System and method for providing position-independent pose estimation

• US 20050078178A1
• Filed: 10/10/2003
• Published: 04/14/2005
• Est. Priority Date: 10/10/2003
• Status: Active Grant

• Current Assignee: Hangzhou Hikvision Digital Technology Company Limited
• Original Assignee: International Business Machines Corporation
• Inventors: Hampapur, Arun, Connell, Jonathan H., Brown, Lisa Marie, Tian, Ying-Li, Senior, Andrew William, Bolle, Rudolf M., Pankanti, Sharathchandra
• Granted Patent: US 7,336,296 B2
• Time in Patent Office: Days
• US Class Current: 348/139
• CPC Class Codes: G06T 2207/10012 Stereo images
• CPC Class Codes: G06T 2207/30196 Human being; Person
• CPC Class Codes: G06T 2207/30244 Camera pose
• CPC Class Codes: G06T 7/70 Determining position or orientation of objects or cameras camera calibration G06T7/80
• CPC Class Codes: H04N 7/181 for receiving images from a plurality of remote sources

Abstract

A system and method to estimate the absolute coarse pose of a rigid body, and more specifically of an individual'"'"'s head as captured by overhead wide-angle cameras, by integrating position and pose information. Position information ideally includes three-dimensional position information provided by range finders or derived from images captured by multiple cameras. The three-dimensional position information can additionally be employed to determine the anchored viewing ray for the object.

Claims

1. A system for providing absolute orientation estimation of an object viewed by at least one camera in a space, the system comprising:

a. one or more cameras, each having a known camera orientation with respect to the space;

b. at least one detection component for detecting the object within one or more camera images and for providing object image location information;

c. at least one relative orientation estimation component for estimating the relative orientation of said object within each camera image using the object image location information; and

d. at least one absolute orientation estimation component for integrating the object image location information and the relative orientation estimate to produce an absolute orientation estimate.

2. The system of claim 1 further comprising at least one position determination component for providing the 3D object position information representing the position of the object in the space.

3. The system of claim 2 further comprising a view determination component which derives ananchored viewing ray of the object from said absolute orientation estimation and said 3D object position information.

4. The system of claim 2 wherein multiple cameras are viewing said object and wherein said 3D object position component derives 3D object position information using wide baseline stereo analysis.

5. The system of claim 2 wherein said position component comprises at least one range finder.

6. The system of claim 1 further comprising at least one roll correction component for correcting roll for the images obtained for each camera.

7. The system of claim 2 further comprising a composite absolute pose estimation component for integrating the absolute orientation estimates from each of said plurality of cameras and said 3D position information into a composite absolute orientation estimate.

8. The system of claim 1 wherein said at least one detection component additionally comprises means for performing background subtraction on each of said images.

9. The system of claim 4 wherein said wide baseline stereo analysis is provided by a triangulation component.

10. A method for providing absolute pose estimation for an object in at least one image captured by at least one camera comprising the steps of:

a. positioning one or more cameras each at a known camera orientation with respect to the space;

b. generating at least one camera image from said one or more cameras;

c. detecting the object within one or more camera images and providing object location information;

d. providing object position information representing the position of the object in the space;

e. estimating the relative pose of said object within each camera image using the object location information; and

f. integrating the object position information and the relative pose estimate to produce an absolute pose estimate.

11. The method of claim 10 wherein said providing object position information comprises deriving said object position information using the object location information and camera orientation information.

12. The method of claim 11 wherein said deriving comprises estimating a 2D line for the position of the object.

13. The method of claim 11 wherein multiple cameras are viewing said object and wherein said deriving comprises deriving 3D object position information using wide baseline stereo analysis.

14. The method of claim 11 wherein multiple cameras are viewing said object and wherein said deriving comprises providing 3D object position information.

15. The method of claim 14 wherein said position component comprises at least one range finder.

16. The method of claim 10 further comprising correcting roll for the images obtained for each camera.

17. The method of claim 14 further comprising integrating absolute pose estimates from each of said plurality of cameras and said 3D position information into a composite absolute pose estimate.

18. The method of claim 10 additionally comprising performing background subtraction on each of said images prior to said detecting.

19. The method of claim 14 further comprising deriving an anchored viewing ray based on said absolute pose estimate and said 3D object position information.

20. A program storage device readable by machine tangibly embodying a program of instructions for said machine to execute a method for providing absolute pose estimation for an object based on images captured by at least one camera in a space, said method comprising the steps of:

a. positioning one or more cameras each at a known camera orientation with respect to the space;

b. generating at least one camera image from said one or more cameras;

c. detecting the object within one or more camera images and providing object location information;

d. providing object position information representing the position of the object in the space;

e. estimating the relative pose of said object within each camera image using the object location information; and

f integrating the object position information and the relative pose estimate to produce an absolute pose estimate for each camera.