News Column

Researchers Submit Patent Application, "Image Recognition System, Image Recognition Method, and Non-Transitory Computer Readable Medium Storing Image...

January 30, 2014



Researchers Submit Patent Application, "Image Recognition System, Image Recognition Method, and Non-Transitory Computer Readable Medium Storing Image Recognition Program", for Approval

By a News Reporter-Staff News Editor at Computer Weekly News -- From Washington, D.C., VerticalNews journalists report that a patent application by the inventor Sakurai, Kazuyuki (Tokyo, JP), filed on November 16, 2011, was made available online on January 16, 2014.

The patent's assignee is NEC Corporation.

News editors obtained the following quote from the background information supplied by the inventors: "One example of a learning apparatus is disclosed in Patent literature 1. As shown in FIG. 14, this learning apparatus includes a storage device 1000, a feature point detection unit 1001, a feature amount calculation unit 1002, a vote learning unit 1003, and a classifier learning unit 1004. This learning apparatus has such a feature that, since the learning apparatus identifies an object by voting of feature points, it is relatively robust regarding the difference in shape or the like of each recognition target.

"The learning apparatus having such a configuration operates as follows.

"The storage device 1000 stores learning images consisting of an image group related to recognition target object and an image group of objects other than the recognition target object. The feature point detection unit 1001 detects a number of feature points from the respective learning images. The feature amount calculation unit 1002 calculates a feature vector as a feature amount related to the feature points and a peripheral image area. The vote learning unit 1003 calculates and stores voting positional information in a parameter space as a voting space for the feature point corresponding to the feature vector calculated from the image related to the recognition target object of the learning image. The classifier learning unit 1004 learns the classifier configured to discriminate whether a given feature point detected in recognition of the recognition target object belongs to the recognition target object using the feature vector.

"Patent literature 2 discloses a learning apparatus aimed at improving an identifying performance. The learning apparatus calculates, for each point on a sample image, local information required to recognize a pattern using a rectangular window set around the point. Further, the learning apparatus calculates, for each point on the sample image, arrangement information that specifies identifying classes of areas in the periphery of the marked point. Then the learning apparatus selects one combined information from a plurality of combined information being generated by combining the local information and the arrangement information, to calculate an identifying parameter for one weak classifier based on the combined information that is selected.

"Non-patent literatures 1-4 also disclose techniques related to image recognition."

As a supplement to the background information on this patent application, VerticalNews correspondents also obtained the inventor's summary information for this patent application: "Technical Problem

"The technique disclosed in Patent literature 1 described above has a problem that it is impossible to accurately recognize an object to be recognized formed of a curved line (e.g., a person). The reason for it is that, if the object to be recognized is an object formed of a curved line (e.g., a person), it is difficult to accurately detect the feature points. More specifically, detection of the feature points used in the technique disclosed in Patent literature 1 is to detect corner points, blobs and the like; however, an object such as a person or the like having a contour line whose curvature changes rather smoothly rarely has such features.

"As stated above, Patent literature 2 discloses a technique aimed at improving the identifying performance. However, the technique disclosed in Patent literature 2 does not disclose a technique of learning a classifier based on an area selected from a partial area set, which is different from the present invention.

"The present invention has been made in order to solve the aforementioned problems, and aims to provide an image recognition system, an image recognition method, and a non-transitory computer readable medium storing an image recognition program which are capable of robustly recognizing even an object to be recognized including a curved line.

"Solution to Problem

"An image recognition system according to a first exemplary aspect of the present invention includes: image recognition means for recognizing 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; partial area determination means for determining a plurality of learning partial areas in a learning image including the object to be recognized; partial area set generation means for generating a learning partial area set based on the learning partial area, 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 learning means for selecting, 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 the 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.

"An image recognition method according to a second exemplary aspect of the present invention includes: 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 the learning partial area, 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 is 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 the 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; and recognizing the object to be recognized included in the input image based on a result of determining the input image by the classifier.

"A non-transitory computer readable medium storing an image recognition program according to a third exemplary aspect of the present invention 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 the learning partial area, 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 is 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 the 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; and recognizing the object to be recognized included in the input image based on a result of determining the input image by the classifier.

"Advantageous Effects of Invention

"According to each exemplary aspect of the present invention stated above, it is possible to provide an image recognition system, an image recognition method, and a non-transitory computer readable medium storing an image recognition program which are capable of robustly recognizing even an object to be recognized including a curved line.

BRIEF DESCRIPTION OF DRAWINGS

"FIG. 1 is a block diagram showing a schematic configuration of an image recognition system according to an exemplary embodiment of the present invention;

"FIG. 2 is a block diagram showing a configuration of the image recognition system according to the exemplary embodiment of the present invention;

"FIG. 3 is a diagram describing search for a partial area from a partial area set;

"FIG. 4 is a diagram showing a method of generating a voting pattern;

"FIG. 5 is a diagram for describing voting;

"FIG. 6 is a flowchart showing a learning operation of the image recognition system according to the exemplary embodiment of the present invention;

"FIG. 7 is a flowchart showing an identifying operation of the image recognition system according to the exemplary embodiment of the present invention;

"FIG. 8 is a diagram showing an example of an original image which is a recognition target;

"FIG. 9 is a block diagram showing a configuration of a specific example of the image recognition system according to the exemplary embodiment of the present invention;

"FIG. 10 is a diagram describing division into partial areas;

"FIG. 11 is a flowchart showing a learning operation of a specific example of the image recognition system according to the exemplary embodiment of the present invention;

"FIG. 12 is a flowchart showing an identifying operation of a specific example of the image recognition system according to the exemplary embodiment of the present invention;

"FIG. 13 is a block diagram showing a hardware configuration of a computer according to the exemplary embodiment of the present invention; and

"FIG. 14 is a block diagram showing a configuration of one example of a learning apparatus according to a related art."

For additional information on this patent application, see: Sakurai, Kazuyuki. Image Recognition System, Image Recognition Method, and Non-Transitory Computer Readable Medium Storing Image Recognition Program. Filed November 16, 2011 and posted January 16, 2014. Patent URL: http://appft.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&u=%2Fnetahtml%2FPTO%2Fsearch-adv.html&r=3067&p=62&f=G&l=50&d=PG01&S1=20140109.PD.&OS=PD/20140109&RS=PD/20140109

Keywords for this news article include: NEC Corporation.

Our reports deliver fact-based news of research and discoveries from around the world. Copyright 2014, NewsRx LLC


For more stories covering the world of technology, please see HispanicBusiness' Tech Channel



Source: Computer Weekly News


Story Tools