CN104021522A - Target image separating device and method based on intensity correlated imaging - Google Patents

Abstract

A target image separation device and separation method based on intensity-correlated imaging. The separation device includes: a laser filter system, a pseudothermo-optic system, a reference arm system, an object arm system, and an algorithm module. The laser, the object optical path barrel detector and the reference optical path detector are triggered and controlled by a synchronous pulse signal sent by the computer to work at the same time. After multiple measurements, the computer collects the results of multiple measurements as the detection matrix and signals, and through compressed sensing The solution algorithm combined with the sparse joint transformation can restore the coefficients of the target in these transformations, and then use these coefficients to obtain different morphological features of the target through the separation algorithm. Different parts of the target image in the field of view are separated and obtained by means of the device of the invention.

Description

Target Image Separation Device and Separation Method Based on Intensity Correlation Imaging

technical field

The invention belongs to the field of optical imaging, in particular to a target image separation device and separation method based on intensity correlation imaging.

Background technique

Correlation imaging, also known as ghost imaging, has attracted more and more researchers' attention in recent years, and belongs to the frontier and hot field of quantum optics. Correlation imaging calculates the image of the object by correlating the light field intensity distribution detected by the array CCD on the reference arm with the light intensity obtained by the barrel detector on the object arm. Correlative imaging can generate an image of an object on the optical path containing the object without requiring a spatially resolved area array detector. Currently, as a new type of imaging technology, it has attracted widespread attention. However, correlative imaging requires an infinite number of samples in theory to obtain a perfect reconstructed image, and it cannot break through the diffraction limit of its aperture. Therefore, scientists have made a lot of efforts to overcome these problems, such as combining correlative imaging with Combined with compressed sensing, some algorithms in compressed sensing are used to restore the image, so that a better image can be obtained under the same sampling times, that is, the sampling efficiency of the image is improved. Some researchers use effective means such as differential detection to improve sampling efficiency. Combining correlative imaging with image sparsity can also achieve super-resolution. Moreover, ghost imaging can not only form a real image of the target, but also obtain the diffraction image of the object. Therefore, the development and application prospects of correlation imaging technology are becoming more and more broad.

Target image separation plays a very important role in the later application of correlation imaging. The use of image separation can achieve target recognition, target feature enhancement, etc.; the features obtained by target image separation are still used in many other fields, such as Correlation imaging can also be used in remote sensing. In this way, image separation can be used to obtain the required image features to reduce the space required for storage and the amount of information transmitted. In this device, different features in the image can be sparsely represented under different transformations to realize the separation of target image features. Therefore, applying this image separation technique to an associated imaging device can promote the development of the device in the application field.

Contents of the invention

The purpose of the present invention is to provide a target image separation device and separation method based on intensity correlation imaging, which uses correlation imaging and target image separation technology to realize the separation of specific features of the target in the field of view, so as to realize the identification of targets in correlation imaging, The enhancement of image features and the reduction of the space required to store useful information, etc., promote the development of associative imaging applications.

In order to achieve the above object, the technical solution of the present invention is as follows:

A target image separation device based on intensity-correlated imaging, characterized in that the device includes a laser, and the direction of the laser light emitted by the laser is a focusing lens, a first aperture, a collimating lens, a second aperture, ground glass, and a beam splitter in sequence , an object to be detected, a lens and a barrel detector, an area array detector is arranged in the reflected light direction of the beam splitter, the output terminals of the area array detector and the barrel detector are connected to the input end of the computer, and the computer The output terminal is connected to the input terminal of the laser, the area detector and the barrel detector, the first diaphragm is located at the focal point of the focusing lens, and the barrel detector is located at the focal point of the lens, so The distance z from the area array detector to the center of the frosted glass and the distance _z1 from the surface of the object to be detected to the center of the frosted glass satisfy the following relationship

$<span\; class="notranslate">\; |</span>\frac{{\mathrm{<span\; class="notranslate">\; zz</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; ></span>\frac{{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; D.</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}$

Among them, D is the lateral size of the spot on the frosted glass, and λ is the wavelength of the laser.

The separation method of the target image separation device based on intensity-correlated imaging, the method includes the following steps:

① After the computer sends a pulse signal to the laser, the object light path barrel detector and the reference light path surface detector, the laser sends out an energy pulse, the barrel detector measures a light intensity, denoted as y ₁ , and the surface detector measures a light intensity The field intensity distribution is recorded as I ₁ , and the second row to the last row in the matrix I ₁ are moved to the first row and then stretched into a row vector (a ₁₁ ...a _1n ) as the first row A of the measurement matrix A ₁ =(a ₁₁ …a _1n ), repeat this process m times, the frosted glass keeps rotating at a constant speed during this process, and get an m×1 vector (y ₁ ,y ₂ ,...,y _m ) ^T as a signal Y and an m×n matrix (A ₁ ,A ₂ ,...,A _m ) ^T are used as the measurement matrix A. For the two-dimensional target image, the second to the last row in the image are also moved behind the first row Then perform the transpose operation to become a one-dimensional vector, express the vector as x=(x ₁ …x _n ) ^T , and the data acquisition process of correlation imaging can be attributed to

②Target image transformation: According to the three different parts of the smoothness, edge and texture of the target image to be separated, wavelet transform can be used respectively curve wave transform discrete cosine transform to represent the target, that is, Among them, θ ₁ , θ ₂ and θ ₃ are the target transformation and The expression coefficient below, combined with the data acquisition process model in formula (2), can be obtained:

③Transform the above formula into the problem of solving convex optimization, that is, the solution optimization function is:

The problem, where τ takes a value according to the information contained in the target, and its range is (0,1), this problem can use the optimal gradient descent algorithm (Wang Shuning, Xu Jun, Huang Xiaolin, Convex Optimization [M]. Tsinghua University Publishing Society: 2013,: 443-446.) to solve and get the target x in the transformation and The lower represents the coefficients θ ₁ , θ ₂ and θ ₃ .

④ After obtaining the transformation in the cascade and After the lower coefficients θ ₁ , θ ₂ and θ ₃ , use the iterative threshold algorithm (Bobin J, Starck J L, Fadili J M, et al. Morphological component analysis: An adaptive thresholding strategy [J]. Image Processing, IEEE Transactions on, 2007 ,16(11):2675-2681.) The iterative process of repeated thresholding of this coefficient can obtain three different characteristics of the target, and the initial threshold in the iteration can be set to the largest of θ ₁ , θ ₂ and θ ₃ A coefficient, the threshold at which the iteration terminates is obtained by computing the median of the target image.

The working process of the device of the present invention is as follows:

The laser, the object optical path barrel detector and the reference optical path detector are simultaneously triggered and controlled to work simultaneously by a synchronous pulse signal sent from the computer.

(1) The light emitted by the laser passes through a filter system to obtain a better laser mode, and then passes through a diaphragm and a rotating ground glass to obtain a speckle field, which is divided into The reflected beam and the transmitted beam, the reflected beam passes through a period of free propagation and reaches the surface detector with high spatial resolution in the far field;

(2) The surface detector receives and records the spatial intensity distribution information of the light field from the frosted glass;

(3) The transmitted light beam passes through the transmissive (or reflective) target after a period of free propagation at the same distance, and then focuses on the barrel detector through the lens, and the barrel detector receives and records the transmitted light (or emitted light) passing through the target Information about the subsequent light intensity;

(4) Input the light intensity distribution information detected by the detector on the reference arm and the light intensity signal detected by the barrel detector on the object arm into the computer;

(5) The components to be separated are calculated by using the signals collected in the barrel detector, the light intensity distribution collected in the area detector and the algorithm module in the computer.

The algorithm process in the described algorithm module is:

The light field intensity distribution matrix collected by the area array detector is stretched into a row vector A _i as a row of the detection matrix A, and the light intensity measured by the barrel detector is y _i , and the required measurement is obtained after multiple measurements Matrix A and signal Y, the data acquisition model is expressed as Y=Ax, we can use three joint transformations Sparsely represent the target, then the calculated result is the target's transformation coefficients θ ₁ , θ ₂ , θ ₃ , and then use the threshold iterative algorithm to separate the coefficients step by step to obtain several morphological features of the target based on different transformations to achieve Separate work.

The advantages of the present invention are:

(1) In correlative imaging, we can use different return time signals to obtain the longitudinal information of the target through a barrel detector. Therefore, the device can be used to observe the transformation law of the components of the three-dimensional target with distance.

(2) In correlative imaging, we can get some information about the shape, feature size and reflectivity of the target, then we can use our device to get some specific image features to realize the recognition of the target or the enhancement of specific features.

(3) Some features of the image we are interested in can be obtained through this device, which can save the space we need for storage and reduce the amount of information in the transmission process.

(4) Since the recovery result of the correlation imaging combines the sparsity of the image, the features obtained by this device have a relatively high resolution.

(5), no radiation, no damage to the optical imaging device.

Description of drawings

FIG. 1 is a schematic structural diagram of a transmission-type object image separation device based on correlation imaging according to the present invention.

FIG. 2 is a schematic structural diagram of a reflective object image separation device based on correlation imaging according to the present invention.

In the figure, 1 is a laser; 2 is a focusing lens; 3 is a stop; 4 is a collimating lens; 5 is a stop; 6 is ground glass; 7 is a beam splitting prism; 11 is the object arm bucket detector; 11 is the reference arm area detector; 12 is the computer acquisition and calculation module.

Figure 3 is the image (64×64 pixels) used as the target in the experiment.

Figure 4 shows the various components obtained. Figure 4(a) and Figure 4(b) are the point features and line features separated by this device.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

Fig. 1 is a schematic structural diagram of a target image separation device based on correlated imaging. It can be seen from the figure that the device includes a laser 1, and along the direction of the laser output from the laser are a focusing lens 2, a first aperture 3, a collimating lens 4, a second aperture 5, a ground glass 6, a beam splitter 7, and an object to be detected. 8. Lens 9 and barrel detector 10, an area array detector 11 is arranged in the reflected light direction of the beam splitter 7, and the output terminals of the area array detector 11 and the barrel detector 10 are connected to the input of the computer 12 end, the output end of the computer 12 is connected to the input end of the laser 1, the area array detector 11 and the barrel detector 10, the first diaphragm 3 is located at the focal point of the focusing lens 2, and the barrel detector The detector 10 is located at the focal point of the lens 9, and the distance z between the area array detector 11 and the center of the ground glass 6 and the distance z1 from the surface of the object 8 to be detected to the center of the ground glass ₆ satisfy the relationship shown below

$<span\; class="notranslate">\; |</span>\frac{{\mathrm{<span\; class="notranslate">\; zz</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; ></span>\frac{{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; D.</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}$

Among them, D is the lateral size of the spot on the frosted glass, and λ is the wavelength of the laser.

The laser 1 , the object optical path barrel detector 10 and the reference optical path detector 11 are simultaneously triggered and controlled to work simultaneously by a synchronous pulse signal sent from the computer, and the collected data is input into the computing module 12 .

After the computer 12 sends a pulse signal to the laser 1, the object optical path barrel detector 10 and the reference optical path surface detector 11, the laser 1 sends an energy pulse, and the barrel detector 10 measures a light intensity, denoted as y ₁ , and the surface detection A light field intensity distribution measured by the device 11 is recorded as I ₁ , and the second row to the last row in the matrix I ₁ are all moved to the first row and then stretched into a row vector (a ₁₁ ...a _1n ) as a measurement matrix The first row A ₁ =(a ₁₁ ...a _1n ), repeat this process m times, the frosted glass 6 keeps rotating at a constant speed during this process, and obtain an m×1 vector (y ₁ ,y ₂ ,... ,y _m ) ^T as the signal Y and an m×n matrix (A ₁ ,A ₂ ,...,A _m ) ^T as the measurement matrix A, for the two-dimensional target image, the second row in the image to the last After a line is moved to the first line, transpose is performed to become a one-dimensional vector, and the vector is expressed as x=(x ₁ …x _n ) ^T , and the data acquisition process of associated imaging can be summarized as

Target image transformation: According to the smoothness, edge and texture of the target image to be separated, wavelet transform can be used respectively curve wave transform discrete cosine transform to represent the target, that is, Among them, θ ₁ , θ ₂ and θ ₃ are the target transformation and The expression coefficient below, combined with the data acquisition process model in formula (2), can be obtained:

Transform the above formula into the problem of solving convex optimization, that is, the solution optimization function is:

The problem, where τ takes a value according to the information contained in the target, and its range is (0,1), this problem can use the optimal gradient descent algorithm (Wang Shuning, Xu Jun, Huang Xiaolin, Convex Optimization [M]. Tsinghua University Publishing Society: 2013,: 443-446.) to solve and get the target x in the transformation and The lower represents the coefficients θ ₁ , θ ₂ and θ ₃ .

get the transformation in the cascade and After the lower coefficients θ ₁ , θ ₂ and θ ₃ , use the iterative threshold algorithm (Bobin J, Starck J L, Fadili J M, et al. Morphological component analysis: Anadaptive thresholding strategy [J]. Image Processing, IEEE Transactions on, 2007, 16(11):2675-2681.) The iterative process of repeated thresholding of this coefficient can obtain three different characteristics of the target, and the initial threshold in the iteration can be set to the largest one among θ ₁ , θ ₂ and θ ₃ Coefficient, the threshold for iteration termination is obtained by calculating the median value of the target image.

The working process of the device of the present invention is as follows:

The laser 1 , the object optical path barrel detector 10 and the reference optical path detector 11 are triggered and controlled to work simultaneously by a synchronous pulse signal sent by the computer 12 .

(1) The light emitted by the laser 1 passes through a filter system 2-4 to obtain a better laser mode, and then passes through a diaphragm 5 and a rotating ground glass 6 to obtain a speckle field, which passes through A beam splitter 7 is divided into a reflected light beam and a transmitted light beam, and the reflected light beam passes through a period of free propagation and reaches the surface detector 11 with high spatial resolution in the far field;

(2) receiving and recording the spatial distribution information of the light intensity of the light field from the frosted glass 6 by the surface detector 11;

(3) The transmitted light beam is transmitted or reflected by the target 8 after the same distance of free propagation, and then focused on the barrel detector 10 by the lens 9, and the barrel detector 10 receives and records the light intensity of the transmitted light after passing through the target information;

(4) Input the light intensity distribution information detected by the detector 11 on the reference arm and the light intensity signal detected by the barrel detector 10 on the object arm into the computer 12;

(5) Using the algorithm module in the computer 12 to calculate the signal collected by the barrel detector 10 and the light intensity distribution collected by the surface detector 11 to obtain the components to be separated. (See Figure 3).

The computing process of algorithm module in described computer 12 is as follows:

Assume that in a measurement process, the light intensity detected by the object arm barrel detector 10 is denoted as y _i , and the light field intensity distribution detected by the reference arm surface detector 11 is a two-dimensional matrix I _ij , the matrix I ₁ in The second row to the last row are all moved to the first row and stretched to become a row vector (a ₁₁ …a _1n ) as the first row A ₁ =(a ₁₁ …a _1n ) of the measurement matrix A. After multiple After measurement, different y _i constitute a detected vector Y, and different A _i can constitute a detection matrix A. Assuming the target is x, for the two-dimensional target image, the second to last row in the image is also moved to the first After one row, the transpose operation is performed to become a one-dimensional vector, which is expressed as x=(x ₁ …x _n ) ^T , considering the noise in the actual experiment, so our sampling model can be adjusted to Y=Ax+ ε, where ε represents the noise in the sampling process, and during computation, we can choose the optimal sparse transformation for the target x At this time, due to the sparsity of the signal, we can solve this kind of problem equivalently as a convex optimization problem, namely:

And because images in nature are more complex, contain more different components, and have a more sparse representation under combined transformation, so using cascaded transformations in the solution process can usually get better results, without loss of generality, we In obtaining the results shown in Figure 3, two different transformations were used, namely the wavelet transform and scale wave transform In this case, the convex optimization problem solved can be transformed into the following form:

Among them, θ ₁ is the transformation of the target x Under the expression coefficient, θ ₂ is the target x in the transformation The following represents the coefficient, ||v|| ₂ represents the Euclidean l ₂ norm of the vector v, ||v|| ₁ represents the Euclidean l ₁ norm of the vector v, τ takes a value according to the information contained in the target, and its range is (0,1).

There are many methods for solving convex optimization problems, such as greedy iterative algorithm, iterative threshold algorithm and so on. In the process of solving the convex optimization problem, the device refers to the gradient projection optimization algorithm (see Wang Shuning, Xu Jun, Huang Xiaolin, Convex Optimization [M]. Tsinghua University Press: 2013,: 443-446.) to modify the optimization goal This method is also an algorithm that solves convex optimization problems and can reach the best speed of linear convergence, and is an optimal gradient descent algorithm.

In the process of solving the optimization problem, we can obtain the representation coefficients θ ₁ and θ ₂ of the target x under the two transformations. Using these representation coefficients, we can obtain different components by iterative thresholding. The specific process is as follows:

① Use these representation coefficients θ ₁ and θ ₂ to do inverse transformation to restore the target to estimate the noise in the target image. The noise can be used to estimate the minimum value λ _min of the threshold, and the representation coefficients lower than the threshold can be omitted. to reduce noise. Select the maximum value of these representation coefficients as the initial threshold λ ₁ , set the number of iterations m, and record the two types of features of points and lines in the target as shown in Figure 2 as {x _j } _j=1,2 .

② has been calculated In the case of , the residual term in the kth iteration process can be calculated by the following formula

${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>{\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&NotEqual;</span>\mathrm{<span\; class="notranslate">\; j</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span>$

Where x is the initial target, that is, the target image recovered by using the coefficients θ ₁ and θ ₂ in ①.

③Calculation and current The representation coefficient corresponding to the component Then thresholding is performed on this coefficient, and its mathematical form is expressed as:

in, Indicates that the coefficients are thresholded, and the threshold is In the first iteration process, its coefficients can be directly replaced by θ ₁ and θ ₂ in ①.

④ Use the coefficients obtained in ③ to do inverse transformation to recover the kth iteration process

⑤ Solve the threshold value that needs to be used in the next iteration process

${\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; min</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 10</span>}<span\; class="notranslate">\; )</span>$

Circularly solve the ②-⑤ process until k=m, and finally get That is, the separated variables we need.

In Figure 3, the picture shown is the object used as the target in the experiment. The target is only composed of two types of features, one is points, and the size of these points is 232.5μm×232.5μm; the other Class features are lines, and the width of these lines is 93 μm. The pixel size of the image is 64×64, because the smooth feature points in the target have very sparse coefficients in the case of wavelet transform, and the edge feature lines have sparse coefficients after the curvelet transform, so when the target is some smooth parts and edges In the case of combination, usually the combination of wavelet transform and curvelet transform can separate these features; when the target still contains some complex periodic texture features, we can also use discrete cosine transform to separate these features. Figure 4(a) and Figure 4(b) show the point feature components and line feature components separated by the device when the number of samples is 1200, and the wavelet transform and curvelet transform are cascaded.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent replacements to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention, and all of them should be included in the scope of the present invention. within the scope of the claims.

Claims (2)

1. The utility model provides a target image separator based on intensity correlation formation of image, characterized in that the device includes laser instrument (1), and the laser direction of following this laser instrument outgoing is focusing lens (2), first diaphragm (3), collimating lens (4), second diaphragm (5), ground glass (6), beam splitter (7), treats detected object (8), lens (9) and bucket detector (10) in proper order, beam splitter (7) the reverberation direction set up area array detector (11), the output termination of area array detector (11) and bucket detector (10) the input of computer (12), this calculation should calculate the input of computer (12), this calculationThe output of machine (12) with the input of laser instrument (1), area array detector (11) and bucket detector (10) link to each other, first diaphragm (3) are located the focus of focusing lens (2), bucket detector (10) be located the focus of lens (9), area array detector (11) to the distance z at ground glass (6) center and wait to detect the distance z at object (8) surface to ground glass (6) center₁Satisfies the relationship shown below

<math> <mrow> <mo>|</mo> <mfrac> <msub> <mi>zz</mi> <mn>1</mn> </msub> <mrow> <mi>z</mi> <mo>-</mo> <msub> <mi>z</mi> <mn>1</mn> </msub> </mrow> </mfrac> <mo>|</mo> <mo>></mo> <mfrac> <msup> <mrow> <mn>2</mn> <mi>D</mi> </mrow> <mn>2</mn> </msup> <mi>&lambda;</mi> </mfrac> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow> </math>

Wherein D is the transverse size of a light spot on the ground glass, and lambda is the wavelength of the laser.

2. A method of separating an object image separating apparatus based on intensity correlation imaging as claimed in claim 1, characterized in that the method comprises the steps of:

firstly, after a computer sends out a pulse signal to a laser (1), a barrel detector (10) and a surface detector (11), the laser (1) sends out an energy pulse, and the barrel detector (10) measures and obtains a light intensity which is marked as y₁The surface detector (11) measures a light field intensity distribution and records the light field intensity distribution as a matrix I₁The matrix I₁After moving to the first row from the second row to the last row in the drawing, the drawing is performed into a row vector (a)₁₁…a_1n) First row A as measurement matrix A₁=(a₁₁…a_1n) Repeating the process m times, and keeping the ground glass in a constant-speed rotation state in the process to obtain an m multiplied by 1 vector (y)₁,y₂,...,y_m)^TAs a matrix (A) of signals Y and m × n₁,A₂,...,A_m)^TAs the measurement matrix a, for a two-dimensional target image, similarly, after moving the second line to the last line in the image to the first line, the two-dimensional target image is transposed to form a one-dimensional vector, and the vector is represented as x = (x)₁…x_n)^TWherein, the symbol T in the upper right corner of the matrix represents the transposition of the matrix, and the data acquisition process of the correlation imaging is summarized as follows:

secondly, converting the target image: wavelet transform can be used separately according to the three different parts of smooth, edge and texture of the target image to be separatedCurvelet transformDiscrete cosine transformTo represent an object, i.e.Wherein, theta₁、θ₂And theta₃Is targeted at transformingAndthe coefficient is obtained by combining the data acquisition process model in the formula (2)To:

thirdly, converting the formula (3) into a problem of solving convex optimization, namely solving an optimization function as follows:

wherein, tau takes values according to the information contained in the target, the range is (0,1), the optimal gradient descent algorithm is used for solving, and the target x is obtained in the transformationAndcoefficient of lower theta₁、θ₂And theta₃；

Obtaining the transformations in cascadeAndcoefficient of lower theta₁、θ₂And theta₃Then, an iterative process of repeatedly thresholding the coefficient by using an iterative threshold algorithm is used to obtain three different characteristics of the target, and an initial threshold in the iterative process is set to be theta₁、θ₂And theta₃The threshold for the termination of the iteration is obtained by calculating the median value of the target image.