Method for representing objects with concentric ring signature descriptors for detecting 3D objects in range images
First Claim
1. A method for representing a 3D object with a descriptor, wherein a model of the 3D object is a 3D point cloud, comprising the steps of:
- locating a local support for each point p in the 3D point cloud, wherein the local support is a spherical volume S of points pi, centered on the point p and within a radius r of the point p according to
S={pi;
∥
pi−
p∥
≦
r}, generating reference x, y, and z axes for the local support;
applying, according to the references x, y, and z axes, a polar grid along an azimuth and a radial directions on an xy plane centered on the point p such that each patch on the grid is a bin for a 2D histogram, wherein the 2D histogram is a 2D matrix F on the grid and each coefficient of the 2D matrix F corresponds to the patch on the grid; and
estimating, for each grid location (k, l), an elevation value F(k, l) by interpolating the elevation values of the 3D points within the patches to produce the descriptor for the point p, wherein the steps are performed in a processor.
1 Assignment
0 Petitions
Accused Products
Abstract
A 3D object is represented by a descriptor, wherein a model of the 3D object is a 3D point cloud. A local support for each point p in the 3D point cloud is located, and reference x, y, and z axes are generated for the local support. A polar grid is applied according to the references x, y, and z axes a along an azimuth and a radial directions on an xy plane centered on the point p such that each patch on the grid is a bin for a 2D histogram, wherein the 2D histogram is a 2D matrix F on the grid and each coefficient of the 2D matrix F corresponds to the patch on the grid. For each grid location (k, l), an elevation value F(k, l) is estimated by interpolating the elevation values of the 3D points within the patches to produce the descriptor for the point p.
19 Citations
20 Claims
-
1. A method for representing a 3D object with a descriptor, wherein a model of the 3D object is a 3D point cloud, comprising the steps of:
-
locating a local support for each point p in the 3D point cloud, wherein the local support is a spherical volume S of points pi, centered on the point p and within a radius r of the point p according to
S={pi;
∥
pi−
p∥
≦
r},generating reference x, y, and z axes for the local support; applying, according to the references x, y, and z axes, a polar grid along an azimuth and a radial directions on an xy plane centered on the point p such that each patch on the grid is a bin for a 2D histogram, wherein the 2D histogram is a 2D matrix F on the grid and each coefficient of the 2D matrix F corresponds to the patch on the grid; and estimating, for each grid location (k, l), an elevation value F(k, l) by interpolating the elevation values of the 3D points within the patches to produce the descriptor for the point p, wherein the steps are performed in a processor. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19)
-
-
20. A method for representing an object with a descriptor, comprising the steps of:
-
locating a local support for each point in a point cloud representing the object; generating reference axes for the local support; applying, according to the references axes, a polar grid along an azimuth and a radial directions on a plane centered on the point such that each patch on the grid is a bin for a histogram, wherein the histogram is a matrix on the grid, and each coefficient of the matrix corresponds to the patch on the grid; and estimating, for each grid location, an elevation value by interpolating the elevation values of the points within the patches to produce the descriptor for the point, wherein the steps are performed in a processor.
-
Specification