Compressed sensing based object imaging system and imaging method therefor

Patent metadata

Field	Value
Publication number	US-11368608-B2
Application number	US-201816618376-A
Country	US
Kind code	B2
Filing date	Jan 18, 2018
Priority date	Jun 1, 2017
Publication date	Jun 21, 2022
Grant date	Jun 21, 2022

Abstract

A compressed sensing based object imaging system and an imaging method thereof. The object imaging system comprises a light source generation unit (11), a filter unit (12), an image generation unit (13), an image acquisition unit (14), and an image reconstruction unit (15). The light source generation unit (11) generates experimental laser; the filter unit (12) filters high frequency scattered light and forms parallel light; the image generation unit (13) generates an experimental image in which an object image (16) and a specific measurement matrix (17) are superimposed; the image acquisition unit (14) performs compression sampling on the generated experimental image; and the image reconstruction unit (15) reconstructs sampling data to restore the object image (16). The imaging method comprises: establishing a sample database comprising the specific target object image (16); training sample images to obtain the specific measurement matrix (17); and simultaneously completing image sampling, image compression and image recognition in an all-optical system. The system and the method can greatly reduce the data volume recorded in image recognition and image matching, thus improving the real-time performance of the system, and providing a possibility of concurrent processing by machine vision and artificial intelligence.

First claim

The invention claimed is:

1. A compressed sensing based object imaging system, comprising a light source generation unit, a filter unit, an image generation unit, an image acquisition unit, and an image reconstruction unit that are sequentially connected,
wherein the light source generation unit generates experimental laser light;
the filter unit controls the attenuation of the light intensity of the laser light, filters off high frequency scattered light, and forms a zero-order diffraction spot, and then controls the diffraction spot to form parallel light;
the image generation unit generates an experimental image in which an object image and a specific measurement matrix are superimposed;
the image acquisition unit performs compressed sampling on the experimental image; and
the image reconstruction unit reconstructs sampling data to restore the object image.

2. The compressed sensing based object imaging system according to claim 1 , wherein the light source generation unit comprises a laser and a mirror that can reflect laser light to change a propagation direction thereof.

3. The compressed sensing based object imaging system according to claim 1 , wherein the filter unit comprises a circular tunable attenuator, a pinhole filter, and a Fourier lens that are sequentially arranged in parallel with an optical path.

4. The compressed sensing based object imaging system according to claim 1 , wherein the image generation unit comprises a first polarizing film, a spatial light modulator loaded with the specific measurement matrix, and a second polarizing film that are sequentially arranged in parallel with an optical path.

5. The compressed sensing based object imaging system according to claim 1 , wherein the image acquisition unit comprises a convergent lens and a single photon detector that are sequentially arranged in parallel with an optical path.

6. The compressed sensing based object imaging system according to claim 1 , wherein the image reconstruction unit comprises a computer.

7. An imaging method based on the system of claim 1 , comprising the following steps:
step S1: emitting experimental laser light by using a laser light source;
step S2: performing attenuation processing on the experimental laser light, filtering off high frequency scattered light to obtain a zero-order diffraction spot, and adjusting the diffraction spot to be parallel light;
step S3: illuminating an object image with the parallel light, and generating, through a spatial light modulator loaded with a specific measurement matrix, an experimental image in which the object image and the specific measurement matrix are superimposed; and
step S4: performing compressed sampling on the experimental image, such that an image reconstruction unit reconstructs the object image according to sampling data.

8. The imaging method according to claim 7 , wherein a calculation method for the specific measurement matrix in step S3 is: establishing a sample library comprising the specific target object image, and training sample images by using a principal component analysis method to obtain the specific measurement matrix.

9. The imaging method according to claim 8 , wherein according to the principal component analysis method, provided that there is a set of N image training samples X each having a size of p×q pixels, each sample has a vector X i that is formed by a pixel grayscale value thereof, and the set of training samples constituted by vectors is X={X 1 , X 2 , . . . , X i , X N }, the mean of the set of training samples is first calculated:

X
_

=

1
N

⁢

∑

i
=
1

N

⁢

⁢

X
i

,

data is further centralized:
{circumflex over (X)} i =X i − X ,

a covariance matrix is calculated for the centralized data:

R
=

1
N

⁢

∑

i
=
1

N

⁢

⁢

(

X
i

-

X
_

)

⁢

(

X
i

-

X
_

)

T

eigenvalues λ i of the covariance matrix R are arranged in a descending order, and eigenvectors u i corresponding to the first m eigenvalues are formed into a principal component matrix U,
U=[u 1 , u 2 , . . . , u m ]

wherein the m components extract features that represent main information of the images; and
the principal component matrix is subjected to rank inversion to form the specific measurement matrix, and according to the principle of compressed sensing, the specific measurement matrix after training with the sample library is expressed as:
Φ= U T =[ϕ 1 , ϕ 2 , . . . , ϕ m ] T (ϕ i =u i ).

10. The imaging method according to claim 7 , wherein acquiring experimental measurement data in step S4 comprises: acquiring an optical signal at a single point and converting the optical signal into an electrical signal, with an output voltage value being expressed as:
y n =φ n x,

where x∈R p×q , x representing the object image, φ n ∈R p×q , n∈{1, 2, . . . , m}, φ n representing an n th dimension of the specific measurement matrix, which implements an n th measurement of the object image x; and
repeating this process m times, such that a measured value Y can be obtained:
Y=[y 1 , y 2 , . . . , y m ] T =[φ 1 , φ 2 , . . . φ m ] T x=Φx,

where Φ∈R m×(p×q) is the specific measurement matrix, and Y∈R m×1 is the measured value.

11. The imaging method according to claim 7 , wherein in step S4, the image reconstruction unit comprises a data screening processing mo

Assignees

• Univ South China Normal

Inventors

• Li Jun
• Lei Miao
• Dai Xiaofang
• Wang Shangyuan
• Zhong Ting
• Liang Chuangxue
• Wang Chen
• Xie Ping
• Wang Ruiqiang

Classifications

• H04N23/55 Primary

  Optical parts specially adapted for electronic image sensors; Mounting thereof · CPC title

• H04N23/56

  provided with illuminating means · CPC title

• G06T7/0002

  Inspection of images, e.g. flaw detection · CPC title

• G01V8/10 Primary

  Detecting, e.g. by using light barriers (by reflection from the object G01S17/00) · CPC title

• G02B5/3033 Primary

  in the form of a thin sheet or foil, e.g. Polaroid · CPC title