Recognition apparatus and recognition method

Patent metadata

Field	Value
Publication number	US-9773189-B2
Application number	US-201514685427-A
Country	US
Kind code	B2
Filing date	Apr 13, 2015
Priority date	Apr 15, 2014
Publication date	Sep 26, 2017
Grant date	Sep 26, 2017

Abstract

A recognition apparatus according to an embodiment of the present invention includes: a candidate region extraction unit configured to extract a subject candidate region from an image; a feature value extraction unit configured to extract a feature value related to an attribute of the image from the subject candidate region extracted by the candidate region extraction unit; an attribute determination unit configured to determine an attribute of the subject candidate region extracted by the candidate region extraction unit on the basis of the feature value extracted by the feature value extraction unit; and a determination result integration unit configured to identify an attribute of the image by integrating determination results of the attribute determination unit.

First claim

What is claimed is:

1. A recognition apparatus comprising:
one or more non-transitory computer-readable storage devices; and
one or more computer processing devices connected to the one or more non-transitory computer-readable storage devices and configured by one or more programs stored in the one or more non-transitory computer-readable storage devices at least to:
extract object candidate regions from an image;
extract a first feature value and a second feature value from each of the extracted object candidate regions, the first feature value being a value used for determining an attribute of the image, the second feature value being a value different from the first feature value and used for determining whether the object candidate region is a region showing an object included in the image or not;
determine the attribute of the image for each of the extracted object candidate regions on a basis of the first feature value, and determine whether the extracted object candidate region is a region showing the object or not each on a basis of the second feature value; and
identify the attribute of the image by integrating determination results regarding the attribute of the image in, among the object candidate regions, object candidate regions determined each as a region showing the object.

2. The recognition apparatus according to claim 1 , wherein the extraction of the object candidate regions is performed on a basis of one of criteria.

3. The recognition apparatus according to claim 2 , wherein each of the criteria has been decided in accordance with the attribute of the image to be determined.

4. The recognition apparatus according to claim 1 , further configured to extract a human body candidate region from the image.

5. The recognition apparatus according to claim 1 , wherein, when the attribute of the image for each of the extracted object candidate regions is learned, object candidate regions are classified in accordance with an attribute of the image.

6. The recognition apparatus according to claim 5 , wherein, when classifying object candidate regions, performing learning so that a region taught as an important object candidate region is preferentially classified.

7. The recognition apparatus according to claim 1 , further configured to perform learning using an object candidate region irrelevant to the attribute of the image.

8. The recognition apparatus according to claim 1 , further configured to employ a method based on a classification tree.

9. The recognition apparatus according to claim 1 , further configured to employ a method based on a similar case data search.

10. The recognition apparatus according to claim 1 , further configured to employ a method based on a Hash method.

11. The recognition apparatus according to claim 1 , wherein the determined attribute is one of an image scene, a behavior of a crowd in an image, a type of an image composition, information on a main subject in an image, and information on a direction of a light source of an image.

12. The recognition apparatus according to claim 1 , wherein an image whose attribute is a determination target is a moving image.

13. The recognition apparatus according to claim 1 , wherein an image whose attribute is a determination target is a distance image.

14. A recognition method comprising the steps of:
extracting object candidate regions from an image;
extracting a first feature value and a second feature value from each of the extracted object candidate regions, the first feature value being a value used for determining an attribute of the image, the second feature value being a value different from the first feature value and used for determining whether the object candidate region is a region showing an object included in the image or not;
determining the attribute of the image for each of the extracted object candidate regions on a basis of the first feature value, and determine whether the extracted object candidate region is a region showing the object or not each on a basis of the second feature value; and
identifying the attribute of the image by integrating determination results regarding the attribute of the image in, among the object candidate regions, object candidate regions determined each as a region showing the object.

15. A non-transitory computer readable storage medium storing a program for causing a computer to execute:
extracting object candidate regions from an image;
extracting a first feature value and a second feature value from each of the extracted object candidate regions, the first feature value being a value used for determining an attribute of the image, the second feature value being a value different from the first feature value and used for determining whether the object candidate region is a region showing an object included in the image or not;
determining the attribute of the image for each of the extracted object candidate regions on a basis of the first feature value, and determine whether the extracted object candidate region is a region showing the object or not each on a basis of the second feature value; and
identifying the attribute of the image by integrating determination results regarding the attribute of the image in, among the object candidate regions, object candidate regions determined each as a region showing the object.

16. The recognition apparatus according to claim 1 , wherein the subject object candidate regions are extracted by repeating processing of coupling two super pixels that are adjacent to each other and have a highest level of similarity therebetween, among plural super pixels generated by dividing the image.

17. The recognition apparatus according to claim 16 , wherein super pixels whose area size is greater than a predetermined value are extracted as the object candidate regions.

Assignees

• Canon Kk

Inventors

• Tate Shunta
• Mori Katsuhiko
• Matsugu Masakazu

Classifications

• G06V20/54 Primary

  of traffic, e.g. cars on the road, trains or boats · CPC title

• G06F18/24147

  Distances to closest patterns, e.g. nearest neighbour classification · CPC title

• G06F18/22

  Matching criteria, e.g. proximity measures · CPC title

• G06F18/217

  Validation; Performance evaluation; Active pattern learning techniques · CPC title

• B60Q9/00

  Arrangement or adaptation of signal devices not provided for in one of main groups B60Q1/00 - B60Q7/00 {, e.g. haptic signalling} · CPC title