image recognition unit that recognizes an object to be recognized included in an input image based on a result of determination by a classifier, the classifier determining a likelihood that an image included in an arbitrary area in the input image including the object to be recognized having an object to be identified is the object to be identified based on a feature amount regarding the area;

a partial area determination unit that determines a plurality of learning partial areas in a learning image including the object to be recognized;

a partial area set generation unit that generates a learning partial area set based on a learning partial area of the plurality of learning partial areas, the learning partial area set including the learning partial area and a plurality of peripheral areas included in a predetermined range with reference to the learning partial area; and

a learning unit that selects, when performing learning of the classifier for the learning partial area, an area including an image suitable to be determined as the object to be identified from a plurality of areas included in the learning partial area set generated by the learning partial area, to learn the classifier so as to determine a likelihood that the image included in the area comprises the object to be identified to be higher based on a feature amount related to the selected area.

2. The image recognition system according to claim 1, wherein the learning unit calculates, when the likelihood that the classifier determines the image included in the area as the object to be identified based on a feature amount regarding the area is learned to become equal to or greater than a first predetermined likelihood, a relative position of a reference point set in advance for the object to be recognized including the object to be identified from the area,

wherein the image recognition system further comprises a voting pattern generation unit that generates a voting pattern indicating distribution of the relative position calculated by the learning unit, and

wherein the image recognition unit comprises:

an identification unit that determines a plurality of input partial areas in the input image to determine by the classifier a likelihood that an image included in an input partial area of the plurality of input partial areas comprises the object to be identified based on a feature amount regarding the input partial area;

a voting unit that votes, when the likelihood that the image included in the input partial area comprises the object to be identified is equal to or greater than a second predetermined likelihood, the voting pattern to a predetermined voting space with reference to the position of the input partial area; and

an object recognition unit that recognizes the object to be recognized based on a voting result to the voting space.

3. The image recognition system according to claim 2, wherein the likelihood equal to or greater than the first predetermined likelihood comprises a highest likelihood among likelihoods determined for the plurality of respective areas included in the learning partial area set.

4. The image recognition system according to claim 2, wherein the classifier determines whether the image included in the area comprises the object to be identified,

wherein the learning unit learns the classifier to determine that the image included in the area is the object to be identified based on a feature amount related to the selected area, and calculates the relative position of the reference point from the area when the classifier is learned to determine that the image included in the area comprises the object to be identified,

wherein the identification unit determines by the classifier whether the image included in the input partial area is the object to be identified, and

wherein the voting unit performs the voting when it is determined that the image included in the input partial area is the object to be identified.

5. The image recognition system according to claim 4, wherein the object to be recognized includes a plurality of objects to be identified,

wherein the voting pattern generation unit generates a plurality of voting patterns so as to correspond to the plurality of respective objects to be identified and to indicate distribution of the relative position of the corresponding objects to be identified, and

wherein the voting unit votes the voting pattern corresponding to the object to be identified when the likelihood that the image included in the input partial area comprises the object to be identified is equal to or greater than the second predetermined likelihood.

6. The image recognition system according to claim 5, wherein the image recognition system further comprises a weight calculation unit that calculates, for each of the plurality of objects to be identified, a weight of the voting pattern corresponding to the object to be identified based on a predetermined calculation criterion, and

wherein the voting unit performs voting while applying a weight corresponding to the voting pattern to the voting pattern.

7. The image recognition system according to claim 5, wherein the image recognition system comprises one classifier, and

wherein the one classifier determines, for each of the plurality of objects to be identified, the likelihood that an image included in the area comprises the object to be identified based on a feature amount regarding the area.

8. The image recognition system according to claim 5, wherein the plurality of input partial areas comprises a first input partial area and a second input partial area including images of objects to be identified different from each other, and

wherein, when a similarity of the first input partial area and the second input partial area becomes equal to or greater than a predetermined similarity, the classifier determines that both of the first input partial area and the second input partial area comprise the object to be identified of the image included in any one of the first input partial area and the second input partial area.

9. The image recognition system according to claim 1, wherein the learning partial area and the plurality of peripheral areas include areas in the learning image.

10. The image recognition system according to claim 1, wherein the plurality of learning partial areas includes areas in the learning image.

11. The image recognition system according to claim 1, wherein the learning unit selects the area including the image suitable to be determined as the object to be identified from the learning partial area set obtained from the learning image.

12. The image recognition system according to claim 1, wherein the learning unit learns the classifier based on the selected area.

13. An image recognition method, comprising:

determining a plurality of learning partial areas in a learning image including an object to be recognized having an object to be identified;

generating a learning partial area set based on a learning partial area of the plurality of learning partial areas, the learning partial area set including the learning partial area and a plurality of peripheral areas included in a predetermined range with reference to the learning partial area;

when performing learning of a classifier that identifies a likelihood that an image included in an arbitrary area in an input image including the object to be recognized comprises the object to be identified based on a feature amount regarding the area for the learning partial area, selecting an area including an image suitable to be determined as the object to be identified from a plurality of areas included in the learning partial area set generated by the learning partial area, to learn the identifier so as to determine a likelihood that the image included in the area is comprises the object to be identified to be higher based on a feature amount related to the selected area; and

recognizing the object to be recognized included in the input image based on a result of determining the input image by the classifier.

14. The image recognition method according to claim 13, wherein the learning partial area and the plurality of peripheral areas include areas in the learning image.

15. The image recognition method according to claim 13, wherein the plurality of learning partial areas includes areas in the learning image.

16. The image recognition method according to claim 13, wherein said selecting includes selecting the area including the image suitable to be determined as the object to be identified from the learning partial area set obtained from the learning image.

17. A non-transitory computer readable medium storing an image recognition program that causes a computer to execute the following processing of:

determining a plurality of learning partial areas in a learning image including an object to be recognized having an object to be identified;

generating a learning partial area set based on a learning partial area of the learning partial areas, the learning partial area set including the learning partial area and a plurality of peripheral areas included in a predetermined range with reference to the learning partial area;

when performing learning of a classifier that identifies a likelihood that an image included in an arbitrary area in an input image including the object to be recognized comprises the object to be identified based on a feature amount regarding the area for the learning partial area, selecting an area including an image suitable to be determined as the object to be identified from a plurality of areas included in the learning partial area set generated by the learning partial area, to learn the identifier so as to determine a likelihood that the image included in the area comprises the object to be identified to be higher based on a feature amount related to the selected area; and

recognizing the object to be recognized included in the input image based on a result of determining the input image by the classifier.

18. The non-transitory computer readable medium according to claim 17, wherein the learning partial area and the plurality of peripheral areas include areas in the learning image.

19. The non-transitory computer readable medium according to claim 17, wherein the plurality of learning partial areas includes areas in the learning image.

20. The non-transitory computer readable medium according to claim 17, wherein said selecting includes selecting the area including the image suitable to be determined as the object to be identified from the learning partial area set obtained from the learning image.