Invariant-Based Dimensional Reduction Of Object Recognition Features, Systems And Methods

• US 20150294188A1
• Filed: 02/19/2015
• Published: 10/15/2015
• Est. Priority Date: 02/19/2014
• Status: Active Grant

• Current Assignee: Nant Holdings IP, LLC (NantWorks, LLC)
• Original Assignee: Nant Holdings IP, LLC (NantWorks, LLC)
• Inventors: Wnuk, Kamil, Sudol, Jeremi, Song, Bing, Siddiqui, Matheen, McKinnon, David
• Granted Patent: US 9,460,366 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 18/2113 by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
• CPC Class Codes: G06F 18/213 Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]
• CPC Class Codes: G06V 10/771 Feature selection, e.g. selecting representative features from a multi-dimensional feature space

Abstract

A sensor data processing system and method is described. Contemplated systems and methods derive a first recognition trait of an object from a first data set that represents the object in a first environmental state. A second recognition trait of the object is then derived from a second data set that represents the object in a second environmental state. The sensor data processing systems and methods then identifies a mapping of elements of the first and second recognition traits in a new representation space. The mapping of elements satisfies a variance criterion for corresponding elements, which allows the mapping to be used for object recognition. The sensor data processing systems and methods described herein provide new object recognition techniques that are computationally efficient and can be performed in real-time by the mobile phone technology that is currently available.

Claims

1. A sensor data processing system comprising:

a controlled sensing environment having environmental parameters having corresponding environmental attributes; and

an image processing engine programmed to;

obtain a first training data set representative of at least one object under a defined environmental state within the controlled environment;

derive a first recognition trait from the first training data set according to a trait extraction algorithm, the first recognition trait comprising a first plurality of elements;

configure the controlled imaging environment to create a second environmental state by varying an environmental attribute through adjusting a corresponding environment parameter;

obtain a second training data set representative of the at least one object under the second environmental state within the controlled environment with respect to the adjusted environmental attribute;

derive a second recognition trait from the second training data set according to the trait extraction algorithm, the new recognition trait comprising a second plurality of elements, and wherein the new trait has a correspondence to the first trait;

identify a mapping that maps the plurality of elements of first and second recognition traits to a new representation space, wherein the mapping of the plurality of elements in the new representation space satisfies trait element variance criteria among corresponding elements in the traits across the first and second training set; and

store the mapping in a memory.

2. The system of claim 1, wherein the controlled environment comprises controlled lighting.

3. The system of claim 1, wherein the controlled environment comprises controlled acoustic properties.

4. The system of claim 1, wherein the trait extraction algorithm comprises an image processing algorithm.

5. The system of claim 4, wherein the image processing algorithm comprises at least one of the following:

SIFT, FAST, FREAK, BRIEF, ORB, BRISK, GLOH, SURF, vSLAM, SLAM, BURST, and DAISY.

6. The system of claim 1, wherein the trait extraction algorithm comprises an automatic speech recognition algorithm.

7. The system of claim 1 wherein the first recognition trait comprise a descriptor.

8. The system of claim 7, wherein the descriptor comprises an image descriptor.

9. The system of claim 8, wherein the first plurality of elements comprise dimensions of the descriptor.

10. The system of claim 1, wherein the first recognition trait comprises a cluster of descriptors.

11. The system of claim 10, wherein the cluster of descriptors comprises a constellation of descriptors.

12. The system of claim 11, wherein the constellation is within a descriptor space.

13. The system of claim 11, wherein the constellation is with respect to 3D space.

14. The system of claim 10, wherein the first plurality of elements comprise at least one of the following:

a descriptor, a descriptor location, and a cluster of descriptors.

15. The system of claim 1, wherein the image processing engine is further programmed to generate a trait vocabulary as a function of descriptor clusters derived from at least the first training data set, the trait vocabulary comprising a corpus of vocabulary atoms representing traits.

16. The system of claim 15, wherein at least one of the first recognition trait and the second recognition trait are associated with at least one of the vocabulary atoms.

17. The system of claim 15, wherein the corpus of vocabulary atoms representing traits comprises at least one cluster shape trait.

18. The system of claim 1, wherein the mapping comprises a dimensionality reduction of the plurality of elements.

19. The system of claim 1, further comprising an object categorization engine configured to classify the at least one object as a type of object based on at least one of the first recognition trait and the second recognition trait.

20. The system of claim 19, wherein the object categorization engine is further configured to store the at least one of the first recognition trait and the second recognition trait with the type of object in a known object database.

21. The system of claim 1, wherein the mapping represents an invariant property as the recognition property with respect to the adjusted environmental parameters as exhibited by at least some of the first plurality of elements.

22. The system of claim 1, wherein the mapping represents a variant property as the recognition property with respect to the adjusted environmental parameters as exhibited by at least some of the first plurality of elements.

23. The system of claim 1, wherein the mapping comprises a non-linear mapping from the first plurality of elements and the second plurality of elements to a third plurality of elements in a new multi-dimensional space.

24. The system of claim 1, wherein the mapping comprises a look-up table.

25. The system of claim 1, wherein the mapping comprises a linear mapping from the first and second plurality of elements to a third plurality of elements in a new invariant space.

26. The system of claim 1, wherein the trait element variance criteria is a function of a variance with respect to the first plurality of elements.

27. The system of claim 1, wherein the mapping includes an inferred state of a real-world environment based on the second recognition trait.

28. The system of claim 1, wherein the adjusted environmental attribute comprises at least one of the following:

a wireless signal, a lighting property, a camera property, a sensor property, an error rate, a gravity field, a shape, a distance, a field of view, a scale, a time, and an orientation.

29. The system of claim 1, wherein the defined environmental state comprises an expected environment state external to the controlled environment.

30. The system of claim 1, where the trait element variance criteria identifies a low variance among the corresponding elements in the traits across the first and second training set, the low variance operating as a function of a low variance threshold.

31. The system of claim 1, where the trait element variance criteria identifies a high variance among the corresponding elements in the traits across the first and second training set, the high variance operating as a function of a high variance threshold.