Object category recognition methods and robots utilizing the same

1. A method for recognizing a category of an object, the method comprising:

calculating a confidence score for a plurality of categories;

determining, by a processor, a preliminary category of a target object, wherein a highest confidence score is associated with the preliminary category;

comparing the highest confidence score to a learning threshold;

if the highest confidence score is less than the learning threshold, estimating properties of the target object and generating a property score for one or more estimated properties of the target object, wherein the property score is different from the confidence score; and

searching a supplemental image collection for supplemental image data using the preliminary category and the one or more estimated properties.

2. The method of claim 1 , wherein the preliminary category of the target object and the highest confidence score is determined by:

obtaining target image data of the target object;

extracting, by a computer, a set of features from the target image data;

comparing the extracted set of features to library features associated with the plurality of categories of an image library stored in a database; and

selecting a category having the highest confidence score as the preliminary category of the target object.

3. The method of claim 2 , further comprising supplementing the image library with retrieved supplemental image data.

4. The method of claim 3 , further comprising determining the preliminary category of the target object using the image library containing the retrieved supplemental image data.

5. The method of claim 3 , further comprising extracting a set of supplemental image data features from the retrieved supplemental image data.

6. The method of claim 1 , wherein the preliminary category of the target object and the confidence score is determined by a scale-invariant feature transform process.

7. The method of claim 6 , wherein the confidence score comprises a degree of correlation resulting from the scale-invariant feature transform process.

8. The method of claim 1 , wherein the one or more estimated properties comprise one or more of a color property or an object pose property.

9. The method of claim 8 , wherein the color property is estimated by evaluating a red value, a green value, and a blue value of individual ones of a plurality of pixels of the target image data, and the color property is an average color of the plurality of pixels.

10. The method of claim 8 , wherein the object pose property is determined by a scale-invariant feature transform process.

11. The method of claim 1 , wherein the supplemental image collection is searched by:

generating a search query based at least in part on the preliminary category and the one or more estimated properties having a property score that is greater than an estimated property threshold; and

searching the supplemental image collection for supplemental image data using the search query.

12. The method of claim 1 , wherein the supplemental image collection is defined by a network comprising a plurality of linked databases.

13. A method for recognizing a category of an object, the method comprising:

determining, by a processor, a preliminary category of a target object and a confidence score associated with the preliminary category, the preliminary category and the confidence score determined by:

obtaining target image data of the target object;

extracting, by the processor, a set of features from the target image data;

comparing the extracted set of features to library features associated with a plurality of categories of an image library stored in a database, and generating the confidence score for one or more categories of the plurality of categories; and

selecting the category having a highest confidence score as the preliminary category of the target object;

comparing the confidence score to a learning threshold; and

if the highest confidence score is less than the learning threshold:

estimating properties of the target object and generating a property score for one or more estimated properties;

comparing the property score for the one or more estimated properties with an estimated property threshold;

generating a search query based at least in part on the preliminary category and the one or more estimated properties having a property score that is greater than the estimated property threshold;

searching the supplemental image collection for supplemental image data using the search query; and

supplementing the image library with retrieved supplemental image data.

14. The method of claim 13 , wherein the preliminary category of the target object and the confidence score is determined by a scale-invariant feature transform process.

15. The method of claim 14 , wherein the confidence score comprises a degree of correlation resulting from the scale-invariant feature transform process.

16. The method of claim 13 , wherein the one or more estimated properties comprise one or more of a color property or an object pose property.

17. The method of claim 16 , wherein the color property is estimated by evaluating a red value, a green value, and a blue value of individual ones of a plurality of pixels of the target image data and the color property is a color associated with color property is an average color of the plurality of pixels.

18. A robot comprising:

an image capturing device;

a processor;

a computer-readable storage medium comprising instructions that, when executed by the processor, causes the processor to:

control the image capturing device to acquire target image data of a target object;

calculate a confidence score for a plurality of categories;

determine a preliminary category of the target object, wherein a highest confidence score is associated with the preliminary category;

compare the confidence score to a learning threshold;

if the highest confidence score is less than the learning threshold:

estimate properties of the target object and generate a property score for one or more estimated properties of the target object wherein the property score is different from the confidence score; and

retrieve supplemental image data from a supplemental image collection using the preliminary category and the one or more estimated properties as search criteria.

19. The robot of claim 18 , wherein the preliminary category of the target object and the highest confidence score is determined by:

obtaining target image data of the target object;

extracting a set of features from the target image data;

comparing the extracted set of features to library features associated with the plurality of categories of an image library stored in a database that is accessible by the processor; and

selecting a category having the highest confidence score as the preliminary category of the target object.

20. The robot of claim 18 , wherein the set of instructions further cause the processor to generate a search query based at least in part on the preliminary category and the one or more estimated properties having a property score that is greater than an estimated property threshold, and search the supplemental image collection for supplemental image data using the search query.