Single-pixel imaging method based on large-capacity spatial multiplexing metasurface

Abstract

The imaging method based on a large-capacity spatial multiplexing metasurface disclosed in the present invention relates to a single-pixel imaging method based on a large-capacity spatial multiplexing metasurface, and belongs to the field of micro-nano optics and single-pixel imaging application technology. The implementation method of the present invention is: using a metasurface with a randomly distributed nanopore structure as a mask in a single-pixel imaging system, it is possible to achieve fast and high-resolution imaging without polarization selection, and its working range can be from the entire visible light to the near-infrared band; by switching the mask pattern by moving the metasurface, it is possible to achieve spatial multiplexing, obtain more mask patterns in a limited spatial area, and improve the information capacity of the metasurface. The present invention can achieve fast and high-resolution single-pixel imaging from the visible light to the near-infrared band. The single-pixel imaging system based on the metasurface is small in size and simple in optical path, which can make the imaging system more compact and integrated, and broaden the application range of the single-pixel imaging system.

Description

Technical Field

[0001] The present invention relates to a single-pixel imaging method, in particular to a single-pixel imaging method based on a large-capacity spatial multiplexing metasurface, and belongs to the technical field of micro-nano optics and single-pixel imaging applications. Background Art

[0002] As a new type of computational imaging technology, single-pixel imaging uses different mask patterns to modulate the target, and then performs correlation operations with the corresponding intensity measurements to reconstruct the target image. Compared with traditional focal plane array technology, single-pixel imaging combines spatial light modulators and single-pixel detectors to replace pixel array detectors such as CCD and CMOS. It has competitive advantages in lower cost, higher detection efficiency, faster response time and wider spectral range. In recent years, application research based on single-pixel imaging has made great progress and has been widely used in terahertz imaging, microscopic imaging, three-dimensional imaging, hyperspectral imaging, time-resolved imaging and remote sensing imaging.

[0003] Metasurface is an artificially designed structure consisting of an array of nanoantennas or nanoresonators with ultra-thin structures and sub-wavelength characteristic dimensions. With the development of current micro-nano processing technology and large-scale integrated design technology, metasurfaces can arbitrarily control the light field through the design of their surfaces, showing a more efficient control capability of the amplitude, phase and polarization of the incident electromagnetic wave. Compared with traditional optical elements that control the light field through phase accumulation during spatial propagation, metasurfaces directly control the light field through the interaction between the incident light and the designed surface, providing great potential for metasurfaces as a substitute for traditional large optical elements. In addition to being able to arbitrarily control the light field, metasurfaces also have the advantages of low cost, high compactness, easy integration and ultra-small pixels, making them a new type of control element for high-resolution imaging.

[0004] At present, single-pixel imaging faces problems such as slow imaging speed and low imaging resolution. Combining single-pixel imaging with micro-nano optics and using metasurfaces as mask control devices in single-pixel imaging can miniaturize the amplitude control devices in the imaging system, increase information capacity, achieve better mask control, and so on. In recent years, the research work that has been reported on combining single-pixel imaging with micro-nano optics has mainly focused on terahertz imaging and optical encryption, lacking analysis and research on the imaging speed and resolution of single-pixel imaging. In addition, in order to improve the imaging speed of single-pixel imaging, some solutions with high projection refresh rates have been proposed, such as high-speed LED lighting modules, rotating masks encoded by cyclic patterns, and static optical structures. However, these research works lack the advantages of metasurfaces such as sub-wavelength feature size and high compactness, making it difficult to achieve multi-pixel control and higher-resolution imaging. Therefore, it is necessary to study a new single-pixel imaging solution that can achieve both high-speed imaging and high-resolution imaging. Metasurfaces can be used as a means to solve this problem by replacing the spatial light modulator in traditional single-pixel imaging. By utilizing their simple amplitude control mechanism, they can overcome the challenges faced by single-pixel imaging in imaging speed and resolution, and achieve the integrability of single-pixel imaging systems.

[0005] Single-pixel imaging often requires a lot of different mask patterns to modulate the target, which requires the metasurface to have a larger spatial capacity. In order to improve space utilization, the method of obtaining more information in a limited space without increasing the number of pixels is called multiplexing technology. Usually, single-pixel imaging can use random binary codes as masks. Metasurfaces can design such random binary codes and obtain more mask patterns through spatial multiplexing. Summary of the invention

[0006] Aiming at the following technical problems existing in the existing single-pixel imaging technology: slow imaging speed, low imaging resolution, poor compactness, and failure to meet the needs of practical applications. The main purpose of the present invention is to provide a single-pixel imaging method based on a large-capacity spatial multiplexing metasurface, which can achieve fast and high-resolution single-pixel imaging from visible light to near-infrared bands. In addition, single-pixel imaging based on metasurfaces can effectively increase the spatial information capacity of metasurfaces, make the imaging system more compact and integrated, and achieve non-polarization selectivity of imaging.

[0007] The objective of the present invention is achieved through the following technical solutions:

[0008] In order to obtain a reconstructed image of the target object in single-pixel imaging, a series of mask patterns are required to amplitude modulate the target object, collect the total amplitude intensity of the corresponding light field after modulation, and then use a single-pixel imaging algorithm to obtain the reconstructed image of the target object. The imaging method based on a large-capacity spatial multiplexing metasurface disclosed in the present invention uses a metasurface with a randomly distributed nanopore structure as a mask in a single-pixel imaging system, which can achieve fast and high-resolution imaging without polarization selection, and its working range can be from the entire visible light to the near-infrared band. By switching the mask pattern by moving the metasurface, spatial multiplexing can be achieved, more mask patterns can be obtained in a limited spatial area, and the information capacity of the metasurface can be improved. The single-pixel imaging system based on the metasurface is small in size and simple in optical path, which can make the imaging system more compact and integrated, and broaden the application range of the single-pixel imaging system.

[0009] The single-pixel imaging method based on a large-capacity spatial multiplexing metasurface disclosed in the present invention comprises the following steps:

[0010] Step 1: Generate a randomly distributed binary mask, and decompose the mask into a series of modulation mask patterns according to the number of pixels required for image modulation, so as to generate more modulation mask patterns in limited spatial information; use the generated series of modulation mask patterns to amplitude modulate the target object to obtain the corresponding total amplitude intensity after modulation, and then use a single-pixel imaging algorithm to perform correlation operations on the modulation pattern and the corresponding detection intensity to test the imaging effect of the mask until a reconstructed image that meets the preset imaging effect is obtained.

[0011] The single-pixel imaging algorithm modulates the target using different mask patterns, and then performs correlation operations with the corresponding intensity measurements to reconstruct the target image. Typically, mask patterns used for single-pixel imaging include Hadamard coding, Fourier coding, or other optimized coding. However, the coding has a fixed amplitude distribution, resulting in no cross-information between different mask patterns, and the information capacity of the space cannot be increased. Random binary coding is a random binary amplitude distribution, and a new mask pattern can be obtained by changing part of the binary coding information. Using this property of binary coding, spatial multiplexing can be achieved and spatial capacity can be increased. Single-pixel imaging algorithms include differential algorithms and compressed sensing algorithms.

[0012] Compressed sensing is a technology for finding sparse solutions to underdetermined linear systems, which is used to acquire and reconstruct sparse or compressible signals. The compressed sensing algorithm can recover the complete original signal from the acquired undersampled signal. Its main reconstruction method is to recover the original information by solving the norm optimization problem. In addition, the compressed sensing algorithm has been widely used in signal processing, imaging, image classification and other fields. For the compressed sensing theory, two conditions are required to recover the original signal: 1. The collected undersampled data contains all the information of the original signal. 2. The compressed sensing algorithm can recover the complete original information from the undersampled information. Single-pixel imaging often requires a large number of sampling times, which leads to an increase in sampling time and complex data processing.

[0013] The model of single pixel imaging is shown in formula (1):

[0014] S i =∫∫P i (x,y)I(x,y)dxdy (1)

[0015] Wherein the subscript i = 1, 2, ..., M is the number of time-arranged measurements, M represents the total number of measurements, P represents the binary matrix of the mask, I represents the intensity distribution of the target object, S represents the corresponding total measured light intensity; x, y represent the spatial coordinates on the object plane. Formula (1) can be used to restore the image by solving the inverse process. Using the compressed sensing algorithm, formula (1) can be transformed into a form for solving the optimal solution:

[0016]

[0017] Where Φ represents the sensing matrix determined by the M-th mask pattern, S represents the M×1 column vector of the intensity signal collected by the detector, I is the N×1 column vector obtained by shaping the target object, and N represents the total number of pixels in the image. represents the L-2 Euclidean norm, Ψ is the regularization term with sparsity constraint on the image, and τ represents the regularization parameter used to adjust the relative weight between residual and sparsity. An appropriate value needs to be selected to achieve optimal reconstruction.

[0018] There are many compressed sensing algorithms for single-pixel imaging. In order to improve the reconstruction quality of the image, as a preferred method, a two-step iterative shrinkage algorithm TWIST is used to reconstruct the image. Each iteration of the two-step iterative shrinkage algorithm TWIST depends on the previous two iterations, rather than just the previous iteration, making the iterative shrinkage process more accurate and faster.

[0019] The implementation method of step 1 includes the following steps:

[0020] Step 1.1: Generate a randomly distributed binary code and decompose the code into a series of mask patterns in sequence, so as to generate more modulation mask patterns in a limited space and improve the spatial capacity.

[0021] Step 1.2: Use the generated series of mask patterns to amplitude modulate the target object to obtain the corresponding total amplitude intensity after modulation, and use the single-pixel imaging algorithm to perform correlation operations on the mask and total intensity:

[0022] S i =∫∫P i (x,y)I(x,y)dxdy (1)

[0023] Wherein the subscript i=1, 2, ..., M is the number of time-arranged measurements, M represents the total number of measurements, P represents the binary matrix of the mask, I represents the intensity distribution of the target object, S represents the corresponding total measured light intensity; x, y represent the spatial coordinates on the object plane, respectively.

[0024] Step 1.3: In order to reduce the number of measurements and improve the reconstruction quality, a compressed sensing algorithm is used. The compressed sensing algorithm reconstructs the object by solving the following optimization problem:

[0025]

[0026] Where Φ represents the sensing matrix determined by the M-th mask pattern, S represents the M×1 column vector of the intensity signal collected by the detector, I is the N×1 column vector obtained by shaping the target object, and N represents the total number of pixels in the image. represents the L-2 Euclidean norm, Ψ is the regularization term with sparsity constraint on the image, and τ represents the regularization parameter used to adjust the relative weight between residual and sparsity. An appropriate value needs to be selected to achieve optimal reconstruction.

[0027] Step 1.4: The two-step iterative shrinkage algorithm TWIST is used to iteratively optimize the target value. Each iteration of this iterative process depends on the previous two iterations, rather than just the previous iteration, making the iterative shrinkage process more accurate and faster.

[0028] Step 1.5: Select the total variation TV as the regularization term with sparsity constraint:

[0029]

[0030] Where dx(I) and dy(I) represent the first-order gradient operations in the horizontal and vertical directions, which can suppress noise while avoiding excessive smoothing of the image.

[0031] Step 1.6: Set appropriate parameters, and after iterative processing from steps 1.4 to 1.5, obtain the reconstructed image of the target object. Then use the iterative total error compensation ITEC algorithm to optimize the reconstruction quality under random mask modulation:

[0032]

[0033] Among them I n (x,y) is the image obtained by n iterations. and Respectively expressed as:

[0034]

[0035]

[0036] Where P r and S r represent the mask pattern and detection intensity in the rth measurement, S 0 It represents the total intensity obtained by the detector under the parallel beam. The process is to compensate the errors of the binary values ​​in the mask pattern respectively, and obtain the reconstructed image that meets the preset effect by iterating a series of mask patterns.

[0037] Step 1.7: After the target image is reconstructed in steps 1.1 to 1.6, the reconstruction quality of the image is tested. The image reconstruction quality is related to the parameters in TWIST and the number of iterations of ITEC, that is, the modulation pattern and the corresponding detection intensity are correlated using a two-step iterative shrinkage algorithm to test the imaging effect of the mask, and the relevant parameters are modified according to the reconstruction quality until a reconstructed image that meets the preset effect is obtained.

[0038] Step 2: The metasurface used to achieve large-capacity spatial multiplexing is formed by punching holes in the gold film at different positions. Among them, the opaque gold film part is equivalent to the pixel with amplitude modulation of "0" in the mask, and the nanohole part is equivalent to the pixel with amplitude modulation of "1" in the mask. It is necessary to optimize the design of the relevant dimensions of the nanohole so that the metasurface can achieve this binary distribution. The relevant geometric dimensions include the thickness H of the gold film, the radius R of the nanohole and the period P of the metasurface unit.

[0039] The metasurface used to achieve large-capacity spatial multiplexing is composed of randomly distributed nanopores. The transmittance of the opaque gold film part and the nanopore part in the metasurface can be expressed by T1 and T2. If T2 is much larger than T1, the metasurface can achieve good binary amplitude modulation.

[0040] When the metasurface unit period P is fixed, the radius R of the simulated nanopore and the thickness H of the gold film are scanned. The transmission intensity of the parallel light after being modulated by the metasurface unit is calculated from the simulated electromagnetic field data to obtain T1 and T2. During the simulation, the metasurface unit period P should be reasonably selected, and the film material should be determined to be gold, and T2 should be ensured to be much larger than T1 as much as possible. The greater the contrast of the metasurface, the better the effect as a binary amplitude modulation mask.

[0041] Metasurfaces are used as amplitude modulation masks in single-pixel imaging methods. Metasurfaces based on nanopore structures have the following advantages: Nanopore structures have the advantages of simple structure and easy manufacturing, which makes it easier to achieve large-area manufacturing, thereby obtaining higher information capacity. In addition, nanopores are centrosymmetric structures, so nanopore-based metasurfaces have polarization-independent characteristics. The combination of gold film and nanopores can achieve good binary modulation effects from the entire visible light to near-infrared band. And because the pixel size of the nanopores in the metasurface reaches the sub-wavelength level, the mask modulation pixel size is significantly reduced, and smaller pixel modulation can be achieved, thereby achieving high-resolution imaging. The single-pixel imaging system based on metasurfaces is small in size and simple in optical path, which can make the imaging system more compact and integrated, and broaden the application range of single-pixel imaging systems.

[0042] Step 3: Encode the binary intensity information of the mask generated in step 1 into the position arrangement of nanoholes in the metasurface. Using the standard micro-nano processing technology of electron beam etching, a metasurface with the random nanohole distribution is produced to achieve good binary amplitude modulation. The metasurface has the property of spatial multiplexing. First, parallel light carrying the image information of the object is incident on a part of the metasurface, which is the first mask pattern. By moving the metasurface to switch different modulation areas, a series of mask patterns can be obtained by moving in the horizontal direction, and a new mask pattern can be obtained for each movement of a metasurface unit period, and the difference between the mask pattern and the previous mask pattern is only one column of pixels. When the metasurface moves to the edge of the overall area in the horizontal direction, a new projection pattern is obtained again by moving a pixel distance in the vertical direction, and the difference between the projection pattern and the previous projection pattern is only one row of pixels. Then adjust the moving direction of the metasurface to move in the opposite direction of the previous horizontal movement. By analogy, spatial multiplexing can be achieved by moving the metasurface, which can obtain more mask patterns in a limited spatial area and improve the information capacity of the metasurface mask.

[0043] Step 4: Realize spatial multiplexing by moving the metasurface in step 3, and obtain a series of mask patterns. Use the modulation pattern and the corresponding total detection intensity to complete image reconstruction using a single-pixel imaging algorithm. In addition, the use of a mobile metasurface for modulation significantly accelerates the switching between masks compared to single-pixel imaging using a DMD as a spatial light modulator. The switching frequency of the mask is determined by the speed of movement. By speeding up the speed of movement, the switching rate of the mask can be increased, thereby achieving single-pixel imaging with a faster imaging speed.

[0044] Beneficial effects:

[0045] 1. The single-pixel imaging method based on a large-capacity spatial multiplexing metasurface disclosed in the present invention realizes the switching of different mask patterns by moving the metasurface. During the movement process, a new mask pattern can be generated by moving only one pixel, and most of the cross information exists between adjacent masks, realizing spatial multiplexing. By utilizing the characteristics of the spatial multiplexing, more mask patterns can be obtained in a limited spatial area, thereby improving the information capacity of the metasurface mask.

[0046] 2. The single-pixel imaging method based on a large-capacity spatial multiplexing metasurface disclosed in the present invention selects a gold film punching method as a method for making a metasurface, which has a simple structure and is easier to achieve large-area manufacturing, thereby obtaining a higher information capacity. The structure of the gold film and the nanopores can achieve a good binary modulation effect from the entire visible light to the near-infrared band that is independent of polarization. The single-pixel imaging system based on the metasurface is small in size and simple in optical path, which can make the imaging system more compact and integrated, and broaden the application range of the single-pixel imaging system.

[0047] 3. The single-pixel imaging method based on large-capacity spatial multiplexing metasurface disclosed in the present invention selects randomly distributed nanopores as amplitude modulation metasurface structural units, encodes the mask that meets the preset imaging effect through the two-dimensional arrangement of sub-wavelength nanopores, and reduces the size of the mask modulation pixel to the nanometer level, which can realize the modulation of smaller pixels and thus achieve high-resolution imaging.

[0048] 4. The single-pixel imaging method based on a large-capacity spatial multiplexing metasurface disclosed in the present invention replaces the traditional single-pixel imaging method of generating a mask by refreshing the DMD with a method of generating a mask by a mobile metasurface. The mobile metasurface method can achieve spatial multiplexing and obtain a series of mask patterns. The image reconstruction is completed using a single-pixel imaging algorithm through the modulation pattern and the corresponding total detection intensity. In addition, the modulation using a mobile metasurface significantly accelerates the switching between masks compared to single-pixel imaging using a DMD as a spatial light modulator. The switching frequency of the mask is determined by the rate of movement. By speeding up the movement rate, the switching rate of the mask can be increased, thereby achieving single-pixel imaging with a faster imaging speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a flow chart of a single-pixel imaging method based on a large-capacity spatial multiplexing metasurface of the present invention;

[0050] Figure 2 is the transmittance of the nanopore unit and the gold film unit in Example 1 of the present invention; wherein, Figure 2 (a) is a schematic diagram of the structure of the nanopore unit and the gold film unit. Figure 2 (b) is the transmittance of the nanopore unit and the gold film unit structure, Figure 2 (c) The contrast of transmittance between nanopore unit and gold film unit structure.

[0051] Figure 3 This is the spatial multiplexing method and single-pixel imaging process of the metasurface in Example 1.

[0052] Figure 4 This is the experimental optical path diagram used in Example 1;

[0053] Among them: 1—light source, 2—aperture, 3—target object, 4—lens, 5—lens, 6—displacement platform, 7—metasurface, 8—lens, 9—single-pixel detector.

[0054] Figure 5 This is a graph of the experimental results obtained in Example 1, where: Figure 5 (a)-(c) are the reconstructed images of different letters under 633nm incident light. Figure 5 (d)-(f) are the reconstructed images of the letter G at different wavelengths. Figure 5 (g) is the reconstructed image of the resolution plate. DETAILED DESCRIPTION

[0055] In order to better illustrate the purpose and advantages of the present invention, the invention is further described below with reference to the accompanying drawings and examples.

[0056] Embodiment 1: Realize broadband, fast, high-resolution, and compact single-pixel imaging without polarization selection.

[0057] like Figure 1 As shown, the single-pixel imaging method based on a large-capacity spatial multiplexing metasurface disclosed in this embodiment is specifically implemented in the following steps:

[0058] Step 1: Generate a randomly distributed binary mask, and decompose the mask into a series of modulation mask patterns according to the number of pixels required for image modulation, so as to generate more modulation mask patterns in limited spatial information; use the generated series of modulation mask patterns to amplitude modulate the target object to obtain the corresponding total amplitude intensity after modulation, and then use a single-pixel imaging algorithm to perform correlation operations on the modulation pattern and the corresponding detection intensity to test the imaging effect of the mask until a reconstructed image that meets the preset imaging effect is obtained.

[0059] The specific implementation method of step 1 includes the following steps:

[0060] The implementation method of step 1 includes the following steps:

[0061] Step 1.1: Generate a randomly distributed binary code and decompose the code into a series of mask patterns in sequence, so as to generate more modulation mask patterns in a limited space and improve the spatial capacity.

[0062] Step 1.2: Use the generated series of mask patterns to amplitude modulate the target object to obtain the corresponding total amplitude intensity after modulation, and use the single-pixel imaging algorithm to perform correlation operations on the mask and total intensity:

[0063] S i =∫∫P i (x,y)I(x,y)dxdy (1)

[0064] Wherein the subscript i=1, 2, ..., M is the number of time-arranged measurements, M represents the total number of measurements, P represents the binary matrix of the mask, I represents the intensity distribution of the target object, S represents the corresponding total measured light intensity; x, y represent the spatial coordinates on the object plane, respectively.

[0065] Step 1.3: In order to reduce the number of measurements and improve the quality of reconstruction, a compressed sensing algorithm is used, which reconstructs the object by solving the following optimization problem:

[0066]

[0067] Where Φ represents the sensing matrix determined by the M-th mask pattern, S represents the M×1 column vector of the intensity signal collected by the detector, I is the N×1 column vector obtained by shaping the target object, and N represents the total number of pixels in the image. represents the L-2 Euclidean norm, Ψ is the regularization term with sparsity constraint on the image, and τ represents the regularization parameter used to adjust the relative weight between residual and sparsity. An appropriate value needs to be selected to achieve optimal reconstruction.

[0068] Step 1.4: The two-step iterative shrinkage algorithm TWIST is used to iteratively optimize the target value. Each iteration of this iterative process depends on the previous two iterations, rather than just the previous iteration, making the iterative shrinkage process more accurate and faster.

[0069] Step 1.5: Select the total variation TV as the regularization term with sparsity constraint:

[0070]

[0071] Where dx(I) and dy(I) represent the first-order gradient operations in the horizontal and vertical directions, which can suppress noise while avoiding excessive smoothing of the image.

[0072] Step 1.6: Set appropriate parameters, and after iterative processing from steps 1.4 to 1.5, obtain the reconstructed image of the target object. Then use the iterative total error compensation ITEC algorithm to optimize the reconstruction quality under random mask modulation:

[0073]

[0074] Among them I n (x,y) is the image obtained by n iterations. and Respectively expressed as:

[0075]

[0076]

[0077] Where P r and S r represent the mask pattern and detection intensity in the rth measurement, S 0 It represents the total intensity obtained by the detector under the parallel beam. The process is to compensate the errors of the binary values ​​in the mask pattern respectively, and obtain the reconstructed image that meets the preset effect by iterating a series of mask patterns.

[0078] Step 1.7: After the target image is reconstructed in steps 1.1 to 1.6, the reconstruction quality of the image is tested. The image reconstruction quality is related to the parameters in TWIST and the number of iterations of ITEC, that is, the modulation pattern and the corresponding detection intensity are correlated using a two-step iterative shrinkage algorithm to test the imaging effect of the mask, and the relevant parameters are modified according to the reconstruction quality until a reconstructed image that meets the preset effect is obtained.

[0079] Step 2: The metasurface used to achieve large-capacity spatial multiplexing is formed by punching holes in the gold film at different positions. Among them, the opaque gold film part is equivalent to the pixel with amplitude modulation of "0" in the mask, and the nanohole part is equivalent to the pixel with amplitude modulation of "1" in the mask. The relevant dimensions of the nanohole need to be optimized so that the metasurface can achieve this binary distribution. The relevant geometric dimensions include the thickness of the gold film H = 100nm, the radius of the nanohole R = 180nm and the period of the metasurface unit P = 500nm.

[0080] Figure 2 is the transmittance of the nanopore unit and the gold film unit in the embodiment of the present invention; wherein, Figure 2 (a) is a schematic diagram of the structure of the nanopore unit and the gold film unit. Figure 2 (b) is the transmittance of the nanopore unit and the gold film unit structure, Figure 2 (c) is the contrast of the transmittance of the nanopore unit and the gold film unit structure. The metasurface used to achieve large-capacity spatial multiplexing is composed of randomly distributed nanopores. The transmittance of the opaque gold film part and the nanopore part in the metasurface is represented by "0" and "1". If the transmittance of "1" is much greater than "0", it means that the metasurface can achieve good binary amplitude modulation.

[0081] Based on the finite difference time domain method FDTD, the simulated electromagnetic field data is used to calculate the transmission intensity of parallel light after being modulated by the metasurface unit, and the transmittance of "0" and "1" is obtained. It can be seen that Figure 2 (b) During simulation, the unit period P of the metasurface should be reasonably selected, that is, P = 500nm, and the film material should be determined to be gold, and the transmittance of "1" should be much greater than "0". The greater the contrast, the better the effect of the metasurface as a binary amplitude modulation mask. Figure 2 In (c), the contrast between the two can be obtained, where the part with the smallest contrast is about 5, which is enough to distinguish the transmittance of "0" and "1" well and realize binary amplitude modulation.

[0082] Step 3: Encode the binary intensity information of the mask generated in step 1 into the position arrangement of nanoholes in the metasurface. Use standard electron beam lithography micro-nano processing technology to produce the metasurface with the random nanohole distribution to achieve good binary amplitude modulation.

[0083] The metasurface also has the property of spatial multiplexing. Figure 3 This is the spatial multiplexing method and single-pixel imaging process of the metasurface in Example 1. The generated metasurface mask is shaped into a corresponding matrix form, and then parallel light carrying the object image information is incident on a part of the metasurface. This area is the first mask pattern, and its matrix form is P1-1 ; and add up the pixels of the modulated amplitude distribution to obtain the total intensity I 1-1 By moving the metasurface to switch different modulation areas, a series of mask patterns can be obtained by moving horizontally. 1-1 To P 1-k ; and each time a metasurface unit period is moved, a new mask pattern can be considered to be obtained, and the difference between the mask pattern and the previous mask pattern is only one column of pixels, such as P 1-1 and P 1-2 There are three columns of information that are the same between them, which realizes spatial multiplexing and increases the information capacity in a certain space. When the metasurface moves horizontally to the edge of the overall area, that is, P 1-k At P, a new projection pattern is obtained by moving one pixel in the vertical direction, and the difference between this projection pattern and the previous projection pattern is only one line of pixels. Then adjust the moving direction of the metasurface in the opposite direction of the previous horizontal movement until P 2-1 By analogy, we can eventually move to P k-k The position of the projection image can be adjusted to obtain k×k different mask patterns. By moving the metasurface to achieve spatial multiplexing, more mask patterns can be obtained in a limited spatial area, thereby improving the information capacity of the metasurface mask. Different projection patterns can modulate the target image to obtain different intensity information. The corresponding intensity information is collected using a single-pixel detector, i.e., I 1-1 To I k-k .

[0084] Step 4: Realize spatial multiplexing by moving the metasurface in step 3, and obtain a series of mask patterns. Use the modulation pattern and the corresponding total detection intensity to complete image reconstruction using a single-pixel imaging algorithm. In addition, the use of a mobile metasurface for modulation significantly accelerates the switching between masks compared to single-pixel imaging using a DMD as a spatial light modulator. The switching frequency of the mask is determined by the speed of movement. By speeding up the speed of movement, the switching rate of the mask can be increased, thereby achieving single-pixel imaging with a faster imaging speed.

[0085] Figure 4The experimental optical path diagram used in the embodiment includes a light source 1, an aperture 2, a target object 3, a lens 4, a lens 5, a displacement platform 6, a metasurface 7, a lens 8, and a single-pixel detector 9. The aperture 2 is used to reduce the parallel light emitted by the light source so that the light can only illuminate a part of the metasurface; the lens 4 and the lens 5 constitute a 4f system, which is used to image the target object onto the front surface of the metasurface; the metasurface 7 can move in the horizontal and vertical directions in the optical path through the displacement platform 6, thereby achieving different amplitude modulations; the lens 8 is used to converge the light beam and transmit the converged light beam to the detector; the single-pixel detector 9 is used to receive the total intensity information transmitted by the optical path.

[0086] In actual imaging, the fast imaging capability of the scheme is first verified. Parallel light with a wavelength of 633nm is incident vertically on the target image and then modulated by the metasurface. The modulated light is collected by a single-pixel detector. The object is reconstructed using a single-pixel imaging algorithm based on the total intensity information collected by the mask modulation and the detector. The reconstructed images of the letters B, I, and T are restored by the algorithm. The experimental results are shown in Figure 2. Figure 5 (a)-(c) are shown. According to the experimental results, this scheme can achieve good image reconstruction, and the imaging speed is determined by the movement speed of the metasurface, which is consistent with the design.

[0087] To verify the wide spectrum imaging capability of the scheme, parallel light with wavelengths of 473nm, 532nm and 633nm was incident respectively, and the object image was reconstructed using the same method as above. The reconstructed images of the letter G at different wavelengths can be obtained. The experimental results are as follows: Figure 5 (d)-(f) According to the experimental results, the scheme can achieve image reconstruction at different wavelengths, which is consistent with the design.

[0088] Finally, the high-resolution imaging capability of the scheme was verified. The resolution plate was used as the target image and the above single-pixel imaging scheme was used to reconstruct the image. The experimental results are shown in the figure. Figure 5 According to the experimental results, the scheme can distinguish a line width of about 2μm. This result is affected by the experimental equipment, but in theory it can achieve imaging with a resolution of 500nm, which is basically consistent with the design.

[0089] It can be seen that the single-pixel imaging method based on large-capacity spatial multiplexing metasurface disclosed in this embodiment uses the metasurface with randomly distributed nanopore structures as a mask in the single-pixel imaging system, which can achieve fast and high-resolution imaging without polarization selection, and its working range can be from the entire visible light to the near-infrared band. By switching the mask pattern by moving the metasurface, spatial multiplexing can be achieved, more mask patterns can be obtained in a limited spatial area, and the information capacity of the metasurface can be improved. The single-pixel imaging system based on the metasurface is small in size and simple in optical path, which can make the imaging system more compact and integrated, and broaden the application range of the single-pixel imaging system.

[0090] The specific description above further illustrates the purpose, technical solutions and beneficial effects of the invention in detail. It should be understood that the above is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A single-pixel imaging method based on a large-capacity spatial multiplexing metasurface, characterized by: The following steps are included: Step 1: Generate a randomly distributed binary mask, and decompose the mask into a series of modulation mask patterns according to the number of pixels required for image modulation, so as to generate more modulation mask patterns in limited spatial information; use the generated series of modulation mask patterns to amplitude modulate the target object to obtain the corresponding total amplitude intensity after modulation, and then use a single-pixel imaging algorithm to perform correlation operations on the modulation pattern and the corresponding detection intensity to test the imaging effect of the mask until a reconstructed image that meets the preset imaging effect is obtained; Step 2: The metasurface used to realize large-capacity spatial multiplexing is formed by punching holes in a gold film at different positions; wherein the opaque gold film portion is equivalent to the pixel with amplitude modulation of "0" in the mask, and the nanohole portion is equivalent to the pixel with amplitude modulation of "1" in the mask; the relevant dimensions of the nanohole need to be optimized and designed so that the metasurface can realize binary distribution; the relevant dimensions include the thickness H of the gold film, the radius R of the nanohole and the period P of the metasurface unit; Step 3: Encode the binary intensity information of the mask generated in step 1 into the position arrangement of nanoholes in the metasurface; Use the standard micro-nano processing technology of electron beam etching to produce a metasurface with random nanohole distribution to achieve good binary amplitude modulation; The metasurface has the property of spatial multiplexing; First, parallel light carrying the image information of the object is incident on a part of the metasurface, which is the first mask pattern; By moving the metasurface to switch different modulation areas, a series of mask patterns can be obtained by moving in the horizontal direction, and a new mask pattern can be obtained each time a distance of a metasurface unit period is moved, and the difference between the mask pattern and the previous mask pattern is only one column of pixels; when the metasurface moves to the edge of the overall area in the horizontal direction, a new projection pattern is obtained again by moving one pixel in the vertical direction, and the difference between the projection pattern and the previous projection pattern is only one row of pixels; then the moving direction of the metasurface is adjusted to move in the opposite direction of the previous horizontal movement; and so on, spatial multiplexing is achieved by moving the metasurface, so that more mask patterns can be obtained in a limited spatial area, thereby improving the information capacity of the metasurface mask.

2. The single-pixel imaging method based on a large-capacity spatial multiplexing metasurface according to claim 1, characterized in that: The method further includes step 4, realizing spatial multiplexing according to the mobile metasurface method of step 3, and obtaining a series of mask patterns, and completing image reconstruction by using a single-pixel imaging algorithm through the modulation pattern and the corresponding total detection intensity; in addition, the modulation by the mobile metasurface method significantly accelerates the switching between masks compared with the single-pixel imaging using DMD as a spatial light modulator; the switching frequency of the mask is determined by the movement rate, and by speeding up the movement rate, the switching rate of the mask can be increased, thereby achieving single-pixel imaging with a faster imaging speed.

3. The single-pixel imaging method based on a large-capacity spatial multiplexing metasurface according to claim 2, characterized in that: The single-pixel imaging algorithm modulates the target using different mask patterns, and then performs correlation operations with the corresponding intensity measurement values ​​to reconstruct the target image; random binary coding is a random binary amplitude distribution, and a new mask pattern can be obtained by changing part of the binary coding information; using this property of binary coding, spatial multiplexing can be achieved and spatial capacity can be increased; Compressed sensing is a technology that seeks sparse solutions to underdetermined linear systems and is used to acquire and reconstruct sparse or compressible signals. The compressed sensing algorithm can recover the complete original signal from the acquired undersampled signal. Its main reconstruction method is to recover the original information by solving the norm optimization problem. The model of single pixel imaging is shown in formula (1): S i =∫∫P i (x,y)I(x,y)dxdy(1) Wherein the subscript i=1, 2, ..., M is the number of time-arranged measurements, M represents the total number of measurements, P represents the binary matrix of the mask, I represents the intensity distribution of the target object, S represents the corresponding total measured light intensity; x, y represent the spatial coordinates on the object plane respectively; Formula (1) can restore the image by solving the inverse process; using the compressed sensing algorithm, Formula (1) can be transformed into a form for solving the optimal solution: Where Φ represents the sensing matrix determined by the M-th mask pattern, S represents the M×1 column vector of the intensity signal collected by the detector, I is the N×1 column vector obtained by shaping the target object, and N represents the total number of pixels in the image; represents the L-2 Euclidean norm, Ψ is the regularization term with sparsity constraint on the image, τ represents the regularization parameter used to adjust the relative weight between residual and sparsity, and an appropriate value needs to be selected to achieve optimal reconstruction; There are many compressed sensing algorithms for single-pixel imaging. In order to improve the reconstruction quality of the image, a two-step iterative shrinkage algorithm TWIST is used to reconstruct the image. Each iteration of the two-step iterative shrinkage algorithm TWIST depends on the previous two iterations, rather than just the previous iteration, making the iterative shrinkage process more accurate and faster; the iterative total error compensation ITEC algorithm is used to optimize the image reconstructed by the compressed sensing algorithm. The algorithm compensates for the errors of the binary values ​​in the mask pattern respectively, and iterates a series of mask patterns until a reconstructed image that meets the preset effect is obtained.

4. The single-pixel imaging method based on a large-capacity spatial multiplexing metasurface according to claim 1, 2 or 3, characterized in that: The implementation method of step 1 includes the following steps: Step 1.1: Generate a randomly distributed binary code and decompose the code into a series of mask patterns in sequence, so as to generate more modulation mask patterns in a limited space and improve the spatial capacity; Step 1.2: Use the generated series of mask patterns to amplitude modulate the target object to obtain the corresponding total amplitude intensity after modulation, and use the single-pixel imaging algorithm to perform correlation operations on the mask and total intensity: S i =∫∫P i (x,y)I(x,y)dxdy(1) Wherein the subscript i=1, 2, ..., M is the number of time-arranged measurements, M represents the total number of measurements, P represents the binary matrix of the mask, I represents the intensity distribution of the target object, S represents the corresponding total measured light intensity; x, y represent the spatial coordinates on the object plane respectively; Step 1.3: In order to reduce the number of measurements and improve the reconstruction quality, a compressed sensing algorithm is used. The compressed sensing algorithm reconstructs the object by solving the following optimization problem: Where Φ represents the sensing matrix determined by the M-th mask pattern, S represents the M×1 column vector of the intensity signal collected by the detector, I is the N×1 column vector obtained by shaping the target object, and N represents the total number of pixels in the image; represents the L-2 Euclidean norm, Ψ is the regularization term with sparsity constraint on the image, τ represents the regularization parameter used to adjust the relative weight between residual and sparsity, and an appropriate value needs to be selected to achieve optimal reconstruction; Step 1.4: Use the two-step iterative shrinkage algorithm TWIST to iteratively optimize the target value. Each iteration of this iterative process depends on the previous two iterations, rather than just the previous iteration, making the iterative shrinkage process more accurate and faster. Step 1.5: Select the total variation TV as the regularization term with sparsity constraint: Where dx(I) and dy(I) represent the first-order gradient operations in the horizontal and vertical directions, which can suppress noise and avoid over-smoothing of the image; Step 1.6: Set appropriate parameters, and after iterative processing from steps 1.4 to 1.5, obtain the reconstructed image of the target object; then use the iterative total error compensation ITEC algorithm to optimize the reconstruction quality under random mask modulation: Among them I n (x,y) is the image obtained by n iterations. and Respectively expressed as: Where P r and S r Represent the mask pattern and detection intensity in the rth measurement, respectively, and S0 represents the total intensity obtained by the detector under the parallel light beam; the process is to compensate the errors of the binary values ​​in the mask pattern respectively, and obtain a reconstructed image that meets the preset effect by iterating a series of mask patterns; Step 1.7: After the target image is reconstructed in steps 1.1 to 1.6, the reconstruction quality of the image is detected; the image reconstruction quality is related to the parameters in TWIST and the number of iterations of ITEC, that is, a two-step iterative shrinkage algorithm is used to perform correlation operations on the modulation pattern and the corresponding detection intensity, and the imaging effect of the mask is tested. The relevant parameters are modified according to the reconstruction quality until a reconstructed image that meets the preset effect is obtained.

5. The single-pixel imaging method based on a large-capacity spatial multiplexing metasurface according to claim 4, characterized in that: The implementation method of step 2 is: The metasurface used to achieve large-capacity spatial multiplexing is composed of randomly distributed nanopores. The transmittance of the opaque gold film part and the nanopore part in the metasurface is represented by T1 and T2. If T2 is greater than T1, the greater the difference, the better the binary amplitude modulation of the metasurface is achieved. When the metasurface unit period P is fixed, the radius R of the simulated nanopore and the thickness H of the gold film are scanned; the transmission intensity of the parallel light after modulation by the metasurface unit is calculated from the simulated electromagnetic field data to obtain T1 and T2; the metasurface unit period P should be reasonably selected during simulation, and the film material should be determined to be gold, so as to ensure that the difference between T2 and T1 is as large as possible. The larger the contrast of the metasurface, the better the effect of it as a binary amplitude modulation mask.

Images