Method for virtual pre-assembly matching of prefabricated beams based on design-measured point cloud models

Patent metadata

Field	Value
Publication number	US-12406382-B2
Application number	US-202218256266-A
Country	US
Kind code	B2
Filing date	Dec 28, 2022
Priority date	May 24, 2022
Publication date	Sep 2, 2025
Grant date	Sep 2, 2025

Abstract

The present invention relates to the field of virtual pre-assembly matching of bridge engineering components based on 3D point clouds, and particularly relates to a method for virtual pre-assembly matching of prefabricated beams based on design-measured point cloud models. An oriented bounding box is computed respectively for 3D point clouds of two prefabricated beams with assembly relationship therebetween, and two point cloud slices and design point cloud are formed; the two point cloud slices are respectively registered with the generated design point cloud by the iterative closest point algorithm; boundary features and corner features of a pre-assembly interface of two components to be assembled are fitted and extracted; and coarse matching and fine matching of the assembly interface are achieved by the Procrustes analysis algorithm and the iterative closest point algorithm in sequence, a matching degree error of the interface is computed, and an assembly result is evaluated.

First claim

What is claimed is:

1. A method for virtual pre-assembly matching of prefabricated beams based on design-measured point cloud models, comprising the following steps:
step (1): computing an oriented bounding box respectively for 3D point clouds of two prefabricated beams with assembly relationship therebetween, and implementing 3D coordinate calibration based on geometric features of the 3D point clouds of the two prefabricated beams; for the calibrated 3D point clouds of the two prefabricated beams, making two point cloud slices at an assembly interface of the two prefabricated beams respectively; and based on an assembly interface jointly contained in the two point cloud slices, generating a discrete design point cloud for design information contained in the assembly interface;
step (2): registering the point cloud slices at the assembly interface of the two prefabricated beams in step (1) with the generated design point cloud respectively by an iterative closest point algorithm; and setting a distance threshold on the basis of a coordinate range of the design point cloud to denoise the two point cloud slices at the assembly interface;
step (3): based on the coordinate range of the design point cloud in step (2), partitioning the two denoised point cloud slices; and selecting a fitting function and a fitting algorithm according to features required to be fitted, to fit and extract boundary features and corner features of an assembly interface of two components to be assembled; and
step (4): based on the fitted boundary features and corner features of the pre-assembly interface of the two components to be assembled in step (3), implementing coarse matching of the assembly interface by a Procrustes analysis algorithm firstly, and then implementing fine matching of the assembly interface by the iterative closest point algorithm, adjusting the 3D point cloud to be in a final assembly posture, computing a matching degree error of the assembly interface, and evaluating an assembly result.

2. The method for virtual pre-assembly matching of prefabricated beams based on design-measured point cloud models according to claim 1 , wherein step (1) comprises the following specific sub-steps:
step 1.1: computing an oriented bounding box respectively for 3D point clouds of two prefabricated beams with assembly relationship therebetween, such that X-axis, Y-axis and z-axis directions of a 3D coordinate system are respectively parallel to a beam width, beam length and beam height directions of a component, thereby achieving coordinate calibration;
step 1.2: for features of an assembly interface of the two prefabricated beams, making point cloud slices (i.e., a sliced point cloud P 1 and a sliced point cloud Q 1 ) parallel to a XOZ coordinate plane respectively at an assembly interface of the two 3D point clouds subjected to coordinate calibration; where a thickness of a slice is twice a density of a measured point cloud; and the expression is as follows:
P 1 ={p 1 ,p 2 , . . . ,p m },Q 1 ={q 1 ,q 2 , . . . ,q n }

step 1.3: generating a discrete design point cloud D based on design information and design drawings of an assembly interface jointly contained in the sliced point cloud P 1 and the sliced point cloud Q 1 , where the expression is as follows:
D={d 1 ,d 2 , . . . ,d j }.

3. The method for virtual pre-assembly matching of prefabricated beams based on design-measured point cloud models according to claim 2 , wherein in step (2), by taking the design point cloud D as a reference, the sliced point cloud P 1 and the sliced point cloud Q 1 are respectively registered with the design point cloud D based on the iterative closest point algorithm through a registration method as follows:
step 2.1: moving centroids of the sliced point cloud P 1 , the sliced point cloud Q 1 and the design point cloud D respectively to an origin of coordinates, that is:

P
c

=

P
1

-

P
_

1

,

Q
c

=

Q
1

-

Q
_

1

,

D
c

=

D
-

D
_

where P c is a moved sliced point cloud P 1 , and P 1 is an average value of coordinates of the sliced point cloud P 1 ;
Q c is a moved sliced point cloud Q 1 , and Q 1 is an average value of coordinates of the sliced point cloud Q 1 ; and
D c is a moved design point cloud D, and D is an average value of coordinates of the design point cloud D;
step 2.2: in the moved design point cloud D c , finding points closest to each point in the moved sliced point cloud P c and the moved sliced point cloud Q c respectively, and minimizing a variance of a distance between corresponding points via an orthogonal rotation matrix and a rigid translation matrix, that is:

argmin
⁢

(

f
⁢

(
p
)

)

=

1
m

⁢

∑

i
=
1

m



R
p

·

p
i

+

T
p

-

d
i



2

argmin
⁢

(

Assignees

• Univ Southeast

Inventors

• Xiong Wen
• Xu Chang
• Zhu Yanjie

Classifications

• G06T2207/10028

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

• G06T3/14 Primary

  Transformations for image registration, e.g. adjusting or mapping for alignment of images · CPC title

• G06T5/70

  Denoising; Smoothing · CPC title

• G06F30/17

  Mechanical parametric or variational design · CPC title

• G06T2207/20021

  Dividing image into blocks, subimages or windows · CPC title