Tomographic Imaging Using Fast Stochastic Hierarchical Bayesian Systems and Methods

• US 20200134888A1
• Filed: 10/29/2019
• Published: 04/30/2020
• Est. Priority Date: 10/29/2018
• Status: Active Grant

• Current Assignee: The Government Of The United States Of America, As Represented By The Secretary Of The Navy
• Original Assignee: The Government Of The United States Of America, As Represented By The Secretary Of The Navy
• Inventors: Raj, Raghu G., Mckay, John, Monga, Vishal
• Granted Patent: US 11,120,583 B2
• Time in Patent Office: Days
• CPC Class Codes: G06T 11/006 Inverse problem, transformation from projection-space into object-space, e.g. transform methods, back-projection, algebraic methods
• CPC Class Codes: G06T 2211/424 Iterative

Abstract

Systems and methods are provided for imaging that demonstrably outperform previous approaches (especially compressive sensing based approaches). Embodiments of the present disclosure provide and solve an imaging cost function via a stochastic approximation approach. By doing so, embodiments of the preset disclosure provide a significant means of generalization and flexibility to adapt to different application domains while being competitive in terms of computational complexity.

Claims

1. An imaging system, comprising:

a sensor configured to sense a scene from a plurality of different angles, thereby generating sensed data; and

an image generator device in communication with the sensor, wherein the image generator device is configured to;

generate a plurality of sensing matrices based on geometry of the angles and a waveform used to sense the scene,

estimate a plurality of coefficients for an image based on the sensing matrices,

estimate a Compound Gaussian (CG) vector based on the plurality of coefficients using a Simultaneous Perturbation Stochastic Approximation (SPSA), and

generate an image based on the estimated CG vector.

2. The imaging system of claim 1, wherein the image generator device is further configured to:

estimate a non-Gaussian component of the CG vector based on the plurality of coefficients using the SPSA.

3. The imaging system of claim 2, wherein the image generator device is further configured to:

multiply columns of an effective sensing matrix with scalar components of the non-Gaussian component of the CG vector.

4. The imaging system of claim 3, wherein the image generator device is further configured to:

estimate a Gaussian component of the CG vector based on the effective sensing matrix, the non-Gaussian component, and the sensed data.

5. The imaging system of claim 4, wherein the image generator device is further configured to:

generate the image based on the estimated CG vector and a dictionary matrix.

6. The imaging system of claim 1, wherein the sensor is a synthetic-aperture radar (SAR).

7. The imaging system of claim 1, wherein the image generator device is configured to generate the plurality of sensing matrices using a Radon transform.

8. A method for generating an image, the method comprising:

generating, using an image generator device, a plurality of sensing matrices based on geometry of a plurality of angles used to sense a scene and a waveform used to sense the scene;

estimating, using the image generator device and based on the sensing matrices, a plurality of coefficients for an image;

estimating, using the image generator device, a Compound Gaussian (CG) vector based on the plurality of coefficients using a Simultaneous Perturbation Stochastic Approximation (SPSA); and

generating, using the image generator device, the image based on the estimated CG vector.

9. The method of claim 8, further comprising:

sensing a scene from a plurality of different angles, thereby generating sensed data.

10. The method of claim 9, further comprising:

estimating a non-Gaussian component of the CG vector based on the plurality of coefficients using the SPSA.

11. The method of claim 10, further comprising:

multiplying columns of an effective sensing matrix with scalar components of the non-Gaussian component of the CG vector.

12. The method of claim 11, further comprising:

estimating a Gaussian component of the CG vector based on the effective sensing matrix, the non-Gaussian component, and the sensed data.

13. The method of claim 12, wherein generating the image comprises:

generating the image based on the estimated CG vector and a dictionary matrix.

14. The method of claim 8, wherein the sensor is a synthetic-aperture radar (SAR).

15. The method of claim 8, wherein generating the plurality of sensing matrices comprises generating the plurality of sensing matrices using a Radon transform.

16. An imaging device, comprising:

a processor;

a memory; and

an image generator device configured to

generate a plurality of sensing matrices based on geometry of a plurality of angles used to sense a scene and a waveform used to sense the scene,

estimate, based on the sensing matrices, a plurality of coefficients for an image,

estimate a Compound Gaussian (CG) vector based on the plurality of coefficients using a Simultaneous Perturbation Stochastic Approximation (SPSA), and

generate an image based on the estimated CG vector.

17. The imaging device of claim 16, wherein the image generator device is further configured to:

estimate a non-Gaussian component of the CG vector based on the plurality of coefficients using the SPSA;

multiply columns of an effective sensing matrix with scalar components of the non-Gaussian component of the CG vector;

estimate a Gaussian component of the CG vector based on the effective sensing matrix, the non-Gaussian component, and sensed data; and

generate the image based on the estimated CG vector and a dictionary matrix.

18. The imaging device of claim 17, further comprising:

a sensor configured to sense a scene from a plurality of different angles, thereby generating the sensed data.

19. The imaging device of claim 18, wherein the sensor is a synthetic-aperture radar (SAR).

20. The imaging device of claim 18, wherein the image generator device is configured to generate the plurality of sensing matrices using a Radon transform.