Granted Patent B2

US 12,223,579 · App. 17/736,382 · Granted Feb 11, 2025

Inventors: Hongyu Liu (Tai Po, HK); Wing Cheong Daniel Ho (Kowloon, HK)

Assignee: City University of Hong Kong

G06T15/00G06T7/60G06T2207/20081G06T2207/20084G06T2207/30196

Abstract

Claims (25)

1. A method for generating a 3D image of a body shape, implemented by a processor, comprising the steps of:

A. obtaining a set of far-field patterns of electromagnetic waves from a given body shape of a human or animal by collecting electromagnetic far-field patterns radiated from physical means, the physical means comprising a body shape model;

B. establishing a one-to-one correspondence between the body shape and the set of far-field patterns of electromagnetic waves;

C. representing the far-field patterns as a shape generating vector;

D. mapping a vector of characteristic parameters to the shape generating vector;

E. producing a body generation model by taking the vector of characteristic parameters and the shape generating vector as input and output data to form a training dataset for machine learning;

F. specifying a characteristic vector of an unknown geometric body shape and obtaining a shape generating vector of the unknown geometric body shape based on the body generation model;

G. obtaining far-field patterns corresponding to the shape generating vector of the unknown geometric body shape; and

H. reconstructing the unknown geometric body shape based on the obtained far-field patterns by an inverse scattering algorithm governed by a Maxwell system,

wherein the body shape comprises a stationary or moving body shape,

wherein, if the body shape is the moving body shape, the Maxwell system is taken to be time-dependent with a moving scatterer, and

wherein the set of far-field patterns of electromagnetic waves comprises electromagnetic spectral data which encodes a color of a scattering object.

2. The method of claim 1 , wherein step B comprises establishing the one-to-one correspondence between the body shape and the set of far-field patterns of electromagnetic waves by treating the body shape as a scattering object.

3. The method of claim 1 , wherein step B comprises representing the far-field patterns through the use of vectorial spherical harmonics.

4. The method of claim 3 , wherein step B is based on Fourier theory, and the shape generating vector consists of corresponding Fourier coefficients of the far field patterns.

5. The method of claim 1 , wherein step B comprises establishing the one-to-one correspondence between the body shape and the set of far-field patterns of electromagnetic waves by treating the body shape as a perfect conductor.

6. The method of claim 1 , wherein step E further comprises training the body generation model using the training dataset by an artificial neural network.

7. The method of claim 6 , wherein the artificial neural network comprises a convolution neural network (CNN).

8. The method of claim 1 , wherein if the body shape is the stationary body shape, the Maxwell system is taken to be time-harmonic with frequencies to be a visible electromagnetic spectrum.

9. The method of claim 1 , wherein the characteristic parameters comprise gender, height, weight, arm span, waist girth, neck girth, abdomen girth, body mass index (BMI), muscle degree, and/or motion trajectory.

10. A processor configured to implement the method of claim 1 .

11. A system for generating a 3D image of a body shape, comprising:

a processor configured to implement the method of claim 1 ;

an input unit configured to input characteristic parameters of a body shape; and

a display unit configured to present the 3D image of the body shape reconstructed by the processor.

Assignments (1)

Continuity (1)

References Cited (18)