3D object imaging method and 3D object imaging system

1 . A 3D object imaging method for a 3D object imaging system, wherein the system includes a server and a user terminal, the method comprising:

executing, by the server, a Neural Radiance Fields (NeRF) model to obtain a 3D object generated by the NeRF model in a virtual space;

generating, by the server, a 3D wrapping mesh, which completely surrounds the 3D object, in the virtual space, wherein the 3D wrapping mesh comprises a plurality of wrapping mesh coordinates;

performing, by the server, a color sampling operation on the 3D object according to the 3D wrapping mesh and a 2D texture mesh mapped to the 3D wrapping mesh to sample a color on the surface of the 3D object, so as to generate a color record file corresponding to the 3D object, wherein performing the color sampling operation comprises: for each of the plurality of wrapping mesh coordinates, generating a plurality of sampling rays across the respective wrapping mesh coordinate with different spherical coordinate angles to obtain a plurality of color values, and encoding the plurality of color values to generate a color vector for the respective wrapping mesh coordinate;

downloading, by the user terminal, the color record file corresponding to the 3D object, wherein the color record file comprises the 2D texture mesh and a plurality of color vectors respectively stored in a plurality of texture pixels of the 2D texture mesh;

determining, by the user terminal, a plurality of display rays respectively corresponding to the plurality of texture pixels of the 2D texture mesh according to a specified viewing angle corresponding to an applied input operation, wherein the plurality of display rays have different spherical coordinate angles;

obtaining, by the user terminal, according to the plurality of display rays, a color value of each of a plurality of display pixels of a 2D display image corresponding to the 3D object at the specified viewing angle from the 2D texture mesh of the color record file; and

displaying, by the user terminal, the plurality of display pixels on a display of the user terminal according to the color value of each of the plurality of display pixels, so as to display the 2D display image, wherein an appearance of the 3D object at the specified viewing angle is seen in real time by the 2D display image.

2 . The 3D object imaging method of claim 1 , wherein the step of performing the color sampling operation on the 3D object according to the 3D wrapping mesh and the 2D texture mesh mapped to the 3D wrapping mesh to sample the color on the surface of the 3D object, so as to generate the color record file corresponding to the 3D object comprises:

generating, by the server, the 2D texture mesh according to the 3D wrapping mesh, where each texture pixel of the 2D texture mesh is mapped to each wrapping mesh coordinate point of the 3D wrapping mesh;

performing, by the server, for each wrapping mesh coordinate, following steps:

generating, according to the wrapping mesh coordinate, a plurality of sampling rays across the wrapping mesh coordinate with different spherical coordinate angles to obtain, by the plurality of sampling rays and the NeRF model, a plurality of color values respectively corresponding to the plurality of sampling rays;

encoding, by the server, by a spherical coordinate distribution function algorithm, the plurality of color values corresponding to the plurality of sampling rays to obtain a color vector corresponding to the wrapping mesh coordinate; and

storing, by the server, the color vector into a texture pixel of the 2D texture mesh mapped to the wrapping mesh coordinate; and

after storing the color vector corresponding to all of the wrapping mesh coordinates of the 3D wrapping mesh into all of the texture pixels of the 2D texture mesh, generating, by the server, a color record file corresponding to the 3D object according to the 3D wrapping mesh and the 2D texture mesh.

3 . The 3D object imaging method of claim 2 , wherein the step of obtaining, according to the display rays, the color value of each of the display pixels of the 2D display image corresponding to the 3D object at the specified viewing angle from the 2D texture mesh of the color record file comprises:

determining, by the user terminal, according to the plurality of display rays and the 2D texture mesh, a first 2D texture mesh of a part of the 2D texture mesh, wherein the first 2D texture mesh corresponds to the 2D display image; and

calculating, by the user terminal, according to a first color vector of each of a plurality of first texture pixels of the first 2D mapping mesh and the display rays, a color value of each of the first texture pixels at the specified viewing angle as a color value of each of the display pixels of the 2D display image by the spherical coordinate distribution function algorithm.

4 . The 3D object imaging method of claim 3 , wherein the spherical coordinate angles of the plurality of display rays comprise an azimuth angle and a polar angle.

5 . The 3D object imaging method of claim 4 , wherein the step of generating, according to the wrapping mesh coordinate, the plurality of sampling rays across the wrapping mesh coordinate with different spherical coordinate angles comprises:

for each sampling ray among the plurality of sampling rays,

generating a line segment of length 2r centered on the wrapping mesh coordinate of the sampling ray according to the corresponding spherical coordinate angle, wherein r is a sampling range constant value of the NeRF model.

6 . The 3D object imaging method of claim 5 , wherein the step of to obtaining, by the plurality of sampling rays and the NeRF model, the plurality of color values respectively corresponding to the plurality of sampling rays comprises:

for each sampling ray among the plurality of sampling rays,

determining i sampling points on the sampling ray, wherein i is a positive integer;

obtaining, by the NeRF model, an inferred NeRF color and a density value of each of the i sampling points; and

integrating the NeRF color values and the density values of the i sampling points to obtain the color value corresponding to the sampling ray.

7 . The 3D object imaging method as in claim 6 , wherein the step of encoding, by the spherical coordinate distribution function algorithm, the plurality of color values corresponding to the plurality of sampling rays to obtain the color vector corresponding to the wrapping mesh coordinate comprises:

calculating N spherical harmonic function bases of each sampling ray according to the spherical coordinate angle of the sampling ray, wherein N is a predetermined base number corresponding to the spherical coordinate distribution function algorithm, where N is 1, 4, 9, 16, or 25;

calculating a three-primary-color contribution vector of each sampling ray according to a three-primary-color vector value of the color value of the sampling ray and the calculated N spherical harmonic function bases; and

calculating an average of the three-primary-color contribution vector of each sampling ray corresponding to the wrapping mesh coordinate as the color vector corresponding to the wrapping mesh coordinate.

8 . The 3D object imaging method of claim 7 , wherein the step of calculating, the color value of each of the first texture pixels at the specified viewing angle by the spherical coordinate distribution function algorithm comprises:

for a target first texture pixel among the plurality of first texture pixels,

calculating, according to a target spherical coordinate angle of a target display ray across the target first texture pixel, a corresponding N target spherical harmonic function bases; and

performing an inner product operation on the N target spherical harmonic function bases and a target first color vector of the target first texture pixel to calculate the color value of the target first texture pixel at the specified viewing angle.

9 . The 3D object imaging method of claim 1 , wherein the 3D object reflects an actual appearance of a real object, and the method further comprises:

photographing the real object from a plurality of viewpoints to capture a plurality of object images; and

inputting the plurality of object images into the NeRF model and the NeRF model performs an inference operation to generate the 3D object reflecting the real object, wherein the 3D object is rotated to simulate the actual appearance of the real object in different viewing angles.

10 . A 3D object imaging system, comprising:

a server; and

a user terminal,

wherein the server executes a Neural Radiance Fields (NeRF) model to obtain a 3D object generated by the NeRF model in a virtual space,

wherein the server generates a 3D wrapping mesh, which completely surrounds the 3D object, in the virtual space, wherein the 3D wrapping mesh comprises a plurality of wrapping mesh coordinates,

wherein the server performs a color sampling operation on the 3D object according to the 3D wrapping mesh and a 2D texture mesh mapped to the 3D wrapping mesh to sample a color on the surface of the 3D object, so as to generate a color record file corresponding to the 3D object, wherein performing the color sampling operation comprises: for each of the plurality of wrapping mesh coordinates, generating a plurality of sampling rays across the respective wrapping mesh coordinate with different spherical coordinate angles to obtain a plurality of color values, and encoding the plurality of color values to generate a color vector for the respective wrapping mesh coordinate,

wherein the user terminal downloads the color record file corresponding to the 3D object, wherein the color record file comprises the 2D texture mesh and a plurality of color vectors respectively stored in a plurality of texture pixels of the 2D texture mesh,

wherein the user terminal determines a plurality of display rays respectively corresponding to the plurality of texture pixels of the 2D texture mesh according to a specified viewing angle, wherein the plurality of display rays have different spherical coordinate angles,

wherein the user terminal obtains, according to the plurality of display rays, a color value of each of a plurality of display pixels of a 2D display image corresponding to the 3D object at the specified viewing angle from the 2D texture mesh of the color record file,

wherein the user terminal displays the plurality of display pixels on a display of the user terminal according to the color value of each of the plurality of display pixels, so as to display the 2D display image, wherein an appearance of the 3D object at the specified viewing angle is seen in real time by the 2D display image.