Method and a device for extracting local features of a three-dimensional point cloud

Patent metadata

Field	Value
Publication number	US-10339409-B2
Application number	US-201515575897-A
Country	US
Kind code	B2
Filing date	Jun 18, 2015
Priority date	Jun 18, 2015
Publication date	Jul 2, 2019
Grant date	Jul 2, 2019

Abstract

A method and a device for extracting local features of a 3D point cloud are disclosed. Angle information and the concavo-convex information about a feature point to be extracted and a point of an adjacent body element are calculated based on a local reference system corresponding to the points of each body element. The feature relation between the two points can be calculated accurately. The property of invariance in translation and rotation is possessed. Since concavo-convex information about a local point cloud is contained during extraction, the inaccurate extraction caused by ignoring concavo-convex ambiguity in previous 3D local feature description is resolved. During normalization processing, exponential normalization processing and second-normal-form normalization are adopted, which solves the problem of inaccurate similarity calculation caused by a circumstance that a few elements in a vector are too large or too small during feature extraction, thus improving accuracy of extracted three-dimensional local features.

First claim

What is claimed is:

1. A method for extracting local features of a 3D point cloud, comprising:
determining the local reference system corresponding to the points of each body element, comprising:
calculating a covariance matrix M;
decomposing the covariance matrix M to obtain three feature vectors;
sorting the three feature vectors in descending order as the roll axis x, the heading axis y and the pitch axis z of the local reference system respectively; and
aligning the three feature vectors for de-ambiguity calculation, to obtain the local reference system corresponding to the points of each body element,
wherein the covariance matrix M is calculated using

M
=

1
Z

⁢

∑

i
:

d
i

≤
R

⁢

(

R
-

d
i

)

⁢

(

p
′

-
p

)

⁢

(

p
′

-
p

)

,

(
1
)

wherein R is the radius of the point cloud sphere, p′ is a point of each body element, p is a local feature point,

d
i

=



p
′

-
p



2

⁢

⁢
and
⁢

⁢
Z

=

∑

i
:

d
i

≤
R

⁢

(

R
-

d
i

)

;

calculating angle information about a local feature point to be extracted and points of each body element in a pre-set point cloud sphere;
calculating concavo-convex information about a curved surface between the local feature point to be extracted and the points of each body element respectively, wherein the pre-set point cloud sphere contains various body elements, and the body elements are adjacent to the local feature point to be extracted;
computing histogram statistics according to the angle information and the concavo-convex information;
generating histograms each corresponding to each body element;
connecting the histograms corresponding to the body elements in the pre-set point cloud sphere on a one-to-one basis, to obtain an extracted vector; and
performing exponential normalization processing and second-normal-form normalization processing on the extracted vector.

2. The method of claim 1 , further comprising:
before the step of calculating angle information about a local feature point, constructing a point cloud sphere with the local feature point to be extracted as a center and a pre-set length as the radius; and
dividing the point cloud sphere along the direction angle, the elevation angle and the radius to obtain a number of body elements adjacent to the local feature point to be extracted.

3. The method of claim 1 , wherein the step of calculating angle information about a local feature point to be extracted and points of each body element in a pre-set point cloud sphere comprises:
determining an angle α between the roll axis of the local reference system corresponding to the points of each body element and the roll axis of the local reference system corresponding to the local feature point, an angle β between the heading axis of the local reference system and the heading axis of the local reference system corresponding to the local feature point, and an angle θ between the pitch axis local of the reference system and the pitch axis of the local reference system corresponding to the local feature point;
calculating cosines of the angles α, β and θ to obtain cos α, cos β and cos θ;
computing a mean of the cosines values to obtain angle information τ of the point of the body element using:

τ
=

cos
⁢

⁢
α

+

Assignees

• Univ Peking Shenzhen Graduate School

Inventors

• Wang Wenmin
• Zhen Mingmin
• Wang Ronggang
• Li Ge
• Dong Shengfu
• Wang Zhenyu
• Li Ying
• Gao Wen

Classifications

• G06V20/64 Primary

  Three-dimensional [3D] objects · CPC title

• G06V10/462

  Salient features, e.g. scale invariant feature transforms [SIFT] · CPC title

• G06K9/00201

  Physics · mapped topic

• G06K9/4671

  Physics · mapped topic

• G06K9/4604 Primary

  Physics · mapped topic