SIMULTANEOUS OBJECT DETECTION AND RIGID TRANSFORM ESTIMATION USING NEURAL NETWORK

• US 20180189580A1
• Filed: 02/23/2017
• Published: 07/05/2018
• Est. Priority Date: 01/03/2017
• Status: Active Grant

• Current Assignee: Qualcomm, Inc.
• Original Assignee: Qualcomm, Inc.
• Granted Patent: US 10,262,218 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 18/2413 based on distances to training or reference patterns
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G06T 2207/20084 Artificial neural networks [ANN]
• CPC Class Codes: G06T 2207/30252 Vehicle exterior; Vicinity of vehicle
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/13 Edge detection
• CPC Class Codes: G06T 7/168 involving transform domain methods
• CPC Class Codes: G06T 7/50 Depth or shape recovery
• CPC Class Codes: G06V 10/44 Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
• CPC Class Codes: G06V 10/454 Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
• CPC Class Codes: G06V 10/764 using classification, e.g. of video objects
• CPC Class Codes: G06V 10/82 using neural networks
• CPC Class Codes: G06V 20/582 of traffic signs

Abstract

A method, a computer-readable medium, and an apparatus for object detection are provided. The apparatus may determine a regression vector using a neural network based on an input image that contains an object. The object may have a planar surface with a known shape. The apparatus may derive a transform matrix based on the regression vector. The apparatus may identify a precise boundary of the object based on the transform matrix. The precise boundary of the object may include a plurality of vertices of the object. To identify the boundary of the object, the apparatus may apply the transform matrix to a determined shape of the object.

Claims

1. A method of object detection, comprising:

determining a regression vector using a neural network based on an input image comprising an object;

deriving a transform matrix based on the regression vector; and

identifying a boundary of the object based on the transform matrix.

2. The method of claim 1, wherein the boundary of the object comprises a plurality of vertices of the object.

3. The method of claim 1, wherein the regression vector comprises 6 dimensions for an affine transform.

4. The method of claim 1, wherein the regression vector comprises 8 dimensions for a perspective transform.

5. The method of claim 1, wherein the object has a planar surface with a known shape.

6. The method of claim 5, wherein the object is a traffic sign.

7. The method of claim 1, wherein the identifying the boundary of the object comprises applying the transform matrix to a shape of the object.

8. The method of claim 7, wherein the shape of the object is determined by the neural network.

9. The method of claim 7, wherein the identifying the boundary of the object is performed without a processing after the transform matrix is applied to the shape of the object.

10. The method of claim 9, wherein the processing comprises one or more of segmentation of the object, edge detection, or corner detection.

11. An apparatus for object detection, comprising:

means for determining a regression vector using a neural network based on an input image comprising an object;

means for deriving a transform matrix based on the regression vector; and

means for identifying a boundary of the object based on the transform matrix.

12. The apparatus of claim 11, wherein the boundary of the object comprises a plurality of vertices of the object.

13. The apparatus of claim 11, wherein the regression vector comprises 6 dimensions for an affine transform.

14. The apparatus of claim 11, wherein the regression vector comprises 8 dimensions for a perspective transform.

15. The apparatus of claim 11, wherein the object has a planar surface with a known shape.

16. The apparatus of claim 11, wherein the means for identifying the boundary of the object is configured to apply the transform matrix to a shape of the object.

17. The apparatus of claim 16, wherein the shape of the object is determined by the neural network.

18. The apparatus of claim 16, wherein the means for identifying the boundary of the object does not perform a processing after the transform matrix is applied to the shape of the object.

19. The apparatus of claim 18, wherein the processing comprises one or more of segmentation of the object, edge detection, or corner detection.

20. An apparatus for object detection, comprising:

a memory; and

at least one processor coupled to the memory and configured to;

determine a regression vector using a neural network based on an input image comprising an object;

derive a transform matrix based on the regression vector; and

identify a boundary of the object based on the transform matrix.

21. The apparatus of claim 20, wherein the boundary of the object comprises a plurality of vertices of the object.

22. The apparatus of claim 20, wherein the regression vector comprises 6 dimensions for an affine transform.

23. The apparatus of claim 20, wherein the regression vector comprises 8 dimensions for a perspective transform.

24. The apparatus of claim 20, wherein the object has a planar surface with a known shape.

25. The apparatus of claim 24, wherein the object is a traffic sign.

26. The apparatus of claim 20, wherein, to identify the boundary of the object, the at least one processor is configured to apply the transform matrix to a shape of the object.

27. The apparatus of claim 26, wherein the shape of the object is determined by the neural network.

28. The apparatus of claim 26, wherein, to identify the boundary of the object, the at least one processor does not perform a processing after the transform matrix is applied to the shape of the object.

29. The apparatus of claim 28, wherein the processing comprises one or more of segmentation of the object, edge detection, or corner detection.

30. A computer-readable medium storing computer executable code, comprising code to:

determine a regression vector using a neural network based on an input image comprising an object;

derive a transform matrix based on the regression vector; and

identify a boundary of the object based on the transform matrix.