Granted Patent B2

US 12,573,217 · App. 18/250,082 · Granted Mar 10, 2026

Inventors: Yeo Jin Yun (Seoul, KR); Gyu Cheol Lee (Seoul, KR); Jong Kyong Park (Seoul, KR); Yu Lee (Seoul, KR); Zu Cheul Lee (Yongin-si, KR); Chul Hee Lee (Seoul, KR)

Assignee: KT CORPORATION

G06V20/64G06T7/13G06T7/60G06T17/20G06V10/24

Abstract

Claims (56)

1 . A server that creates a spatial model, comprising:

an input unit configured to input a panoramic image of an indoor space to a boundary point detection unit that derives, from the panoramic image of the indoor space, boundary point information that includes a plurality of boundary points between a plurality of planes for the indoor space;

a point cloud creation unit configured to create a point cloud for the panoramic image;

a division unit, implemented by hardware or a combination of hardware and software, the division unit being configured to:

set at least one polygonal region corresponding to the indoor space according to the plurality of boundary points based on the boundary point information,

determine whether the created point cloud is located inside the polygonal region, and

divide the point cloud by extracting the point cloud located inside the polygonal region based on the determination result;

a texture image creation unit configured to create a texture image by projecting the divided point cloud onto a plane corresponding to the divided point cloud;

a mesh model creation unit configured to:

extract geometric information from the point cloud based on the boundary point information,

filter the extracted geometric information from the point cloud to minimize a number of vertices, and

create a mesh model based on the extracted geometric information with the minimum number of vertices by:

performing a meshing operation on a wall surface based on geometric information corresponding to the wall surface extracted from the point cloud, and

performing a meshing operation on a ceiling surface and a floor surface based on geometric information corresponding to the ceiling surface and the floor surface extracted from the point cloud; and

a spatial model creation unit configured to create the spatial model for the indoor space based on the texture image and the mesh model.

2 . The server of claim 1 , wherein the point cloud creation unit is further configured to create the point cloud having points equal in number to a resolution of the panoramic image.

3 . The server of claim 1 , wherein the division unit is further configured to:

select a reference boundary point for each plane from among the plurality of boundary points included in the boundary point information, and

divide the point cloud for each plane based on the selected reference boundary point.

4 . The server of claim 3 , wherein the division unit is further configured to divide a point cloud corresponding to a wall surface from the point cloud based on the number of vectors between boundary points from the plurality of boundary points corresponding to a ceiling surface.

5 . The server of claim 4 , wherein the division unit is further configured to divide, from the point cloud, a point cloud corresponding to a ceiling surface and a floor surface determined as an inside of the polygonal region corresponding to a shape of the indoor space.

6 . The server of claim 1 , wherein the spatial model creation unit is further configured to create the spatial model by matching information from the boundary point information corresponding to the texture image with information from the boundary point information corresponding to the mesh model.

7 . The server of claim 1 , further comprising:

a correction unit, implemented by hardware or a combination of hardware and software, the correction unit being configured to:

align, based on reference position information, a plurality of spatial models respectively corresponding to a plurality of panoramic images obtained by photographing the indoor space at a plurality of positions,

calculate a displacement difference of boundary point information matched between the plurality of aligned spatial models, and

correct the plurality of spatial models based on the calculated displacement difference.

8 . The server of claim 7 , wherein, if there is a missing texture image among texture images constituting a reference model in the plurality of spatial models, the correction unit is further configured to obtain the missing texture image from the plurality of spatial models and correct the reference model based on the obtained texture image.

9 . A method for creating a spatial model by a server, the method comprising:

deriving boundary point information including a plurality of boundary points between a plurality of planes for an indoor space from a panoramic image of the indoor space;

creating a point cloud for the panoramic image;

setting at least one polygonal region corresponding to the indoor space according to the plurality of boundary points based on the boundary point information;

determine whether the created point cloud is located inside the polygonal region;

dividing the point cloud by extracting the point cloud located inside the polygonal region based on the determination result;

creating a texture image by projecting the divided point cloud onto a plane corresponding to the divided point cloud;

extracting geometric information from the point cloud based on the boundary point information;

filtering the extracted geometric information from the point cloud to minimize a number of vertices;

creating a mesh model based on the extracted geometric information with the minimum number of vertices by:

performing a meshing operation on a wall surface based on geometric information corresponding to the wall surface extracted from the point cloud, and

performing a meshing operation on a ceiling surface and a floor surface based on geometric information corresponding to the ceiling surface and the floor surface extracted from the point cloud; and

creating the spatial model for the indoor space based on the texture image and the mesh model.

10 . The method of claim 9 , wherein the creating a point cloud includes creating the point cloud having points equal in number to a resolution of the panoramic image.

11 . The method of claim 9 , wherein the dividing the point cloud includes:

selecting a reference boundary point for each plane from among the plurality of boundary points included in the boundary point information; and

dividing the point cloud for each plane based on the reference boundary point.

12 . The method of claim 11 , wherein the dividing the point cloud includes:

dividing a point cloud corresponding to a wall surface from the point cloud based on the number of vectors between boundary points from the plurality of boundary points corresponding to a ceiling surface.

13 . The method of claim 11 , wherein the dividing the point cloud includes:

dividing, from the point cloud, a point cloud corresponding to a ceiling surface and a floor surface determined as an inside of the polygonal region corresponding to a shape of the indoor space.

14 . The method of claim 9 , wherein the creating the spatial model includes:

creating the spatial model by matching information from the boundary point information corresponding to the texture image with information from the boundary point information corresponding to the mesh model.

15 . The method of claim 9 , further comprising:

aligning, based on reference position information, a plurality of spatial models respectively corresponding to a plurality of panoramic images obtained by photographing the indoor space at a plurality of positions;

calculating a displacement difference of boundary point information matched between the plurality of aligned spatial models; and

correcting the plurality of spatial models based on the calculated displacement difference.

16 . The method of claim 9 , wherein the method is performed by a computing device executing a sequence of instructions for creating a spatial model.

Assignments (2)

Priority Claims (1)

Continuity (1)

References Cited (34)