Cognitive method for visual classification of very similar planar objects

• US 9,977,990 B1
• Filed: 01/08/2017
• Issued: 05/22/2018
• Est. Priority Date: 01/08/2017
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Marder, Mattias
• Primary Examiner(s): Seth, Manav
• Application Number:
  US15/401,040
• Time in Patent Office: 499 Days
• CPC Class Codes: G06F 16/532 Query formulation, e.g. graphical querying
• CPC Class Codes: G06F 16/5838 using colour
• CPC Class Codes: G06F 18/22 Matching criteria, e.g. proximity measures
• CPC Class Codes: G06F 18/23213 with fixed number of clusters, e.g. K-means clustering
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]
• CPC Class Codes: G06V 10/75 Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
• CPC Class Codes: G06V 10/763 Non-hierarchical techniques, e.g. based on statistics of modelling distributions

Abstract

A cognitive system and method for visual classification of similar planar objects is disclosed. The method uses a query image and a list of candidate templates as the input, and produces the most probable candidate for the query image. The system uses the mutual saliency among a sample of points in the query templates, and selects those points with the highest saliency. The corresponding candidate templates to the points with the highest saliency are then compared to the query image, and those with the highest similarity are kept. The system has applications to industrial and commercial settings where processes require object recognition from image data.

Claims

1. An apparatus for cognitive visual recognition, the apparatus comprising:

a memory device comprising a list of candidate templates;

an optical recording device configured to capture a digital query image and store the digital query image on the memory device; and

a processor coupled to the memory device and configured to;

retrieve the digital query image from the memory device,

retrieve the list of candidate templates from the memory device,

align each candidate template in the list of candidate templates with the digital query image,

select a set of sample points of each candidate template in the list of candidate templates,

measure a mutual saliency of the set of sample points,

select a set of highest mutual saliency points from the set of sample points based on the mutual saliency,

measure the similarity of each candidate template in the list of candidate templates to the query image at the set of highest mutual saliency points, and

generate a list of highest similarity candidates based on the measured similarity.

2. The apparatus of claim 1, wherein selecting a set of highest mutual saliency points from the set of sample points comprises retaining a percentage of the set of sample points.

3. The apparatus of claim 1, wherein selecting a set of highest mutual saliency points from the set of sample points comprises K-means clustering.

4. The apparatus of claim 1, wherein the digital query image comprises a representation of a planar object.

5. The apparatus of claim 1, wherein the digital query image comprises a region of interest.

6. An apparatus for cognitive visual recognition, the apparatus comprising:

a memory storage device comprising a list of candidate templates; and

a digital imager configured to;

capture a digital image, and

store the digital image on the memory storage device; and

a processor configured to;

receive the digital image,

receive the list of candidate templates,

align each candidate template in the list of candidate templates with the digital image,

select a set of sample points of each candidate template in the list of candidate templates and measure their mutual saliency,

select a set of highest mutual saliency points from the set of sample points, measure the similarity of each candidate template in the list of candidate templates to the image at the set of highest mutual saliency points, and

remove at least a lowest similarity candidate from the list of candidate templates.

7. The apparatus of claim 6, wherein selecting a set of highest saliency points from the set of sample points comprises retaining a percentage of the points.

8. The apparatus of claim 6, wherein selecting a set of highest saliency points from the set of sample points comprises K-means clustering.

9. The apparatus of claim 6, wherein the digital image comprises a representation of a planar object.

10. The apparatus of claim 6, wherein the digital image comprises a region of interest.

11. A method for cognitive visual recognition comprising:

a. receiving a digital query image from a digital imaging device;

b. receiving a list of candidate templates;

c. aligning the list of candidate templates with the digital query image;

d. selecting a set of sample points in the list of candidate templates and measuring their mutual saliency;

e. selecting a set of highest mutual saliency points from the set of sample points;

f. measuring a similarity of each candidate template in the list of candidate template to the digital query image at the set of highest mutual saliency points; and

g. generating a list of highest similarity candidates based on the measured similarity.

12. The method of claim 11, wherein selecting a set of highest saliency points from the set of sample points comprises keeping a percentage of the points.

13. The method of claim 11, wherein selecting a set of highest saliency points from the set of sample points comprises K-means clustering.

14. The method of claim 11, wherein measuring the similarity of candidate templates comprises measuring an image feature, wherein the image feature is selected from the group consisting of gray value, color value, and SIFT.

15. The method of claim 11, further comprising performing steps a through g iteratively until a termination condition is reached.

16. The method of claim 15, wherein the termination condition is one candidate remaining.

17. The method of claim 15, wherein the termination condition is two candidates remaining.

18. The method of claim 15, wherein the termination condition is a top candidate is significantly more similar than any other hypotheses.

19. The method of claim 11, wherein the digital query image comprises a representation of a planar object.

20. The method of claim 11, wherein the digital query image comprises a region of interest.