LEARNING DEVICE AND METHOD, RECOGNITION DEVICE AND METHOD, AND PROGRAM

• US 20110135192A1
• Filed: 10/29/2010
• Published: 06/09/2011
• Est. Priority Date: 10/03/2008
• Status: Active Grant

• Current Assignee: Jun Yokono
• Original Assignee: Sony Corporation (Sony Group Corp.)
• Inventors: YOKONO, Jun
• Granted Patent: US 9,275,305 B2
• Time in Patent Office: Days
• US Class Current: 382/160
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06V 10/446 using Haar-like filters, e.g. using integral image techniques
• CPC Class Codes: G06V 10/449 Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
• CPC Class Codes: G06V 10/507 Summing image-intensity values; Histogram projection analysis
• CPC Class Codes: G06V 10/774 Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06V 40/103 Static body considered as a whole, e.g. static pedestrian or occupant recognition

Abstract

A learning device includes: a generating unit configured to generate an image having different resolution from an input image; an extracting unit configured to extract a feature point serving as a processing object from an image generated by the generating unit; a calculating unit configured to calculate the feature amount of the feature point by subjecting the feature point to filter processing employing a predetermined filter; and an identifier generating unit configured to generate an identifier for detecting a predetermined target object from the image by statistical learning employing the feature amount; with the filter including a plurality of regions, and the calculating unit taking the difference value of difference within the regions as the feature amount.

Claims

1. A learning device comprising:

generating means configured to generate an image having different resolution from an input image;

extracting means configured to extract a feature point serving as a processing object from an image generated by the generating means;

calculating means configured to calculate the feature amount of the feature point by subjecting the feature point to filter processing employing a predetermined filter; and

identifier generating means configured to generate an identifier for detecting a predetermined target object from the image by statistical learning employing the feature amount;

wherein the filter includes a plurality of regions, and the calculating means take the difference value of difference within the regions as the feature amount.

2. The learning device according to claim 1, wherein the regions of the predetermined filter have a rectangular shape.

3. The learning device according to claim 1, wherein the calculating means execute convolution operation in a differential function of a predetermined order of a Gaussian function in an arbitrary angle to calculate the summation of the absolute values of the operation results.

4. The learning device according to claim 1, wherein the calculating means execute convolution operation in a differential function of a predetermined order of a Gaussian function in an arbitrary angle to calculate the maximum value of the absolute values of the operation results.

5. The learning device according to claim 1, wherein the calculating means calculate the summation or average value of the pixel values of pixels positioned within the region for each of the plurality of regions, calculate the difference of the summation or average value thereof, and take the absolute value of the difference value thereof as the feature amount.

6. The learning device according to claim 1, wherein the calculating means create a histogram from the pixel values of pixels positioned within the region for each of the plurality of regions, calculate the difference of the values thereof, and take the absolute value of the difference value thereof as the feature amount.

7. The learning device according to claim 1, wherein the calculating means create a histogram from the pixel values of pixels positioned within the region for each of the plurality of regions, perform normalization so that the sum of the histogram thereof becomes 1, calculate the difference of the values of the normalized histogram, and take the absolute value of the difference value thereof as the feature amount.

8. The learning device according to claim 1, wherein the calculating means apply a filter applied to a feature point serving as a reference point to a plurality of feature points positioned in the neighborhood of the feature point serving as the reference thereof, and take the summation or average value of the feature amounts calculated from these plurality of feature points as the feature amount of the feature point serving as the reference.

9. A learning method comprising the steps of:

generating an image having different resolution from an input image;

extracting a feature point from the generated image;

calculating the feature amount of the feature point by subjecting the feature point to filter processing employing a predetermined filter; and

generating an identifier for detecting a predetermined target object from the image by statistical learning employing the feature amount;

wherein the filter includes a plurality of regions, and the difference value of the difference within the regions thereof is taken as the feature amount.

10. A computer-readable program causing a computer to execute processing comprising the steps of:

generating an image having different resolution from an input image;

extracting a feature point from the generated image;

calculating the feature amount of the feature point by subjecting the feature point to filter processing employing a predetermined filter; and

generating an identifier for detecting a predetermined target object from the image by statistical learning employing the feature amount;

wherein the filter includes a plurality of regions, and the difference value of the difference within the regions thereof is taken as the feature amount.

11. A recognition device comprising:

generating means configured to generate an image having different resolution from an input image;

extracting means configured to extract a feature point from an image generated by the generating means;

calculating means configured to calculate the feature amount of the feature point by subjecting the feature point to filter processing employing a predetermined filter; and

detecting means configured to substitute the feature amount for an identifier for detecting a predetermined target object from an image, obtained by statistical learning, to detect the target object from the input image;

wherein the filter includes a plurality of regions, and the calculating means take the difference value of difference within the regions as the feature amount.

12. The recognition device according to claim 11, wherein the regions of the predetermined filter have a rectangular shape.

13. The recognition device according to claim 11, wherein the calculating means execute convolution operation in a differential function of a predetermined order of a Gaussian function in an arbitrary angle to calculate the summation of the absolute values of the operation results.

14. The recognition device according to claim 11, wherein the calculating means execute convolution operation in a differential function of a predetermined order of a Gaussian function in an arbitrary angle to calculate the maximum value of the absolute values of the operation results.

15. The recognition device according to claim 11, wherein the calculating means calculate the summation or average value of the pixel values of pixels positioned within the region for each of the plurality of regions, calculate the difference of the summation or average value thereof, and take the absolute value of the difference value thereof as the feature amount.

16. The recognition device according to claim 11, wherein the calculating means create a histogram from the pixel values of pixels positioned within the region for each of the plurality of regions, calculate the difference of the values thereof, and take the absolute value of the difference value thereof as the feature amount.

17. The recognition device according to claim 11, wherein the calculating means create a histogram from the pixel values of pixels positioned within the region for each of the plurality of regions, perform normalization so that the sum of the histogram thereof becomes 1, calculate the difference of the values of the normalized histogram, and take the absolute value of the difference value thereof as the feature amount.

18. The recognition device according to claim 11, wherein the calculating means apply a filter applied to a feature point serving as a reference point to a plurality of feature points positioned in the neighborhood of the feature point serving as the reference thereof, and take the summation or average value of the feature amounts calculated from these plurality of feature points as the feature amount of the feature point serving as the reference.

19. A recognition method comprising the steps of:

generating an image having a different scale coefficient from an input image;

extracting a feature point from the generated image;

calculating the feature amount of the feature point by subjecting the feature point to filter processing employing a predetermined filter; and

substituting the feature amount for an identifier for detecting a predetermined target object from an image, obtained by statistical learning, to detect the target object from the input image;

wherein the filter includes a plurality of regions, and the difference value of difference within the regions is taken as the feature amount.

20. A computer-readable program causing a computer to execute processing comprising the steps of:

generating an image having a different scale coefficient from an input image;

extracting a feature point from the generated image;

calculating the feature amount of the feature point by subjecting the feature point to filter processing employing a predetermined filter; and

substituting the feature amount for an identifier for detecting a predetermined target object from an image, obtained by statistical learning, to detect the target object from the input image;

wherein the filter includes a plurality of regions, and the difference value of difference within the regions is taken as the feature amount.

21. A learning device comprising:

a generating unit configured to generate an image having different resolution from an input image;

an extracting unit configured to extract a feature point serving as a processing object from an image generated by the generating unit;

a calculating unit configured to calculate the feature amount of the feature point by subjecting the feature point to filter processing employing a predetermined filter; and

an identifier generating unit configured to generate an identifier for detecting a predetermined target object from the image by statistical learning employing the feature amount;

wherein the filter includes a plurality of regions, and the calculating unit take the difference value of difference within the regions as the feature amount.

22. A recognition device comprising:

a generating unit configured to generate an image having different resolution from an input image;

an extracting unit configured to extract a feature point from an image generated by the generating unit;

a calculating unit configured to calculate the feature amount of the feature point by subjecting the feature point to filter processing employing a predetermined filter; and

a detecting unit configured to substitute the feature amount for an identifier for detecting a predetermined target object from an image, obtained by statistical learning, to detect the target object from the input image;

wherein the filter includes a plurality of regions, and the calculating unit takes the difference value of difference within the regions as the feature amount.