Reconstruction method, reconstruction device, and generation method for generating a three-dimensional (3D) model

1. A reconstruction method for reconstructing a three-dimensional (3D) model using a plurality of two-dimensional (2D) images obtained by a plurality of cameras disposed in different positions and attitudes capturing a same target space in a 3D space, the method comprising:

converting the plurality of 2D images into a plurality of first low-resolution images, respectively, by lowering a resolution of each of the plurality of 2D images;

converting a plurality of camera parameters of the plurality of cameras into a plurality of converted camera parameters, respectively, along with the conversion of the plurality of 2D images into the plurality of first low-resolution images;

reconstructing a first 3D point cloud using the plurality of first low-resolution images and the plurality of converted camera parameters, the first 3D point cloud including a plurality of 3D points;

generating a new 3D point by feature point matching using a plurality of high-resolution images obtained based on the plurality of 2D images and each having a resolution that is higher than a resolution of any of the plurality of first low-resolution images;

generating a second 3D point cloud by adding the new 3D point generated to the first 3D point cloud; and

reconstructing the 3D model using the second 3D point cloud.

2. The reconstruction method according to claim 1 , further comprising:

determining whether or not a predetermined condition is satisfied;

when determining that the predetermined condition is not satisfied, repeating the following until the predetermined condition is satisfied:

a process of converting the plurality of 2D images into a plurality of second low-resolution images having a resolution different from the resolution of the plurality of first low-resolution images obtained in a last round of conversion; and

a process of reconstructing a first 3D point cloud including a plurality of 3D points using the plurality of second low-resolution images; and

in the generating of the second 3D point cloud, the second 3D point cloud including 3D points increased by adding the new 3D point to the first 3D point cloud generated as a result of the repeating until the predetermined condition is satisfied is generated.

3. The reconstruction method according to claim 2 , wherein

the plurality of second low-resolution images obtained in the conversion that is repeated has a resolution higher than the resolution of the plurality of first low-resolution images.

4. The reconstruction method according to claim 2 , wherein

the predetermined condition is that a total number of 3D points included in an obtained 3D point cloud exceeds a predetermined number.

5. The reconstruction method according to claim 1 , wherein

in the generating of the new 3D point, the new 3D point is generated by feature point matching using an empty pixel which is a pixel that does not correspond to the plurality of 3D points included in the first 3D point cloud, among pixels of the plurality of high-resolution images.

6. A reconstruction device, comprising:

a processor which reconstructs a three-dimensional (3D) model using a plurality of two-dimensional (2D) images obtained by a plurality of cameras disposed in different positions and attitudes capturing a same target space in a 3D space, wherein

the processor performs operations including:

converting the plurality of 2D images into a plurality of first low-resolution images, respectively, by lowering a resolution of each of the plurality of 2D images;

converting a plurality of camera parameters of the plurality of cameras into a plurality of converted camera parameters, respectively, along with the conversion of the plurality of 2D images into the plurality of first low-resolution images;

reconstructing a first 3D point cloud using the plurality of first low-resolution images and the plurality of converted camera parameters, the first 3D point cloud including a plurality of 3D points;

generating a new 3D point by feature point matching using a plurality of high-resolution images obtained based on the plurality of 2D images and each having a resolution that is higher than a resolution of any of the plurality of first low-resolution images;

generating a second 3D point cloud by adding the new 3D point generated to the first 3D point cloud; and

reconstructing the 3D model using the second 3D point cloud.

7. The reconstruction device according to claim 6 , wherein

in the generating of the new 3D point, the new 3D point is generated by feature point matching using an empty pixel which is a pixel that does not correspond to the plurality of 3D points included in the first 3D point cloud, among pixels of the plurality of high-resolution images.

8. A generation method, comprising:

obtaining two-dimensional (2D) images generated by photographing a space from different viewpoints with at least one camera;

reducing resolutions of each of the 2D images to generate first images, respectively;

generating second images from each of the 2D images, respectively such that a resolution of each of the second images is higher than a resolution of any one of the first images;

generating first three-dimensional (3D) points based on the first images and a plurality of camera parameters from the at least one camera, the first 3D points indicating respective first positions in the space;

generating a new 3D point by feature point matching based on the second images, the new 3D point indicating a second position in the space;

generating second 3D points by adding the new 3D point generated to the first 3D points; and

generating a 3D model of the space based on the second 3D points.

9. The generation method according to claim 8 , wherein

in the generating of the new 3D point, the new 3D point is generated by feature point matching using an empty pixel which is a pixel that does not correspond to the first 3D points, among pixels of the second images.