PRIMITIVE QUADRIC SURFACE EXTRACTION FROM UNORGANIZED POINT CLOUD DATA

• US 20110304619A1
• Filed: 08/03/2010
• Published: 12/15/2011
• Est. Priority Date: 06/10/2010
• Status: Abandoned Application

• Current Assignee: Autodesk, Inc.
• Original Assignee: Autodesk, Inc.
• Inventors: Fu, Yan, Yang, Jin, Zhu, Xiaofeng, Yuan, Zhenggang
• Application Number:
  US12/849,670
• Publication Number:
  US 20110304619A1
• Time in Patent Office: Days
• US Class Current: 345/420
• CPC Class Codes: G06T 17/00 Three dimensional [3D] modelling, e.g. data description of 3D objects
• CPC Class Codes: G06V 20/64 Three-dimensional objects

Abstract

A method, apparatus, system, article of manufacture, and data structure provide the ability to extract a primitive quadric surface from point cloud data. Point cloud data is obtained in 3D space. The point cloud data is segmented to create a disjoined surface and a smooth surface segment based on spatial connectivity and surface smoothness. One or more shapes are extracted from the point cloud data using geometric fitting. The geometric fitting searches for one or more quadric surface parameters of a given type of model that provides a best agreement between selected points from the point cloud data and a resultant model.

Claims

1. A computer implemented method for extracting a primitive quadric surface from point cloud data, comprising:

(a) obtaining point cloud data in three-dimensional (3D) space;

(b) segmenting the point cloud data to create a disjoined surface and a smooth surface segment based on spatial connectivity and surface smoothness;

(c) extracting one or more shapes from the point cloud data using geometric fitting;

wherein the geometric fitting comprises searching for one or more quadric surface parameters of a given type of model that provides a best agreement between one or more selected points from the point cloud data and a resultant model; and

(d) outputting the resultant model that includes the extracted shapes.

2. The computer implemented method of claim 1, wherein the segmenting comprises:

(a) selecting a point in the point cloud data with a minimum residual as a current seed;

(b) if all of the points in the point cloud data have been segmented, return a segmentation result;

(c) obtain neighboring points of the current seed using k-nearest neighbor processing to create a current region;

(d) adding one or more points from the point cloud that satisfy a condition to the current region;

(d) adding one or more points from the point cloud with residuals less than a threshold value to a potential seed points queue;

(e) if the potential seed points queue is not empty, select a top seed in the queue as the next seed and return to step (c); and

(f) adding the current region to the segmentation result and returning to step (a).

3. The computer implemented method of claim 1, wherein linear least square fitting is used to perform the geometric fitting for planes and spheres.

4. The computer implemented method of claim 1, wherein non-linear least square fitting is used to perform the geometric fitting for cylinders, cones, and tori.

5. The computer implemented method of claim 1, wherein:

the geometric fitting decomposes the one or more quadric surface parameters into two parts to reduce a dimension of parameter spaces;

the two parts comprise a first part that is solved by iterative optimization and a second part that is directly solved.

6. The computer implemented method of claim 5, wherein the quadric surface parameters for a cylinder comprises five parameters {right arrow over (v)}=(α

, β

, ρ

, γ

, r), where ρ

is an orthogonal distance from an origin to an axis of the cylinder, γ

is an angle between u and a unit normal vector (n=u sin γ

+v cos γ

) from the origin to the axis and r is a radius of the cylinder.

7. The computer implemented method of claim 6, wherein the five parameters are initially obtained by:

estimating a cylinder direction in a first step; and

estimating the radius and position of the cylinder in a second step.

8. The computer implemented method of claim 5, wherein the quadric surface parameters for a cone comprises six parameters (α

, β

, ρ

, γ

, l, ω

), where a=a(α

,β

) represents a unit vector of a cone axis, ρ

is an orthogonal distance from an origin to the cone axis, γ

is an angle between u and a unit normal vector (n=u sin γ

+v cos γ

) from the origin to the cone axis, p_o is a projected point of the origin on the cone axis, distance l is a position of the cone apex from an apex point c to p_o, and ω

, 0<

ω

<

π

/2 is an angle between the cone axis and a cone surface.

9. The computer implemented method of claim 8, wherein the six parameters are initially obtained by:

estimating the cone axis in a first step; and

estimating the position of the cone apex and the angle between the cone axis and the cone surface in a second step.

10. The computer implemented method of claim 5, wherein the quadric surface parameters for a torus comprises seven parameters (α

, β

, ρ

, γ

, l, r, R), wherein (α

, β

, ρ

, γ

) are used to define a position and direction of a symmetry axis of the torus, r and R are a minor radius and a major radius respectively, and l specifies a distance from a torus center and a projection point p_o of an origin to the symmetric axis.

11. The computer implemented method of claim 10, wherein the seven parameters are initially obtained by:

estimating the symmetry axis in a first step; and

estimating a radius center position, major radius, and minor radius in a second step.

12. A computer modeling system for extracting a primitive quadric surface from point cloud data, in a computer system comprising:

(a) a computer having a memory; and

(b) an application executing on the computer, wherein the application is configured to;

(i) obtain point cloud data in three-dimensional (3D) space;

(ii) segment the point cloud data to create a disjoined surface and a smooth surface segment based on spatial connectivity and surface smoothness;

(iii) extract one or more shapes from the point cloud data using geometric fitting;

wherein the geometric fitting comprises searching for one or more quadric surface parameters of a given type of model that provides a best agreement between one or more selected points from the point cloud data and a resultant model; and

(iv) output the resultant model that includes the extracted shapes.

13. The computer modeling system of claim 12, wherein the application is configured to segment by:

(a) selecting a point in the point cloud data with a minimum residual as a current seed;

(b) if all of the points in the point cloud data have been segmented, return a segmentation result;

(c) obtain neighboring points of the current seed using k-nearest neighbor processing to create a current region;

(d) adding one or more points from the point cloud that satisfy a condition to the current region;

(d) adding one or more points from the point cloud with residuals less than a threshold value to a potential seed points queue;

(e) if the potential seed points queue is not empty, select a top seed in the queue as the next seed and return to step (c); and

(f) adding the current region to the segmentation result and returning to step (a).

14. The computer modeling system of claim 12, wherein linear least square fitting is used to perform the geometric fitting for planes and spheres.

15. The computer modeling system of claim 12, wherein non-linear least square fitting is used to perform the geometric fitting for cylinders, cones, and tori.

16. The computer modeling system of claim 12, wherein:

the geometric fitting decomposes the one or more quadric surface parameters into two parts to reduce a dimension of parameter spaces;

the two parts comprise a first part that is solved by iterative optimization and a second part that is directly solved.

17. The computer modeling system of claim 16, wherein the quadric surface parameters for a cylinder comprises five parameters {right arrow over (v)}=(α

, β

, ρ

, γ

, r), where ρ

is an orthogonal distance from an origin to an axis of the cylinder, γ

is an angle between u and a unit normal vector (n=u sin γ

+v cos γ

) from the origin to the axis and r is a radius of the cylinder.

18. The computer modeling system of claim 17, wherein the five parameters are initially obtained by:

estimating a cylinder direction in a first step; and

estimating the radius and position of the cylinder in a second step.

19. The computer modeling system of claim 16, wherein the quadric surface parameters for a cone comprises six parameters (α

, β

, ρ

, γ

, l, ω

), where a=a(α

,β

) represents a unit vector of a cone axis, ρ

is an orthogonal distance from an origin to the cone axis, γ

is an angle between u and a unit normal vector (n=u sin γ

+v cos γ

) from the origin to the cone axis, p_o is a projected point of the origin on the cone axis, distance l is a position of the cone apex from an apex point c to p_o, and ω

, 0<

ω

<

π

/2 is an angle between the cone axis and a cone surface.

20. The computer modeling system of claim 19, wherein the six parameters are initially obtained by:

estimating the cone axis in a first step; and

estimating the position of the cone apex and the angle between the cone axis and the cone surface in a second step.

21. The computer modeling system of claim 16, wherein the quadric surface parameters for a torus comprises seven parameters (α

, β

, ρ

, γ

, l, r, R), wherein (α

, β

, ρ

, γ

) are used to define a position and direction of a symmetry axis of the torus, r and R are a minor radius and a major radius respectively, and l specifies a distance from a torus center and a projection point p_o of an origin to the symmetric axis.

22. The computer modeling system of claim 21, wherein the seven parameters are initially obtained by:

estimating the symmetry axis in a first step; and

estimating a radius center position, major radius, and minor radius in a second step.