Object symmetry axis detection method based on RGB-D camera

Patent metadata

Field	Value
Publication number	US-10607106-B2
Application number	US-201816093658-A
Country	US
Kind code	B2
Filing date	Apr 17, 2018
Priority date	Nov 21, 2017
Publication date	Mar 31, 2020
Grant date	Mar 31, 2020

Abstract

An object symmetry axis detection method based on an RGB-D camera includes: obtaining three-dimensional point cloud data of an image of a target object in natural scenes, identifying data corresponding to a point cloud of an area where the target object is located from the three-dimensional point cloud data according to a color difference threshold segmentation method, establishing a spherical coordinate system centering about a centroid of the target point cloud data, selecting n 2 candidate symmetry planes with a spherical coordinate system segmentation method, and calculating a symmetry axis of the target point cloud data according to the n 2 candidate symmetry planes as the symmetry axis of the target object.

First claim

What is claimed is:

1. An object symmetry axis detection method based on an RGB-D camera, comprising:
calibrating a color camera of the RGB-D camera and a depth camera of the RGB-D camera to obtain camera parameters of the RGB-D camera;
obtaining a color image of a target object with the color camera of the RGB-D camera and a depth image of the target object with the depth camera of the RGB-D camera;
mapping the depth image into a color image pixel coordinate system according to the camera parameters of the RGB-D camera to obtain an aligned depth image, wherein the color image is located in the color image pixel coordinate system; and processing the color image and the aligned depth image into three-dimensional point cloud data, wherein the three-dimensional point cloud data comprises three-dimensional coordinates and color information of a point cloud;
identifying target point cloud data from the three-dimensional point cloud data according to a color difference threshold segmentation method, wherein the target point cloud data is data corresponding to a first point cloud, the first point cloud is located in an area having the target object;
seeking a centroid of the target point cloud data, establishing a spherical coordinate system centering about the centroid, and selecting n 2 candidate symmetry planes with a spherical coordinate system segmentation method, wherein n is an integer and n is greater than 3;
calculating a symmetry axis of the target point cloud data according to the n 2 candidate symmetry planes, and determining the symmetry axis of the target point cloud data as a symmetry axis of the target object.

2. The method according to claim 1 , wherein the step of calculating the symmetry axis of the target point cloud data according to the n 2 candidate symmetry planes comprises the following substeps:
calculating a score of each candidate symmetry plane of the n 2 candidate symmetry planes according to a preset scoring strategy;
determining a candidate symmetry plane with a lowest score as a symmetry plane of the target object;
calculating the symmetry axis of the target point cloud data according to the symmetry plane of the target object.

3. The method according to claim 2 , wherein the step of calculating the symmetry axis of the target point cloud data according to the symmetry plane of the target object comprises calculating:

l
→

=

n
p

→

×

p
v

-

p
o



p
v

-

p
o



wherein {right arrow over (l )}is a symmetry axis vector of the target point cloud data, {right arrow over (n p )}is a normal vector of the symmetry plane of the target object, p v is coordinates of a viewpoint, p o is coordinates of the centroid of the target point cloud data.

4. The method according to claim 2 , wherein the step of calculating the score of each candidate symmetry plane of the n 2 candidate symmetry planes according to the preset scoring strategy comprises:
determining a candidate point for the each candidate symmetry plane, wherein the candidate point is a point having a symmetric partner in the each candidate symmetry plane;
calculating a score of each candidate point according to the preset scoring strategy;
determining a sum of the scores of all the candidate points as the score of the each candidate symmetry plane.

5. The method according to claim 4 , wherein, the step of calculating the score of each candidate point according to the preset scoring strategy, comprises:
calculating a symmetry point of the each candidate point with respect to the each candidate symmetry plane;
determining a neighbor point nearest to the symmetry point of the each candidate point according to a KNN classification algorithm, and determining a distance between the symmetry point of the each candidate point and the neighbor point;
calculating x score =d min +ω·α, wherein x score is the score of the each candidate point, d min is the distance between the symmetry point of the each candidate point and the neighbor point, α is an angle between a normal vector of the symmetry point of the each candidate point and a normal vector of the neighbor point, ω is a weight coefficient.

6. The method according to claim 4 , wherein the step of determining the candidate point for the each candidate symmetry plane comprises:
calculating (p v −p), p v is coordinates of a viewpoint, p is a point on one side of the each candidate symmetry plane, is a normal vector at the point;
when (p v −p)>0, determining the point to have a symmetric partner in the each candidate symmetry plane, the point is a candidate point.

7. The method according to claim 1 , wherein the step of identifying target point cloud data from the three-dimensional point cloud data according to the color difference threshold segmentation method comprises:
identifying foreground point cloud data from the three-dimensional point cloud data according to the color difference threshold segmentation method;
calculating

d
i

=

∑

j
=
1

k

⁢

⁢

1
k

⁢

(

x
i

-

x
j

)

2

+

(

y
i

-

y
j

)

2

+

(

Assignees

• Univ Jiangnan

Inventors

• Huang Min
• Li Hao
• Zhu Qibing
• Guo Ya

Classifications

• G06T2207/10024

  Color image · CPC title

• G06T2207/10028

  Range image; Depth image; 3D point clouds · CPC title

• G06T7/66

  of image moments or centre of gravity · CPC title

• G06T2207/30128

  Food products · CPC title

• G06T7/68 Primary

  of symmetry · CPC title