Object identification apparatus, object identification method, and program

• US 9,679,221 B2
• Filed: 05/21/2013
• Issued: 06/13/2017
• Est. Priority Date: 08/23/2012
• Status: Active Grant

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Mase, Ryota
• Primary Examiner(s): HU, FRED H.
• Application Number:
  US14/422,842
• Publication Number:
  US 20150220810A1
• Time in Patent Office: 1,484 Days
• Field of Search: None
• CPC Class Codes: G06T 3/00 Geometric image transformations in the plane of the image
• CPC Class Codes: G06T 5/80 Geometric correction
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]
• CPC Class Codes: G06V 10/751 Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
• CPC Class Codes: G06V 10/757 Matching configurations of points or features
• CPC Class Codes: G06V 10/759 Region-based matching

Abstract

An input image showing a same object as an object shown in a reference image is identified more accurately. A difference area in the input image is determined by converting a difference area in the reference image, on a basis of geometric transformation information calculated by an analysis using a local descriptor. By matching a descriptor extracted from the difference area in the input image with the difference area in the reference image, fine differences that cannot be identified by conventional matching using only a local descriptor can be distinguished and images showing a same object can be exclusively identified.

Claims

1. An object identification apparatus comprising:

a processor; and

a memory coupled to the processor;

wherein the memory stores instructions that cause the processor to;

determine whether or not respective descriptors of feature points extracted from an input image and respective descriptors of feature points extracted from a reference image correctly correspond to one another;

extract a first descriptor of an area in the input image corresponding to a position of an image area obtained by performing a geometric transformation for correcting a geometric deviation between the input image and the reference image on a prescribed area of the reference image when a score based on the number of combinations of descriptors determined to correspond correctly is equal to or larger than a prescribed value;

match the first descriptor with a second descriptor extracted from the prescribed area of the reference image; and

output a matching result,

wherein the first descriptor and the second descriptor are respectively parts of the descriptor of the input image and the descriptor of the reference image that have been used in the determination of whether or not the descriptors of feature points extracted from an input image and respective descriptors of feature points extracted from a reference image correctly correspond to one another.

2. The object identification apparatus according to claim 1, wherein information regarding the prescribed area of the reference image is stored.

3. The object identification apparatus according to claim 1, wherein the prescribed area of the reference image is an area including a feature point in the reference image, to which a descriptor is determined to correspond erroneously.

4. The object identification apparatus according to claim 1, wherein the prescribed area of the reference image is an area including a feature point in the reference image, to which a descriptor is determined to correspond erroneously, in an area showing an object.

5. The object identification apparatus according to claim 1, wherein the prescribed area of the reference image includes an area in which a difference between an image obtained by performing a geometric transformation for correcting a geometric deviation between the input image and the reference image on the input image and the reference image is equal to or greater than a prescribed value.

6. The object identification apparatus according to claim 1, wherein the prescribed area of the reference image is an area which includes a feature point, to which a descriptor is determined to correspond erroneously, in an area showing an object, and which includes an area in which a difference between an image obtained by performing a geometric transformation for correcting a geometric deviation between the input image and the reference image on the input image and the reference image is equal to or greater than a prescribed value.

7. The object identification apparatus according to claim 1, wherein the prescribed area of the reference image is an area including an area whose degree of similarity with a prescribed pattern image is equal to or greater than a prescribed value in the reference image.

8. An object identification method comprising:

a local descriptor matching step of determining whether or not respective descriptors of feature points extracted from an input image and respective descriptors of feature points extracted from a reference image correctly correspond to one another;

an input image difference area descriptor extracting step of extracting a descriptor of an area in the input image corresponding to a position of an image area obtained by performing a geometric transformation for correcting a geometric deviation between the input image and the reference image on a prescribed area of the reference image when a score based on the number of combinations of descriptors determined to correspond correctly in the determining step is equal to or larger than a prescribed value; and

a descriptor matching step of matching the descriptor extracted in the extracting step with a descriptor extracted from the prescribed area of the reference image, and outputting a matching result,

wherein a descriptor of the input image and a descriptor of the reference image that are used in a matching in the descriptor matching step are respectively parts of the descriptor of the input image and the descriptor of the reference image that have been used in the determination in the local descriptor matching step.

9. A non-transitory computer-readable medium that stores a program causing a computer to:

determine whether or not respective descriptors of feature points extracted from an input image and respective descriptors of feature points extracted from a reference image correctly correspond to one another;

extract a first descriptor of an area in the input image corresponding to a position of an image area obtained by performing a geometric transformation for correcting a geometric deviation between the input image and the reference image on a prescribed area of the reference image when a score based on the number of combinations of descriptors determined to correspond correctly is equal to or larger than a prescribed value;

match the first descriptor with a second descriptor extracted from the prescribed area of the reference image; and

output a matching result,

wherein the first descriptor and the second descriptor are respectively parts of the descriptor of the input image and the descriptor of the reference image that have been used in the determination of whether or not the descriptors of feature points extracted from an input image and respective descriptors of feature points extracted from a reference image correctly correspond to one another.