REGISTRATION OF 3D POINT CLOUD DATA BY CREATION OF FILTERED DENSITY IMAGES

• US 20090232388A1
• Filed: 03/12/2008
• Published: 09/17/2009
• Est. Priority Date: 03/12/2008
• Status: Abandoned Application

• Current Assignee: Harris Corporation (L3Harris Technologies, Inc.)
• Original Assignee: Harris Corporation (L3Harris Technologies, Inc.)
• Inventors: Minear, Kathleen, Blask, Steven G.
• Application Number:
  US12/046,862
• Publication Number:
  US 20090232388A1
• Time in Patent Office: Days
• US Class Current: 382/154
• CPC Class Codes: G06T 2207/10028 Range image; Depth image; 3D point clouds
• CPC Class Codes: G06T 2207/10032 Satellite or aerial image; Remote sensing
• CPC Class Codes: G06T 2207/20068 Projection on vertical or horizontal image axis
• CPC Class Codes: G06T 2207/30181 Earth observation
• CPC Class Codes: G06T 3/147 using affine transformations
• CPC Class Codes: G06T 7/32 using correlation-based methods

Abstract

Method (300) for registration of two or more of frames of three dimensional (3D) point cloud data (200-i, 200-j). A density image for each of the first frame (frame i) and the second frame (frame j) is used to obtain the translation between the images and thus image-to-image point correspondence. Correspondence for each adjacent frame is determined using correlation of the ‘filtered density’ images. The translation vector or vectors are used to perform a coarse registration of the 3D point cloud data in one or more of the XY plane and the Z direction. The method also includes a fine registration process applied to the 3D point cloud data (200-i, 200-j). Corresponding transformations between frames (not just adjacent frames) are accumulated and used in a ‘global’ optimization routine that seeks to find the best translation, rotation, and scale parameters that satisfy all frame displacements.

Claims

1. A method for registration of a plurality of frames of three dimensional (3D) point cloud data concerning a target of interest, comprising:

acquiring at least a first frame and a second frame, each containing 3D point cloud data collected for a selected object;

creating a density image for each of said first frame and said second frame respectively by projecting said 3D point cloud data from each of said first frame and said second frame to a two dimensional (2D) plane;

using said density images obtained from said first frame and said second frame to determine at least one translation vector;

performing a coarse registration of said 3D point cloud data in at least one of said XY plane and said Z plane using said at least one translation vector.

2. The method according to claim 1, further comprising exclusively selecting for registration a sub-volume of said 3D point cloud data from each frame which sub-volume includes less than a total volume of said 3D point cloud data.

3. The method according to claim 1, further comprising selecting said density images for each of said first frame and said second frame to be an XY density images by setting to zero a z coordinate value of each data point in a 3D point cloud contained in said first and second frame.

4. The method according to claim 1, further comprising selecting said density images for said first frame and said second frame to be XZ density images by setting to zero a y coordinate value of each data point in a 3D point cloud contained in said first and second frame.

5. The method according to claim 1, further comprising filtering each of said density images to obtain a filtered density image for each of said first frame and said second frame, prior to determining said translation vector.

6. The method according to claim 5, further comprising selecting said filtering to include a median filtering.

7. The method according to claim 5, further comprising selecting said filtering to include an edge enhancement filtering.

8. The method according to claim 5, wherein said step of determining said at least one translation vector further comprises performing a cross-correlation of said filtered density image obtained from said first frame and said filtered density image obtained from said second frame.

9. The method according to claim 8, further comprising determining said at least one translation vector based on a peak value resulting from the cross-correlation of said filtered density image from said first frame and said filtered density image of said second frame.

10. The method according to claim 1, further comprising performing a coarse registration of said 3D point cloud data from said first frame and said second frame in both said XY plane and in a Z axis direction.

11. The method according to claim 10, further comprising performing a fine registration process on said 3D point cloud data from said first frame and said second frame.

12. The method according to claim 11, wherein said fine registration process further comprises defining a plurality of sub-volumes within each of said first and second frames.

13. The method according to claim 12, wherein said fine registration process further comprises identifying one or more qualifying ones of said sub-volumes which include selected arrangements of 3D point cloud data.

14. The method according to claim 13, wherein said step of identifying qualifying ones of said sub-volumes further comprises calculating a set of eigen values for each of said sub-volumes.

15. The method according to claim 14, wherein said step of identifying qualifying ones of said sub-volumes further comprises calculating a set of eigen-metrics using said eigen values to identify sub-volumes containing 3D point clouds that have a blob-like arrangement.

16. The method according to claim 13, further comprising identifying qualifying data points in said qualifying ones of said sub-volumes.

17. The method according to claim 16, further comprising selecting said qualifying data points to include a plurality of pairs of data points, each said pair of data points comprising a first data point in said first frame that most closely matches a position of a corresponding second data point in said second frame.

18. The method according to claim 17, further comprising performing an optimization routine on said 3D point cloud data from said first frame and said second frame to determine a global rotation and translation vector applicable to all points in said second frame that minimizes an error as between said plurality of data point pairs.

19. A method for registration of a plurality of frames of three dimensional (3D) point cloud data concerning a target of interest, comprising:

acquiring at least a first frame and a second frame, each containing 3D point cloud data collected for a selected object;

creating a density image for each of said first frame and said second frame respectively by projecting said 3D point cloud data from each of said first frame and said second frame to a two dimensional (2D) plane;

using said density images obtained from said first frame and said second frame to determine at least one translation vector;

performing a coarse registration of said 3D point cloud data in at least one of said XY plane and said Z plane using said at least one translation vector;

selecting said density images for each of said first frame and said second frame to be XY density images formed by setting to zero a z coordinate value of each data point in a 3D point cloud contained in said first and second frame; and

selecting said density images for said first frame and said second frame to be XZ density images formed by setting to zero a y coordinate value of each data point in a 3D point cloud contained in said first and second frame.

20. A method for registration of a plurality of frames of three dimensional (3D) point cloud data concerning a target of interest, comprising:

acquiring at least a first frame and a second frame, each containing 3D point cloud data collected for a selected object;

creating a density image for each of said first frame and said second frame respectively by projecting said 3D point cloud data from each of said first frame and said second frame to a two dimensional (2D) plane;

using said density images obtained from said first frame and said second frame to determine at least one translation vector;

performing a coarse registration of said 3D point cloud data in at least one of said XY plane and said Z plane using said at least one translation vector;

selecting said density images for each of said first frame and said second frame to be XY density images formed by setting to zero a z coordinate value of each data point in a 3D point cloud contained in said first and second frame;

selecting said density images for said first frame and said second frame to be XZ density images formed by setting to zero a y coordinate value of each data point in a 3D point cloud contained in said first and second frame;

filtering each of said density images to obtain a filtered density image for each of said first frame and said second frame, prior to determining said translation vector.

21. The method according to claim 20, wherein said step of determining said at least one translation vector further comprises performing a cross-correlation of said filtered density image obtained from said first frame and said filtered density image obtained from said second frame.

22. The method according to claim 21, further comprising performing a fine registration process on said 3D point cloud data from said first frame and said second frame.

23. The method according to claim 22, wherein said fine registration process further comprises defining a plurality of sub-volumes within each of said first and second frames.

24. The method according to claim 23, wherein said fine registration process further comprises identifying one or more qualifying ones of said sub-volumes which include selected arrangements of 3D point cloud data.

25. The method according to claim 24, wherein said step of identifying qualifying ones of said sub-volumes further comprises calculating a set of eigen-metrics using said eigen values to identify sub-volumes containing 3D point clouds that have a blob-like arrangement.