Incremental learning for dynamic feature database management in an object recognition system

1. A method performed by a mobile imaging device, comprising:

obtaining one or more images of a scene;

recognizing that a target object is present in the one or more images based on features identified from the one or more images;

selecting a candidate feature from the features identified from the one or more images, the candidate feature including a feature not represented in a database associated with a set of features;

determining that the candidate feature belongs to the target object;

determining a feature score associated with the candidate feature;

determining at least one difference score of the candidate feature to at least one feature represented in the database associated with the set of features; and

updating the database associated with the set of features to represent the candidate feature including adding a representation of the candidate feature to the database based on the feature score satisfying a feature threshold, the candidate feature belonging to the target object, and the at least one difference score satisfying a difference threshold.

2. The method of claim 1 , wherein the target object comprises a portion of a user of the mobile imaging device present in the one or more images.

3. The method of claim 2 , wherein the portion of the user comprises a face of the user.

4. The method of claim 3 , wherein the set of features associated with the database are associated with the face of the user, and wherein the set of features associated with the database are identified from images previously obtained by the mobile imaging device.

5. The method of claim 1 , wherein the set of features associated with the database is stored in the database.

6. The method of claim 5 , further comprising:

determining a second set of features from the set of features stored in the database based on the features identified from the one or more images;

wherein determining the second set of features from the set of features stored in the database includes determining, from the set of features stored in the database, a feature that includes a descriptor determined to be similar to a descriptor of at least one feature of the features identified from the one or more images and a keypoint location that is proximate to a keypoint location of the at least one feature of the features identified from the one or more images.

7. The method of claim 5 , wherein updating the database associated with the set of features to represent the candidate feature includes adding the candidate feature to the set of features stored in the database.

8. A mobile imaging device, comprising:

a memory;

a processor coupled to the memory, the processor configured to:

obtain one or more images of a scene;

recognize that a target object is present in the one or more images based on features identified from the one or more images;

select a candidate feature from the features identified from the one or more images, the candidate feature including a feature not represented in a database associated with a set of features;

determine that the candidate feature belongs to the target object;

determine a feature score associated with the candidate feature;

determine at least one difference score of the candidate feature to at least one feature represented in the database associated with the set of features; and

update the database associated with the set of features to represent the candidate feature including adding a representation of the candidate feature to the database based on the feature score satisfying a feature threshold, the candidate feature belonging to the target object, and the at least one difference score satisfying a difference threshold.

9. The mobile imaging device of claim 8 , wherein the processor is further configured to determine depth information for the candidate feature from the one or more images, and wherein the processor is configured to use the depth information to determine that the candidate feature belongs to the target object.

10. The mobile imaging device of claim 8 , wherein the target object comprises a portion of a user of the mobile imaging device present in the one or more images.

11. The mobile imaging device of claim 10 , wherein the portion of the user comprises a face of the user.

12. The mobile imaging device of claim 11 , wherein the set of features associated with the database are associated with the face of the user, and wherein the set of features associated with the database are identified from images previously obtained by the mobile imaging device.

13. The mobile imaging device of claim 8 , wherein the set of features associated with the database is stored in the database.

14. The mobile imaging device of claim 13 , wherein the processor is further configured to:

determine a second set of features from the set of features stored in the database based on the features identified from the one or more images;

wherein determining the second set of features from the set of features stored in the database includes determining, from the set of features stored in the database, a feature that includes a descriptor determined to be similar to a descriptor of at least one feature of the features identified from the one or more images and a keypoint location that is proximate to a keypoint location of the at least one feature of the features identified from the one or more images.

15. The mobile imaging device of claim 13 , wherein the processor is configured to:

recognize that the target object is present in the one or more images based on at least one identified feature in a first set of features corresponding with at least one feature of the set of features stored in the database;

wherein adding the representation of the candidate feature to the database includes adding the candidate feature to the set of features stored in the database based on the feature score satisfying the feature threshold, the candidate feature belonging to the target object, and the at least one difference score satisfying the difference threshold.

16. The mobile imaging device of claim 13 , wherein updating the database associated with the set of features to represent the candidate feature includes adding the candidate feature to the set of features stored in the database.

17. A mobile imaging device, comprising:

a memory;

a processor coupled to the memory, the processor configured to:

obtain one or more images of a scene;

identify a first set of features from the one or more images;

compare the first set of features with a set of stored features existing in an object recognition system;

recognize that a target object is present in the one or more images based on at least one of the features in the first set of features corresponding with at least one feature of the set of stored features;

select a first candidate feature from the first set of features, the first candidate feature not among the features corresponding with the set of stored features;

determine a feature score of the first candidate feature;

determine that the first candidate feature belongs to the target object;

determine at least one difference score of the first candidate feature to at least one feature of the set of stored features; and

update the object recognition system to add the first candidate feature to the set of stored features based on: the feature score satisfying a feature threshold, the first candidate feature belonging to the target object, and the at least one difference score satisfying a difference threshold.

18. The mobile imaging device of claim 17 , wherein the target object comprises a portion of a user of the mobile imaging device present in the one or more images.

19. The mobile imaging device of claim 18 , wherein the portion of the user comprises a face of the user.

20. The mobile imaging device of claim 19 , wherein the feature score is based on at least one of: a feature local curvature, a feature edge strength, or a feature corner strength.

21. The mobile imaging device of claim 19 , wherein the object recognition system includes stored features associated with the face of the user, and wherein the stored features are identified from images previously obtained by the mobile imaging device.

22. The mobile imaging device of claim 21 , wherein the object recognition system resides on the mobile imaging device.

23. The mobile imaging device of claim 17 , wherein the processor is further configured to determine depth information for the first candidate feature from the one or more images, and wherein the processor is configured to use the depth information to determine that the first candidate feature belongs to the target object.

24. The mobile imaging device of claim 23 , wherein the processor is further configured to determine pose information for the mobile imaging device, and wherein the processor is configured to use the pose information to determine that the first candidate feature belongs to the target object.

25. The mobile imaging device of claim 17 , wherein the processor is configured to determine the at least one difference score based on at least one of: a difference in one or more keypoint locations of the first candidate feature and one or more corresponding keypoint locations of the at least one of the stored features, or a difference of a descriptor of the first candidate feature and a descriptor of a corresponding stored feature.

26. The mobile imaging device of claim 17 , wherein the processor is further configured to:

determine a second candidate feature from the stored features of the object recognition system that is similar to the first candidate feature; and

update the object recognition system to remove the second candidate feature from the stored features.

27. The mobile imaging device of claim 26 , wherein the second candidate feature is determined to be similar to the first candidate feature based on a comparison of descriptors of the first and second candidate features satisfying a descriptor threshold and a comparison of keypoint locations of the first and second candidate features satisfying a distance threshold.