Captured image recognition device, captured image recognition system, and captured image recognition method

• US 9,842,259 B2
• Filed: 12/21/2011
• Issued: 12/12/2017
• Est. Priority Date: 01/17/2011
• Status: Active Grant

• Current Assignee: Panasonic Intellectual Property Management Co., Ltd. (Panasonic Holdings Corporation)
• Original Assignee: Panasonic Intellectual Property Management Co., Ltd. (Panasonic Holdings Corporation)
• Inventors: Yamazaki, Ryuji
• Primary Examiner(s): Chen, Xuemei
• Application Number:
  US13/979,864
• Publication Number:
  US 20130308825A1
• Time in Patent Office: 2,183 Days
• Field of Search: 348 1408, 3482401, 348143, 348159, 348169, 348345, 351240, 353 79, 382276
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69
• CPC Class Codes: H04N 7/18 Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast

Abstract

Provided is a captured image recognition device that enables the performance of an image recognition function to be sufficiently evinced. A field-of-view splitting estimation unit (31) estimates the splitting of the field of view of a camera unit (20) using a captured image (S1). On the basis of the estimated splitting of the field of view, a candidate application selection unit (32) selects, from among a plurality of image recognition applications, candidates for an image recognition application that is favorable or able to execute processing with respect to a current captured image. An image recognition processing unit (40) executes the image recognition application selected by a user from among the candidate applications. As a result, the performance of an image recognition function can be sufficiently evinced as a result of it being possible to execute an image recognition application that is suitable to the current captured image.

Claims

1. A captured-image recognition apparatus comprising:

a memory that stores instructions; and

a processor that, when executing the instructions stored in the memory, performs a process comprising;

estimating a visual field ratio of a target in a captured image, using at least two kinds of detecting methods that detect different kinds of targets, the visual field ratio being a ratio of a target size in the captured image to a field-angle size of the captured image;

selecting, based on the estimated visual field ratio, at least one candidate image recognition application from a plurality of image recognition applications that recognize different kinds of targets;

performing image recognition by executing an image recognition application selected from among the at least one candidate image recognition application; and

optimizing the visual filed ratio based upon the selected image recognition application.

2. The captured-image recognition apparatus according to claim 1, wherein,

in the estimating, the processor extracts a target in the captured image by the detecting methods.

3. The captured-image recognition apparatus according to claim 1, wherein

the detecting methods include person detection that detects a person from the captured image, and face detection that detects a face from the captured image, and

in the estimating, the processor estimates the visual field ratio based on the person detection result and the face detection result.

4. The captured-image recognition apparatus according to claim 3, wherein,

in the estimating, the processor estimates the visual field ratio to be smaller for a higher success rate of the person detection.

5. The captured-image recognition apparatus according to claim 3, wherein,

in the estimating, the processor estimates the visual field ratio to be larger for a higher success rate of the face detection.

6. The captured-image recognition apparatus according to claim 3, wherein,

in the estimating, the processor performs the person detection using multiple templates having different sizes and estimates the visual field ratio based on degrees of success or failure of the person detection using the respective sizes.

7. The captured-image recognition apparatus according to claim 1, wherein:

the image recognition applications are set with proper visual field ratios, respectively; and

in the selecting, the processor selects the image recognition application based on the proper visual field ratio set for each of the image recognition applications and the estimated visual field ratio.

8. The captured-image recognition apparatus according to claim 1, wherein,

in the selecting, the processor selects an image recognition application in consideration of a zoom range of an imaging apparatus configured to take the captured image.

9. The captured-image recognition apparatus according to claim 1, wherein,

in the selecting, the processor selects, as the at least one candidate image recognition application, at least one of a congestion detection application, a person detection application, a movement-recognition application, a facial-recognition application, a visual-line estimation application, a movement-line detection application, a cloth search application, a sex estimation application, an age estimation application, and a facial-expression estimation application.

10. The captured-image recognition apparatus according to claim 1, wherein, the at least one candidate image recognition application selected in the selecting includes a facial-recognition application, when the estimated visual field ratio is greater than a first predetermined value;

and the at least one candidate image recognition application selected in the selecting includes a person detection application when the estimated visual field ratio is smaller than a second predetermined value, and

the executing executes the facial-recognition application, when the estimated visual field ratio is greater than the first predetermined value, and executes the person detection application when the estimated visual field ratio is smaller the second predetermined value.

11. The captured image recognition apparatus according to claim 1, wherein in the estimating, the processor extracts a target in the captured image by the detecting methods, and estimates the visual field ratio based on a visual field angle of a camera and a positional relationship between the camera and the target.

12. The captured image recognition apparatus according to claim 1, further comprising storing the estimated visual field ratio and updating the stored estimated visual field ratio based on the optimized visual field ratio.

13. The captured image recognition apparatus according to claim 1, wherein each of the plurality of image recognition applications corresponds to a range of visual field ratios.

14. A server apparatus comprising:

a first memory that stores a database comprising a plurality of image recognition applications that recognize different kinds of targets; and

a captured-image recognition apparatus,

the captured-image recognition apparatus comprising;

a second memory that stores instructions; and

a processor that, when executing the instructions stored in the second memory, performs a process comprising;

estimating a visual field ratio of a target in a captured image, using at least two kinds of detecting methods that detect different kinds of targets, the visual field ratio being a ratio of a target size in the captured image to a field-angle size of the captured image;

selecting, based on the estimated visual field ratio, at least one candidate image recognition application from the plurality of image recognition applications;

performing image recognition by executing an image recognition application selected from among the at least one candidate image recognition application; and

optimizing the visual filed ratio based upon the selected image recognition application.

15. A captured-image recognition system comprising:

a server apparatus that includes a first memory storing a database comprising a plurality of image recognition applications that recognize different kinds of targets, a second memory storing instructions, and a processor executing the instructions stored in the second memory; and

a network camera connected to the server apparatus in a network,

wherein;

the processor selects at least one candidate image recognition application from the plurality of image recognition applications, based on a visual field ratio of a target detected by using at least two detecting methods, the at least two detecting methods detecting different kinds of the targets in an image captured by the network camera, the visual field ratio being estimated as a ratio of a target size in the captured image to a field-angle size of the captured image, and the visual field ratio being optimized based on the selected at least one candidate image recognition application;

the processor presents information on the selected at least one candidate image recognition application to a user of the network camera;

the processor transmits a program for an image recognition application selected by the user to the network camera; and

the network camera performs an image recognition process by executing the program for the image recognition application selected by the user,

wherein, the at least one candidate image recognition application selected by the processor includes a facial-recognition application, when the estimated visual field ratio is greater than a first predetermined value; and

the at least one candidate image recognition application selected by the processor includes a person detection application when the estimated visual field ratio is smaller than a second predetermined value.

16. A captured-image recognition method comprising:

estimating a visual field ratio of a target in a captured image using at least two kinds of detecting methods that detect different kinds of targets, the visual field ratio being a ratio of a target size in the captured image to a field-angle size of the captured image;

selecting at least one candidate image recognition application corresponding to the estimated visual field ratio, from a plurality of image recognition applications that recognize different kinds of targets;

performing image recognition by executing an image recognition application selected from the at least one candidate image recognition application; and

optimizing the visual field ratio based upon the selected image recognition application,

wherein, the at least one candidate image recognition application selected in the selecting includes a facial-recognition application, when the estimated visual field ratio is greater than a first predetermined value; and

the at least one candidate image recognition application selected in the selecting includes a person detection application when the estimated visual field ratio is smaller than a second predetermined value, and

the executing executes the facial-recognition application, when the estimated visual field ratio is greater than the first predetermined value, and executes the person detection application when the estimated visual field ratio is smaller than the second predetermined value.

17. The captured image recognition method according to claim 16, wherein each of the plurality of image recognition applications corresponds to a range of visual field ratios.

18. A non-transitory computer readable recording medium storing a program that causes a computer to execute a process comprising:

estimating a visual field ratio of a target in a captured image using at least two kinds of detecting methods that detect different kinds of targets, the visual field ratio being a ratio of a target size in the captured image to a field-angle size of the captured image;

selecting at least one candidate image recognition application corresponding to the estimated visual field ratio, from a plurality of image recognition applications that recognize different kinds of targets;

performing image recognition by executing an image recognition application selected from the at least one candidate image recognition application; and

optimizing the visual field ratio based upon the selected image recognition application.

19. The non-transitory computer readable according medium according to claim 18, wherein, the at least one candidate image recognition application selected in the selecting includes a facial-recognition application, when the estimated visual field ratio is greater than a first predetermined value;

and the at least one candidate image recognition application selected in the selecting includes a person detection application when the estimated visual field ratio is smaller than a second predetermined value, and

the executing executes the facial-recognition application, when the estimated visual field ratio is greater than the first predetermined value, and executes the person detection application when the estimated visual field ratio is smaller than the second predetermined value.

20. The non-transitory computer readable recording medium according to claim 18, further comprising storing the estimated visual field ratio and updating the stored estimated visual field ratio based upon the optimized visual field ratio.