Method, apparatus, computer device and storage medium for three-dimensional reconstruction of indoor structure

Patent metadata

Field	Value
Publication number	US-12530849-B2
Application number	US-202318453238-A
Country	US
Kind code	B2
Filing date	Aug 21, 2023
Priority date	Sep 13, 2022
Publication date	Jan 20, 2026
Grant date	Jan 20, 2026

Abstract

The disclosure provides a method, an apparatus and a computer device for three-dimensional reconstruction of an indoor structure based on a two-dimensional video, wherein the method comprises: generating a color point cloud of three-dimensional structure of a house by using the video data of the house; regularizing the color point cloud of the house; extracting a plan of the house based on the color point cloud of the house; and constructing a three-dimensional model of the house in real life. The disclosure realizes autonomous and rapid construction of a three-dimensional model of a house by the general public, and realizes a low-cost, efficient and convenient three-dimensional visual digital representation of the house.

First claim

What is claimed is:

1 . A method for three-dimensional reconstruction of an indoor structure based on a two-dimensional video, characterized in that the method comprises:
S1: generating a color point cloud of three-dimensional structure of a house using a captured panoramic video data of the house; S2: regularizing the color point cloud of the three-dimensional structure of the house to obtain a regularized color point cloud; S3: vertically projecting the regularized color point cloud of the three-dimensional structure of the house onto a horizontal plane, fitting a straight line according to a density of projected points, performing region division based on the fitted straight line and a spatial division algorithm to get a vector polygon units that divides space, and splicing the vector polygon units to get vector room plan data, wherein the vector room plan data is a vector line map layer obtained by projecting the regularized color point cloud of the three-dimensional structure of the house onto a two-dimensional plane; S4: performing reconstruction of a three-dimensional model of the house based on the vector room plan data, wherein step S1 comprises following steps: S1.1: establishing a sequence image set on the captured panoramic video data of the house by means of a video stream frame extraction technique, wherein the sequence image set is a sequence of images with temporal correlation; S1.2: performing image multi-view projection correction on the sequence image set to obtain a sequence multi-view image; S1.3: based on the obtained sequence multi-view image, generating a three-dimensional sparse point cloud by sequence multi-view image dense matching and structure from motion reconstruction techniques; S1.4: generating a dense three-dimensional point cloud based on the three-dimensional sparse point cloud using a patch-based multi-view stereo method as the color point cloud of the three-dimensional structure of the house, wherein step S1.2 comprises: converting an original panoramic image into a three-dimensional panoramic spherical space by using an equirectangular projected spherical panoramic model;

{

x
pano
q

=

w
·

θ
q

/
2
⁢
π

+

(

w
/
2

-
1

)

y
pano
q

=

(

h
/
2

-
1

)

-

h
·

ϕ
q

/
π

,

(

-
π

<
θ
≤
π

,

-
π

/
2

<
ϕ
≤

π
/
2

)

(
1
)

{

θ
q

=

(

2
⁢

π
·

x
pano
q

+

2

Assignees

• Univ Wuhan

Inventors

• Chen Jiangping
• Wang Siyuan
• Zhao Pengcheng
• Zhang Bin
• Wang Bohua

Classifications

• G06T2219/2016

  Rotation, translation, scaling · CPC title

• G06T2219/2012

  Colour editing, changing, or manipulating; Use of colour codes · CPC title

• G06T2210/56

  Particle system, point based geometry or rendering · CPC title

• G06T2210/04

  Architectural design, interior design · CPC title

• G06T2207/30244

  Camera pose · CPC title