Systems and methods for tracking an object

• US 9,665,804 B2
• Filed: 11/12/2014
• Issued: 05/30/2017
• Est. Priority Date: 11/12/2014
• Status: Active Grant

• Current Assignee: Qualcomm, Inc.
• Original Assignee: Qualcomm, Inc.
• Inventors: Qi, Yingyong, Mohamed, Magdi Abuelgasim, Sarkis, Michel Adib
• Primary Examiner(s): NAKHJAVAN, SHERVIN K
• Application Number:
  US14/539,617
• Publication Number:
  US 20160133022A1
• Time in Patent Office: 930 Days
• Field of Search: None
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/30201 Face
• CPC Class Codes: G06T 7/246 using feature-based methods, e.g. the tracking of corners or segments
• CPC Class Codes: G06T 7/74 involving reference images or patches

Abstract

A method for tracking an object by an electronic device is described. The method includes detecting an object position in an initial frame to produce a detected object position. The method also includes measuring one or more landmark positions based on the detected object position or a predicted object position. The method further includes predicting the object position in a subsequent frame based on the one or more landmark positions. The method additionally includes determining whether object tracking is lost. The method also includes avoiding performing object detection for the subsequent frame in a case that object tracking is maintained.

Claims

1. A method for tracking an object by an electronic device, comprising:

detecting an object position in an initial frame to produce a detected object position;

determining one or more landmark positions based on the detected object position or a predicted object position;

training a velocity model for object position prediction from landmarks of a previous frame, wherein the training comprises;

iteratively solving the velocity model until a displacement vector of the landmarks is lower than a predefined threshold;

predicting the object position in a subsequent frame based on the velocity model and the one or more landmark positions;

determining whether object tracking is lost in a current frame; and

avoiding performing object detection for the subsequent frame in response to object tracking being maintained for the current frame.

2. The method of claim 1, wherein the landmark positions are within a structure of the object.

3. The method of claim 1, wherein object position detection and object position prediction are independent.

4. The method of claim 1, wherein predicting the object position is based on one or more landmark positions from a previous frame.

5. The method of claim 1, wherein a structure of the landmark positions is predefined.

6. The method of claim 1, wherein determining whether object tracking is lost is based on at least one of one or more landmark positions and object position prediction.

7. The method of claim 1, wherein a velocity model is utilized during prediction and a displacement model is utilized while determining the one or more landmark positions, and wherein each of the velocity model and displacement model corresponds to a structure of the object.

8. The method of claim 7, wherein the velocity model and the displacement model do not include any hard constraints.

9. The method of claim 1, wherein the method operates on one frame at a time of a single image sequence.

10. The method of claim 1, further comprising training a displacement model for determining the one or more landmark positions from a predicted bounding box position at a current frame or from a detected bounding box position at the current frame.

11. An electronic device for tracking an object, comprising:

a processor;

memory in electronic communication with the processor; and

instructions stored in the memory, the instructions being executable by the processor to;

detect an object position in an initial frame to produce a detected object position;

determine one or more landmark positions based on the detected object position or a predicted object position;

train a velocity model for object position prediction from landmarks of a previous frame, wherein the instructions being executable to train comprise instructions being executable to;

iteratively solve the velocity model until a displacement vector of the landmarks is lower than a predefined threshold;

predict the object position in a subsequent frame based on the velocity model and the one or more landmark positions;

determine whether object tracking is lost in a current frame; and

avoid performing object detection for the subsequent frame in response to object tracking being maintained for the current frame.

12. The electronic device of claim 11, wherein the landmark positions are within a structure of the object.

13. The electronic device of claim 11, wherein object position detection and object position prediction are independent.

14. The electronic device of claim 11, wherein predicting the object position is based on one or more landmark positions from a previous frame.

15. The electronic device of claim 11, wherein a structure of the landmark positions is predefined.

16. The electronic device of claim 11, wherein determining whether object tracking is lost is based on at least one of one or more landmark positions and object position prediction.

17. The electronic device of claim 11, wherein a velocity model is utilized during prediction and a displacement model is utilized while determining the one or more landmark positions, and wherein each of the velocity model and displacement model corresponds to a structure of the object.

18. The electronic device of claim 17, wherein the velocity model and the displacement model do not include any hard constraints.

19. The electronic device of claim 11, wherein the electronic device operates on one frame at a time of a single image sequence.

20. The electronic device of claim 11, wherein the instructions are further executable by the processor to train a displacement model for determining the one or more landmark positions from a predicted bounding box position at a current frame or from a detected bounding box position at the current frame.

21. A computer-program product for tracking an object, comprising a non-transitory tangible computer-readable medium having instructions thereon, the instructions comprising:

code for causing an electronic device to detect an object position in an initial frame to produce a detected object position;

code for causing the electronic device to determine one or more landmark positions based on the detected object position or a predicted object position;

code for causing the electronic device to train a velocity model for object position prediction from landmarks of a previous frame, wherein the code for causing the electronic device to train a velocity model comprises;

code for causing the electronic device to iteratively solve the velocity model until a displacement vector of the landmarks is lower than a predefined threshold;

code for causing the electronic device to predict the object position in a subsequent frame based on the velocity model and the one or more landmark positions;

code for causing the electronic device to determine whether object tracking is lost in a current frame; and

code for causing the electronic device to avoid performing object detection for the subsequent frame in response to object tracking being maintained for the current frame.

22. The computer-program product of claim 21, wherein the landmark positions are within a structure of the object.

23. The computer-program product of claim 21, wherein object position detection and object position prediction are independent.

24. The computer-program product of claim 21, wherein the code for causing the electronic device to predict the object position is based on one or more landmark positions from a previous frame.

25. An apparatus for tracking an object, comprising:

means for detecting an object position in an initial frame to produce a detected object position;

means for determining one or more landmark positions based on the detected object position;

means for training a velocity model for object position prediction from landmarks of a previous frame, wherein the means for training comprise;

means for iteratively solving the velocity model until a displacement vector of the landmarks is lower than a predefined threshold;

means for predicting the object position in a subsequent frame based on the velocity model and the one or more landmark positions;

means for determining whether object tracking is lost in a current frame; and

means for avoiding performing object detection for the subsequent frame in response to object tracking being maintained for the current frame.

26. The apparatus of claim 25, wherein the landmark positions are within a structure of the object.

27. The apparatus of claim 25, wherein object position detection and object position prediction are independent.

28. The apparatus of claim 25, wherein the means for predicting the object position is based on one or more landmark positions from a previous frame.