Apparatus for recognizing object in image

• US 20080247651A1
• Filed: 04/07/2008
• Published: 10/09/2008
• Est. Priority Date: 04/09/2007
• Status: Active Grant

• Current Assignee: DENSO Corporation, Hironobu Fujiyoshi
• Original Assignee: DENSO Corporation, Hironobu Fujiyoshi
• Inventors: Takaki, Masanari, Fujiyoshi, Hironobu
• Granted Patent: US 8,027,514 B2
• Time in Patent Office: Days
• US Class Current: 382/219
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]
• CPC Class Codes: G06V 20/582 of traffic signs

Abstract

An object recognition apparatus sets each of keypoints extracted from a typical image and one of keypoints extracted from an input image, having scale invariant features similar to each other, as a typical corresponding point and an object corresponding point matching with each other, and produces a position vector directed from a typical reference point of the typical image to each typical keypoint. The apparatus determines a position of an object reference point in the input image from a position of each object corresponding point and the vector of the typical corresponding point matching with the object corresponding point. When the positions of the object reference point are concentrated, the apparatus judges that an object picture having the object corresponding points in the input image matches with the typical image, and the apparatus recognizes the picture as an object indicated by the typical image.

Claims

1. An object recognition apparatus comprising:

a typical feature point extracting unit that extracts a plurality of typical feature points from a typical image;

a first feature calculating unit that calculates a feature of each typical feature point from image data of the typical image;

an object feature point extracting unit that extracts a plurality of object feature points from an input image;

a second feature calculating unit that calculates a feature of each object feature point from image data of the input image;

a corresponding point setting unit that calculates similarities between the feature of one typical feature point and the features of the object feature points for each typical feature point, judges based on the similarities whether or not the features of both one typical feature point and one object feature point in each pair are similar to each other, and sets one typical feature point and one object feature point in each pair, having the features similar to each other, as a typical corresponding point and an object corresponding point matching with each other;

an information producing unit that produces position information indicating positions of the typical feature points in the typical image;

a judging unit that judges based on the position information of the typical corresponding points whether or not a positional relationship between or among the object corresponding points is substantially similar to a positional relationship between or among the typical corresponding points; and

a recognizing unit that recognizes an object picture having the object corresponding points in the input image as an object indicated by the typical image when the judging unit judges that the positional relationships are substantially the same as each other.

2. The apparatus according to claim 1, wherein the information producing unit is adapted to set a typical reference point in the typical image and to set a position vector indicating a positional difference between the typical reference point and each typical feature point as the position information of the typical feature point, the judging unit is adapted to determine a position of an object reference point matching with the typical reference point in the input image from the position of each object corresponding point and the position vector of the typical corresponding point matching with the object corresponding point and to judge whether or not the positions of the object reference point are concentrated at a concentration higher than a predetermined concentration value, and the recognizing unit is adapted to recognize the object when the positions of the object reference point are concentrated at a concentration higher than the predetermined concentration value.

3. The apparatus according to claim 2, wherein the judging unit is adapted to set the predetermined concentration value in accordance with the number of typical feature points.

4. The apparatus according to claim 2, wherein the judging unit is adapted to set a small region including one position of the object reference point in the input image each time the position of the object reference point is determined by one object corresponding point on condition that the position of the object reference point is not placed in any of other small regions already set, and the judging unit is adapted to judge that the positions of the object reference point are concentrated at a concentration higher than the predetermined concentration value when the positions of the object reference point are placed in one of the small regions at a predetermined possibility or more.

5. The apparatus according to claim 2, wherein

the judging unit is adapted to adjust the position vector of the typical corresponding point matching with each object corresponding point to an adjusted position vector so as to compensate a scale difference or a rotation difference between the input image and the typical image, and to determine the position of the object reference point from the position of each object corresponding point and the adjusted position vector of the typical corresponding point matching with the object corresponding point.

6. The apparatus according to claim 1, wherein the information producing unit is adapted to set typical position vectors indicating relative positions of one typical feature point to the other typical feature points as the position information of the typical feature point for each typical feature point, the judging unit is adapted to calculate object position vectors indicating relative positions of each object corresponding point to the other object corresponding points, to determine the positional relationship among the typical corresponding points from the typical position vectors of the typical corresponding points, to determine the positional relationship among the object corresponding points from the object position vectors of the object corresponding points, and to judge whether or not the relationships are substantially similar to each other, and the recognizing unit is adapted to recognize the object in response to the judgment of the judging unit.

7. The apparatus according to claim 6, wherein the judging unit is adapted to select three of the typical corresponding points, to set a typical triangle determined by the three typical corresponding points, to set an object triangle determined by three object corresponding points matching with the three typical corresponding points, to judge based on the typical position vectors of the three typical corresponding points and the object position vectors of the three object corresponding points whether or not the object triangle is substantially similar to the typical triangle, and the recognizing unit is adapted to recognize the object when the object triangle is substantially similar to the typical triangle.

8. The apparatus according to claim 1, wherein the object picture of the input image has a scale or a rotation different from that of the typical image.

9. An object recognition apparatus comprising:

a typical feature point extracting unit that extracts a plurality of typical feature points from a typical image;

a first feature calculating unit that calculates a feature of each typical feature point from image data of the typical image;

an object feature point extracting unit that extracts a plurality of object feature points from an input image;

a second feature calculating unit that calculates a feature of each object feature point from image data of the input image;

a corresponding point setting unit that calculates similarities between the feature of one typical feature point and the features of the object feature points for each typical feature point, judges based on the similarities whether or not the features of both one typical feature point and one object feature point in each pair are similar to each other, and sets one typical feature point and one object feature point in each pair, having the features similar to each other, as a typical corresponding point and an object corresponding point matching with each other;

a judging unit that judges whether or not a ratio of the number of object corresponding points to the number of typical feature points is higher than a predetermined value; and

a recognizing unit that recognizes an object picture having the object corresponding points in the input image as an object indicated by the typical image when the judging unit judges that the ratio is higher than the predetermined value.