3D object descriptors

• US 8,229,928 B2
• Filed: 02/27/2009
• Issued: 07/24/2012
• Est. Priority Date: 02/27/2009
• Status: Expired due to Fees

• Current Assignee: Empire Technology Development LLC (Allied Inventors Management, LLC)
• Original Assignee: Empire Technology Development LLC (Allied Inventors Management, LLC)
• Inventors: Yun, Il Dong
• Primary Examiner(s): Ehichioya, Fred I
• Assistant Examiner(s): MACKES, KRIS E
• Application Number:
  US12/395,441
• Publication Number:
  US 20100223299A1
• Time in Patent Office: 1,243 Days
• Field of Search: 707/737, 382/154
• US Class Current: 707/737
• CPC Class Codes: G06F 16/532 Query formulation, e.g. graphical querying
• CPC Class Codes: G06V 20/64 Three-dimensional objects

Abstract

Techniques for matching a 3D object are provided. In some embodiments, a 3D object retrieval apparatus includes an input unit configured to receive as input at least one sample view of a 3D query object to be inputted, and an object descriptor determining unit configured to generate at least one object representative image of the 3D query object from the at least one sample view and to determine a query object descriptor of the 3D query object from the at least one object representative image.

Claims

1. A 3D object computing apparatus comprising:

a microprocessor;

an input unit in communication with the microprocessor, configured to receive as input at least one sample view of a 3D query object to be inputted;

an object descriptor determining unit electrically coupled to the microprocessor and configured to generate at least one object representative image of the 3D query object from the at least one sample view and to determine a query object descriptor of the 3D query object from the at least one object representative image;

an object classifying unit electrically coupled to the microprocessor, the object classifying unit configured to;

retrieve candidate object descriptors of two or more 3D candidate objects each of the candidate object descriptors being associated with at least one object representative image;

compare the candidate object descriptors to obtain similarities therebetween;

classify the candidate object descriptors into at least one class based on the similarities;

use a clustering algorithm to generate at least one class representative image from the candidate object descriptors of the two or more 3D candidate objects;

determine at least one class descriptor, wherein determining the at least one class descriptor comprises determining at least one class image descriptor and concatenating the at least one class image descriptor; and

an object matching unit configured to;

compare the query object descriptor of the 3D query object with at least one class descriptor respectively identifying classes, each of the classes being associated with at least one candidate object descriptor of the 3D candidate objects, to find a matched class thereby;

compare the query object descriptor of the 3D query object with the at least one candidate object descriptor associated with the matched class to find a matched object thereby.

2. The apparatus of claim 1, further comprising a database that stores candidate object descriptors of 3D candidate objects and an object matching unit configured to compare the query object descriptor of the 3D query object with the candidate object descriptors of the 3D candidate objects to select at least one of the 3D candidate objects matched with the 3D query object as a matched object.

3. The apparatus of claim 2, wherein the database further stores object representative images of the 3D candidate objects, and the apparatus further including a class descriptor determining unit configured to generate at least one class representative image from the object representative images of the 3D candidate objects having candidate object descriptors classified into the at least one class to determine a class descriptor of the at least one class representative image of the at least one class.

4. The apparatus of claim 1, wherein the object descriptor determining unit is further configured to apply a clustering algorithm to the at least one sample view of the 3D query object to generate the at least one object representative image of the 3D query object.

5. The apparatus of claim 4, wherein the clustering algorithm includes an affinity propagation algorithm.

6. The apparatus of claim 1, wherein the object descriptor determining unit is further configured to determine an image descriptor for each of the at least one object representative image of the 3D query object and to determine the query object descriptor of the 3D query object based on the image descriptor.

7. The apparatus of claim 6, wherein the object descriptor determining unit is further configured to concatenate the image descriptor for each of the at least one object representative image of the 3D query object so as to determine the query object descriptor of the 3D query object.

8. A method performed under the control of an object retrieval apparatus, the method comprising:

receiving at least one sample view of a 3D query object;

using a clustering algorithm to generate at least one object representative image of the 3D query object based on the at least one sample view of the 3D query object;

determining an image descriptor for each of the at least one object representative image;

determining a query object descriptor of the 3D query object from the respective image descriptor for each of the at least one object representative image;

retrieving candidate object descriptors of two or more 3D candidate objects, each of the candidate object descriptors being associated with at least one object representative image;

comparing the candidate object descriptors to obtain similarities therebetween;

classifying the candidate object descriptors into at least one class based on the similarities;

using a clustering algorithm to generate at least one class representative image from the candidate object descriptors of the two or more 3D candidate objects;

determining at least one class descriptor, wherein determining the at least one class descriptor comprises determining at least one class image descriptor and concatenating the at least one class image descriptor;

comparing the query object descriptor of the 3D query object with at least one class descriptor respectively identifying classes, each of the classes being associated with at least one candidate object descriptor of the 3D candidate objects, to find a matched class thereby;

comparing the query object descriptor of the 3D query object with the at least one candidate object descriptor associated with the matched class to find a matched object thereby.

9. The method of claim 8 further comprising:

retrieving at least one object representative image of the matched object.

10. The method of claim 9, wherein comparing the candidate object descriptors includes:

measuring distance between the query object descriptor of the 3D query object and each of the candidate object descriptors of the 3D candidate objects.

11. The method of claim 10, wherein comparing the query object descriptor of the 3D query object further includes:

finding as the matched object, one of the 3D candidate objects having the candidate object descriptor with the shortest distance from the query object descriptor of the 3D query object.

12. A computer-readable storage medium whose contents cause a processor to:

receive at least one sample view of a 3D query object;

use a clustering algorithm to generate at least one object representative image of the 3D query object based on the at least one sample view of the 3D query object;

determine an image descriptor for each of the at least one object representative image;

determine a query object descriptor of the 3D query object from the respective image descriptor for each of the at least one object representative image;

determine a query object descriptor of the 3D query object from the candidate object descriptor for each of the at least one object representative image;

retrieve candidate object descriptors of at least one 3D candidate object, each of the candidate object descriptors being associated with at least one object representative image;

compare the candidate object descriptors to obtain similarities therebetween;

classify the candidate object descriptors into at least one class based on the similarities;

use a clustering algorithm to generate at least one class representative image from the candidate object descriptors of the two or more 3D candidate objects;

determine at least one class descriptor, wherein determining the at least one class descriptor comprises determining at least one class image descriptor and concatenating the at least one class image descriptor;

compare the query object descriptor of the 3D query object with at least one class descriptor respectively identifying classes, each of the classes being associated with at least one candidate object descriptor of the 3D candidate objects, to find a matched class thereby; and

compare the query object descriptor of the 3D query object with the at least one candidate object descriptor associated with the matched class to find a matched object thereby.

13. The computer-readable storage medium of claim 12, wherein the mathematical algorithm includes an affinity propagation algorithm.

14. The computer-readable storage medium of claim 12 wherein comparing the candidate object descriptors comprises measuring the distance between the query object descriptor of the 3D query object and each of the candidate object descriptors of the 3D candidate objects.

15. The computer-readable storage medium of claim 12 wherein finding a matched class comprises finding as the matched class one of the class descriptors having the shortest distance from the query object descriptor.

16. The computer readable-storage medium of claim 12 wherein finding a matched object comprises finding as the matched object one of the candidate object descriptors having the shortest distance from the query object descriptor.