Multi-wavelength single-pixel imaging method and system based on graphene detector

Abstract

The invention relates to the technical field of optical imaging, in particular to a multi-wavelength single-pixel imaging method and system based on a graphene detector, and the method comprises the steps: constructing a binary mask pattern; encoding an image by using the mth binary mask pattern to obtain a single-pixel measurement value of image encoding; selecting a training set; encoding the images of the training set to obtain K marked data pairs; a loss function is constructed, a neural network can learn from statistical information of K pairs of marked data pairs to map single-pixel measurement values lower than a set sampling rate to an image in a training set, and an optimized neural network mapping function model and a binary mask pattern are obtained by minimizing the loss function; the imaging system comprises a visible light single-pixel imaging light path and a terahertz single-pixel imaging light path. The method can be applied to single-pixel imaging of visible light and terahertz wave bands, the imaging quality of single-pixel imaging can be improved, and the imaging speed can be increased.

Description

Technical Field

[0001] This invention relates to the field of optical imaging technology, and more specifically, to a multi-wavelength single-pixel imaging method and system based on a graphene detector. Background Technology

[0002] From the visible light to the terahertz band, various detectors are designed and fabricated based on the material physics properties of different wavelength ranges. For the visible light band, mainstream detection technologies are mainly based on silicon-based semiconductor photodetectors, such as complementary metal-oxide-semiconductor (CMOS) and charge-coupled devices (CCD). Although these devices exhibit excellent photoelectric conversion efficiency and quantum efficiency within specific wavelength ranges, their spectral response range is limited by the intrinsic bandgap characteristics of the material, making it difficult to achieve wide-spectrum detection. The terahertz band (0.1-10 THz) shows significant application prospects in biomedical imaging, security detection, and non-destructive testing due to its non-ionizing radiation characteristics and good penetration ability to most non-polar media. Currently, terahertz detection mainly relies on thermal detectors, photon detectors, and electronic detectors, but these detector schemes have some technical bottlenecks: thermal detectors generally exhibit long response times and low sensitivity at room temperature, limiting real-time imaging capabilities; photon detectors and electronic detectors are usually limited to operating frequencies below 1 THz, and high spatial resolution imaging is difficult to achieve in the high-frequency band due to diffraction-limited effects. Graphene, as a zero-bandgap two-dimensional carbon material, possesses ultra-high carrier mobility and unique electrical and optical properties, providing a novel material platform for developing broadband-response, high-speed photodetectors. However, existing technologies still suffer from the following problems: current single-pixel imaging systems struggle to simultaneously achieve single-pixel imaging of visible light and terahertz waves, and exhibit significant limitations in imaging quality and sampling efficiency, failing to meet the application requirements of high spatial resolution imaging. Summary of the Invention

[0003] The purpose of this invention is to overcome the limitations of existing single-pixel imaging technologies in terms of imaging quality and imaging speed in visible light and terahertz waves, and to provide a multi-wavelength single-pixel imaging method and system based on a graphene detector, which can be applied to single-pixel imaging in the visible light and terahertz wave bands, and can improve the imaging quality and imaging speed of single-pixel imaging.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A multi-wavelength single-pixel imaging method based on a graphene detector is provided, comprising the following steps: S1. Construction M A binary mask pattern, in which... P ={ P m |m = 1, 2, 3, ..., M} represents a string consisting of elements 0 and 1. M A binary mask pattern; S2. Using the first m A binary mask pattern P m Encoding a pixel with a resolution of N × N Image O To obtain the image O Encoded single-pixel measurement value; S3. Select an image dataset as the training set for the neural network; where, for a dataset with a number of images... K The training set is defined as { | k = 1, 2, 3, ..., K},in, Indicates the first training set k The pixel resolution is N × N Images; S4. Following the encoding process of step S2, the images in the training set from step S3 are used to obtain the training set. K For the labeled data pair Ω T = {( I k , ) | k = 1, 2, 3, ..., K}; S5. The loss function is constructed using the root mean square error between the output image and the input image of the neural network. The neural network can then... K For the labeled data pair Ω T Learning from statistical information to map single-pixel measurements below the set sampling rate to The optimized neural network mapping function model and the optimized binary mask pattern are obtained by minimizing the loss function.

[0005] The present invention provides a multi-wavelength single-pixel imaging method based on a graphene detector, which can obtain efficient binary mask patterns through optimized learning from neural networks. It can obtain more image information with a small number of mask patterns, thereby meeting the high-quality imaging requirements under low sampling rates. The present invention also has excellent generalization performance and can be applied to single-pixel imaging in the visible light and terahertz bands.

[0006] Further, in step S1:

[0007] In the formula, Indicates the first m A binary mask pattern P m The i Line number j Column elements; Indicates the first m The first binary mask pattern i Line number j The trainable weights of the column, wherein the trainable weights follow a truncated normal distribution with a mean of 0 and a standard deviation of 0.1; N Indicates a row element or a column element; M Represents the number of binary masks; defines the sampling rate SR = M / .

[0008] Further, in step S2:

[0009] In the formula, This represents the convolution operator. I m Indicates the image O Encoded single-pixel measurements; therefore, for using M A binary mask pattern P ={ P m | m = 1, 2, 3, ..., M} to the image O Encode to obtain a length of M Single pixel measurement value I ={ I m | m = 1, 2, 3, ..., M}; In step S4:

[0010] In the formula, I k Indicates the first k Zhang Image quilt M A binary mask pattern P Encoded single-pixel measurements.

[0011] Further, in step S5: the optimized neural network mapping function model and the optimized binary mask pattern are obtained by minimizing the loss function as follows:

[0012] In the formula, This represents a set of weights and bias parameters. The neural network mapping function model defined by parameters, This represents the optimized neural network mapping function model. This represents the optimized binary mask pattern.

[0013] This invention also provides an imaging system applied to the aforementioned multi-wavelength single-pixel imaging method based on a graphene detector, comprising a visible light single-pixel imaging optical path and a terahertz single-pixel imaging optical path, wherein: the visible light single-pixel imaging optical path includes a visible light source, a beam expander, a digital micromirror device, a projection mirror group, a terahertz modulator, an off-axis parabolic mirror group, and a graphene detector arranged sequentially, and further includes a single-pixel imaging device communicatively connected to the graphene detector and the digital micromirror device; wherein, an optimized binary mask pattern is used as input to the digital micromirror device, and an optimized neural network mapping function model is used as invoked by the single-pixel imaging device; in the visible light single-pixel imaging optical path, the visible light source emits a light beam, and the light beam passes through... After being expanded by the beam expander, the beam propagates to the digital micromirror device to form structured visible light. This structured visible light is projected onto the surface of the terahertz modulator by the projection mirror group and then reflected to the off-axis parabolic mirror group. Next, the structured visible light passes through the off-axis parabolic mirror group and the object to be imaged within it, obtaining visible light carrying object information, which is then converged to the graphene detector. The light is then transmitted through the graphene detector to the single-pixel imaging device, where it is used for visible light single-pixel imaging using an optimized neural network mapping function model. The terahertz single-pixel imaging optical path can be used to perform terahertz single-pixel imaging at the single-pixel imaging device using the optimized neural network mapping function model.

[0014] Furthermore, the terahertz single-pixel imaging optical path includes a terahertz source located at the terahertz modulator, as well as the off-axis parabolic mirror group, a graphene detector, and a single-pixel imaging device. In the terahertz single-pixel imaging optical path, the terahertz source emits a light beam, which passes through the terahertz modulator and propagates to the off-axis parabolic mirror group. Then, the light beam passes through the off-axis parabolic mirror group and the object to be imaged located in the off-axis parabolic mirror group to obtain a terahertz beam carrying object information, which is then converged to the graphene detector. The beam is then transmitted through the graphene detector to the single-pixel imaging device, where terahertz single-pixel imaging is performed using an optimized neural network mapping function model.

[0015] Furthermore, the projection lens group includes a first projection lens, a reflector, and a second projection lens arranged in sequence; the off-axis parabolic mirror group includes a first off-axis parabolic mirror and a second off-axis parabolic mirror; the visible light single-pixel imaging optical path also includes a first chopper disposed between the visible light source and the beam expander, and the first chopper is communicatively connected to the single-pixel imaging device.

[0016] Furthermore, the single-pixel imaging device includes a computer, a data acquisition unit, a lock-in amplifier, and a current amplifier connected in sequence. The current amplifier is electrically connected to the graphene detector, and the digital micromirror device is electrically connected to the data acquisition unit. The optimized neural network mapping function model is stored in the computer and can be called.

[0017] Furthermore, the method for fabricating the graphene detector includes the following steps: Transferring the graphene film to the substrate layer; Photoresist was coated onto a substrate sample with a graphene film, and then exposed and developed using maskless patterning. The developed sample was subjected to plasma etching, and then the photoresist was removed to obtain a patterned graphene layer. Photoresist was coated onto a substrate sample with a patterned graphene layer, maskless overlay was performed, and development was carried out to complete the positioning photolithography patterning of the electrode area on the sample. Ti thin film and Au thin film are deposited sequentially in the electrode area patterned by positioning photolithography, and then the thin film outside the electrode area is peeled off to obtain the first electrode and the second electrode; The sample after obtaining the first electrode and the second electrode is annealed to obtain the graphene detector.

[0018] Furthermore, the graphene detector includes a substrate layer and a patterned graphene layer disposed on the substrate layer. The two ends of the patterned graphene layer are coated with a first electrode and a second electrode, both of which are Ti / Au electrodes.

[0019] Compared with the prior art, the beneficial effects of the present invention are: This invention relates to a multi-wavelength single-pixel imaging method and system based on a graphene detector. It can obtain efficient binary mask patterns through optimized learning from a neural network, acquiring more image information with a small number of mask patterns, thus meeting the high-quality imaging requirements under low sampling rates. The imaging method also possesses excellent generalization performance, applicable to single-pixel imaging in both visible and terahertz bands. The imaging system combines a graphene detector with a single-pixel imaging method, significantly improving imaging speed and achieving superior imaging results. Furthermore, the imaging system can simultaneously achieve multi-wavelength single-pixel imaging in both visible and terahertz bands, exhibiting high-speed modulation and efficient imaging capabilities. Attached Figure Description

[0020] Figure 1 This is a flowchart of a multi-wavelength single-pixel imaging method based on a graphene detector according to the present invention. Figure 2 This is a connection diagram of an imaging system according to the present invention, in which the arrows indicate the signal transmission direction; Figure 3 This is a schematic diagram of an imaging system according to the present invention, where the arrows indicate the propagation direction of the light source; Figure 4 This is a flowchart of the preparation method of the graphene detector of the present invention; Figure 5 This is a schematic diagram of the graphene detector of the present invention; Figure 6 The images show the results of single-pixel imaging of 532 nm visible light using a graphene detector at different sampling rates. Figure 7 The images show the results of single-pixel imaging at 2.52 THz terahertz based on graphene detectors at different sampling rates. Figure 8 The switching response characteristics of the graphene detector at different wavelengths are shown in the curves.

[0021] In the attached diagram: 100, visible light source; 200, beam expander; 300, digital micromirror device; 410, first projection lens; 420, reflector; 430, second projection lens; 510, terahertz modulator; 520, first chopper; 530, second chopper; 610, first off-axis parabolic mirror; 620, second off-axis parabolic mirror; 700, graphene detector; 710, substrate; 720, patterned graphene layer; 730, first electrode; 740, second electrode; 800, terahertz source; 910, computer; 920, data acquisition device; 930, lock-in amplifier; 940, current amplifier. Detailed Implementation

[0022] The present invention will be further described below with reference to specific embodiments. The accompanying drawings are for illustrative purposes only, representing schematic diagrams rather than actual physical objects, and should not be construed as limiting the scope of this patent. To better illustrate the embodiments of the present invention, some components in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0023] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present patent. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0024] Example 1 like Figure 1 The figure shown is an embodiment of a multi-wavelength single-pixel imaging method based on a graphene detector according to the present invention, which includes the following steps: S1. Construction M A binary mask pattern, in which... P ={ P m | m = 1, 2, 3, ..., M} represents a string consisting of elements 0 and 1. M A binary mask pattern; wherein:

[0025] In the formula, Indicates the first m A binary mask pattern P m The i Line number j Column elements; Indicates the first m The first binary mask pattern i Line number j The trainable weights of the column follow a truncated normal distribution with a mean of 0 and a standard deviation of 0.1. N Indicates a row element or a column element; M Indicates the number of binary masks; It should be noted that “0” and “1” in the formula are used to simulate the encoding of the object to be imaged by the mask pattern projected by the digital micromirror device 300. “0” and “1” can be used to represent the non-flipped state and the flipped state of the mirror array in the digital micromirror device 300, respectively. Define sampling rate SR = M / It should be noted that when M = N × N At this time, the sampling rate SR = 100%, which results in a high-quality and near-perfect image; S2. Using the first m A binary mask pattern P m Encoding a pixel with a resolution of N × N Image O To obtain the image O Encoded single-pixel measurements; where:

[0026] In the formula, This represents the convolution operator. I m Indicates the image O Encoded single-pixel measurements; therefore, for using M A binary mask pattern P ={ P m | m = 1, 2, 3, ..., M} to the image O Encode to obtain a length of M Single pixel measurement value I ={ I m | m = 1, 2, 3, ..., M}; S3. Select an image dataset as the training set for the neural network; where, for a dataset with a number of images... K The training set is defined as { | k = 1, 2, 3, ..., K},in, Indicates the first training set k The pixel resolution is N × NThe training set can be selected from the publicly available image dataset STL-10, but is not limited to this image dataset. The images in the image dataset are labeled images (Ground Truth), and the labeled images can be considered to be high-quality images. S4. Following the encoding process of step S2, the images in the training set from step S3 are used to obtain the training set. K For the labeled data pair Ω T = {( I k , ) | k = 1, 2, 3, ..., K};in:

[0027] In the formula, I k Indicates the first k Zhang Image quilt M A binary mask pattern P Encoded single-pixel measurement value; S5. The loss function is constructed using the root mean square error between the output image and the input image of the neural network. The neural network can then... K For the labeled data pair Ω T Learning from statistical information to map single-pixel measurements below the set sampling rate to The optimized neural network mapping function model and the optimized binary mask pattern are obtained by minimizing the loss function; where: The optimized neural network mapping function model and the optimized binary mask pattern are obtained by minimizing the loss function as follows:

[0028] In the formula, This represents a set of weights and bias parameters. The neural network mapping function model defined by parameters, This represents the optimized neural network mapping function model. This represents the optimized binary mask pattern; In this embodiment, the sampling rate is set to a low sampling rate, which means fewer masks, such as a sampling rate SR of less than 50%. This allows the neural network to... K For the labeled data pair Ω T Learning from statistical information to map single-pixel measurements at low sampling rates to high-quality images, i.e. During neural network training, measurements obtained using a low sampling rate are mapped to the original image. Ik , x k The model is trained in this way, and the trained model can be used to conduct image reconstruction experiments on the actual measurement values ​​of the object to be imaged at this sampling rate.

[0029] In this embodiment, the neural network uses an Adam optimizer with a learning rate of 0.001 to optimize the weights and bias parameters, completes the setting of the sampling rate, and maps the neural network to the function model. The optimized neural network mapping function model is obtained through training. By inputting multi-wavelength single-pixel measurements from the imaging system into the neural network... I This allows for the reconstruction of images at each wavelength. ,like Figure 6 The results show visible light single-pixel imaging at sampling rates of 1.5625%, 6.25%, 12.5%, 25%, and 50%, as well as... Figure 7 The results are terahertz single-pixel imaging at sampling rates of 1.5625%, 6.25%, 12.5%, 25%, and 50%.

[0030] The present invention provides a multi-wavelength single-pixel imaging method based on a graphene detector, which can obtain efficient binary mask patterns through optimized learning from neural networks. It can obtain more image information with a small number of mask patterns, thereby meeting the high-quality imaging requirements under low sampling rates. The present invention also has excellent generalization performance and can be applied to single-pixel imaging in the visible light and terahertz bands.

[0031] Example 2 like Figures 2 to 3 The image shown is a first embodiment of an imaging system of the present invention, applied to a multi-wavelength single-pixel imaging method based on a graphene detector as described in Embodiment 1. The imaging system includes a visible light single-pixel imaging optical path and a terahertz single-pixel imaging optical path, wherein: The visible light single-pixel imaging optical path includes a visible light source 100, a beam expander 200, a digital micromirror device 300, a projection mirror group, a terahertz modulator 510, an off-axis parabolic mirror group, and a graphene detector 700 arranged sequentially. It also includes a single-pixel imaging device that is communicatively connected to the graphene detector 700 and the digital micromirror device 300. The optimized binary mask pattern is used as input to the digital micromirror device 300, and the optimized neural network mapping function model is used by the single-pixel imaging device. In the visible light single-pixel imaging optical path, the visible light source 100 emits a beam, which is expanded by the beam expander 200 and then propagates to the digital micromirror device 300 to form structured visible light. The structured visible light is projected onto the surface of the terahertz modulator 510 by the projection mirror group and then reflected to the off-axis parabolic mirror group. Then, the structured visible light passes through the off-axis parabolic mirror group and the object to be imaged in the off-axis parabolic mirror group to obtain visible light carrying object information and converges to the graphene detector 700. It is then transmitted to the single-pixel imaging device through the graphene detector 700. Visible light single-pixel imaging is performed at the single-pixel imaging device through the optimized neural network mapping function model. The terahertz single-pixel imaging optical path can be used to perform terahertz single-pixel imaging at a single-pixel imaging device using an optimized neural network mapping function model. In this embodiment, the terahertz single-pixel imaging optical path includes a terahertz source 800 located at the terahertz modulator 510, an off-axis parabolic mirror group, a graphene detector 700, and a single-pixel imaging device. In the terahertz single-pixel imaging optical path, the terahertz source 800 emits a light beam, which passes through the terahertz modulator 510 and propagates to the off-axis parabolic mirror group. Then, the light beam passes through the off-axis parabolic mirror group and the object to be imaged located in the off-axis parabolic mirror group to obtain a terahertz beam carrying object information, which is then converged to the graphene detector 700. The beam is then transmitted to the single-pixel imaging device through the graphene detector 700, where terahertz single-pixel imaging is performed using an optimized neural network mapping function model.

[0032] like Figure 2 and Figure 3 As shown, the projection lens group includes a first projection lens 410, a reflector 420, and a second projection lens 430 arranged sequentially; the off-axis parabolic mirror group includes a first off-axis parabolic mirror 610 and a second off-axis parabolic mirror 620, wherein both the first off-axis parabolic mirror 610 and the second off-axis parabolic mirror 620 are off-axis parabolic mirrors with an off-axis angle of 90°; the visible light single-pixel imaging optical path also includes a first chopper 520 disposed between the visible light source 100 and the beam expander 200, and the terahertz single-pixel imaging optical path also includes a second chopper 530 disposed between the terahertz source 800 and the terahertz modulator 510, and both the first chopper 520 and the second chopper 530 are communicatively connected to the single-pixel imaging device.

[0033] In this embodiment, the terahertz modulator 510 is a light-controlled terahertz modulator. The light-controlled terahertz modulator is opaque to visible light but permeable to terahertz waves. When visible light shines on the light-controlled terahertz modulator, it generates an ultra-thin conductive layer that blocks the transmission of terahertz waves. Therefore, by using a computer 910 to print a series of optimized binary mask patterns... The visible light input to the digital micromirror device 300 is modulated by the digital micromirror device 300 and then irradiates the light-controlled terahertz modulator, enabling spatial encoding of the terahertz wave.

[0034] like Figure 2 and Figure 3 As shown, the single-pixel imaging device includes a computer 910, a data acquisition unit 920, a lock-in amplifier 930, and a current amplifier 940, which are connected in sequence. The current amplifier 940 is electrically connected to the graphene detector 700, and the digital micromirror device 300 is electrically connected to the data acquisition unit 920. The optimized neural network mapping function model is stored in the computer 910 and can be called. In this embodiment, the data acquisition unit 920 is a data acquisition card, and the digital micromirror device 300 is electrically connected to the data acquisition card, which can be used to synchronize trigger signals.

[0035] like Figure 3 As shown, in the visible light single-pixel imaging optical path, the visible light source 100 emits a visible light beam. The emitted visible light beam passes through the first chopper 520, and is then expanded by the beam expander 200 before propagating to the mirror array surface of the digital micromirror device 300 to form structured visible light. The structured visible light passes sequentially through the first projection lens 410, the reflector 420, and the second projection lens 430 before being projected onto the surface of the terahertz modulator 510. It is then reflected to the first off-axis parabolic mirror 610, where it is collimated and illuminates the first object to be imaged. The visible light carrying object information is converged by the second off-axis parabolic mirror 620 to the graphene detector 700. The graphene detector 700 converts the optical signal into an electrical signal, which is read out by the current amplifier 940 and the lock-in amplifier 930. The data acquisition unit 920 acquires the electrical signal and transmits it to the optimized neural network mapping function model in the computer 910. Perform visible light single-pixel imaging, for example Figure 6 As shown.

[0036] like Figure 3 As shown, in the terahertz single-pixel imaging optical path, the terahertz source 800 emits a terahertz beam. The emitted terahertz beam passes through the second chopper 530, then through the terahertz modulator 510 illuminated by visible light, and propagates to the first off-axis parabolic mirror 610. The terahertz beam is collimated by the first off-axis parabolic mirror 610 and illuminates the second object to be imaged. The terahertz beam carrying object information is converged by the second off-axis parabolic mirror 620 to the graphene detector 700. The graphene detector 700 converts the optical signal into an electrical signal, which is read out by the current amplifier 940 and the lock-in amplifier 930. The data acquisition unit 920 acquires the electrical signal and transmits it to the optimized neural network mapping function model in the computer 910. Perform terahertz single-pixel imaging, for example Figure 7 As shown.

[0037] The imaging system of this invention combines a graphene detector with a single-pixel imaging method, which can significantly improve the imaging speed and obtain better imaging results. The imaging system of this invention can simultaneously realize multi-wavelength single-pixel imaging of visible light and terahertz, and has the performance of high-speed modulation and high-efficiency imaging.

[0038] It should be noted that visible light and terahertz light can be modulated using different chopper frequencies, not limited to physical chopper modulation but also electrical modulation. This can be combined with a lock-in amplifier 930 to detect changes in the photocurrent of the graphene detector 700 caused by variations in the intensity of multi-wavelength light, thus achieving photoelectric detection of visible light and terahertz light within the same optical path. For example... Figure 8 As shown, the photovoltage switching response characteristics of the graphene detector 700 under illumination and without illumination are displayed in visible light (wavelength of 532 nm) and terahertz (wavelengths of 118 μm and 883 μm).

[0039] Example 3 This embodiment is a second embodiment of an imaging system. This embodiment is similar to Embodiment 2, except that, as shown in the example... Figures 4 to 5 As shown, the fabrication method of the graphene detector 700 includes the following steps: I. A graphene film is transferred to a substrate 710 using a wet transfer process; wherein the substrate 710 includes a Si wafer and a SiO2 layer covering the surface of the Si wafer; II. Place the substrate sample with graphene film in a spin coater, then drop positive photoresist onto it, and spin coat a uniform photoresist film onto the sample. III. Using a maskless lithography machine, patterned exposure is completed according to preset exposure parameters; IV. Immerse the exposed sample in the developing solution to dissolve the photoresist in the exposed area and complete the development process; V. Place the developed sample in the oxygen atmosphere chamber of the plasma cleaner and perform etching for 5 minutes to remove the graphene film in the exposed area. VI. Immerse the etched sample in acetone solution to remove residual photoresist and obtain patterned graphene layer 720; VII. Place the substrate sample with the patterned graphene layer 720 in a spin coater, then drop positive photoresist onto it, and spin coat a uniform photoresist film onto the sample. VIII. Using a maskless lithography machine, complete the maskless overlay of the electrode area according to the preset exposure parameters; IX. Immerse the exposed sample in the developer to dissolve the photoresist in the electrode area and complete the positioning photolithographic patterning of the electrode area on the sample; X. Using a thermal evaporation deposition process, a 20 nm Ti thin film and a 100 nm Au thin film are sequentially deposited in the electrode area with a positioning photolithography pattern to obtain a Ti / Au thin film; XI Immerse the sample with the deposited Ti / Au film in an acetone solution until the Ti / Au film outside the electrode area is completely peeled off to obtain the first electrode 730 and the second electrode 740. XII. The sample after obtaining the first electrode 730 and the second electrode 740 was subjected to high-temperature annealing in a single-temperature zone tube furnace to remove residual impurities on the sample surface and improve the ohmic contact characteristics between graphene and metal electrodes, thus obtaining the graphene detector 700.

[0040] like Figure 5 As shown, the graphene detector 700 includes a substrate layer 710 and a patterned graphene layer 720 disposed on the substrate layer 710. The two ends of the patterned graphene layer 720 are plated with a first electrode 730 and a second electrode 740. In this embodiment, the first electrode 730 and the second electrode 740 are both Ti / Au electrodes, and the substrate layer 710 is a SiO2 / Si substrate.

[0041] In the specific implementation of the above embodiments, the technical features can be combined in any non-contradictory way. For the sake of brevity, not all possible combinations of the above technical features are described. However, as long as the combination of these technical features is not contradictory, it should be considered to be within the scope of this specification.

[0042] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A multi-wavelength single-pixel imaging method based on a graphene detector, characterized in that, Includes the following steps: S1. Construction M A binary mask pattern, in which... P ={ P m | m = 1, 2, 3, ..., M } represents a string consisting of elements 0 and 1. M A binary mask pattern; S2. Using the first m A binary mask pattern P m Encoding a pixel with a resolution of N × N Image O To obtain the image O Encoded single-pixel measurement value; S3. Select an image dataset as the training set for the neural network; where, for a dataset with a number of images... K The training set is defined as { | k = 1, 2, 3, ..., K },in, Indicates the first training set k The pixel resolution is N × N Images; S4. Following the encoding process of step S2, the images in the training set from step S3 are used to obtain the training set. K For the labeled data pair Ω T = {( I k , ) | k = 1, 2, 3, ..., K }; S5. The loss function is constructed using the root mean square error between the output image and the input image of the neural network. The neural network can then... K For the labeled data pair Ω T Learning from statistical information to map single-pixel measurements below the set sampling rate to The optimized neural network mapping function model and the optimized binary mask pattern are obtained by minimizing the loss function.

2. The multi-wavelength single-pixel imaging method based on a graphene detector according to claim 1, characterized in that, In step S1: In the formula, Indicates the first m A binary mask pattern P m The i Line number j Column elements; Indicates the first m The first binary mask pattern i Line number j The trainable weights of the column; N Indicates a row element or a column element; M Indicates the number of binary masks; Define sampling rate SR = M / .

3. The multi-wavelength single-pixel imaging method based on a graphene detector according to claim 2, characterized in that, In step S2: In the formula, This represents the convolution operator. I m Indicates the image O Encoded single-pixel measurements; therefore, for using M A binary mask pattern P ={ P m | m = 1, 2, 3, ..., M } to the image O Encode to obtain a length of M Single pixel measurement value I ={ I m | m = 1, 2, 3, ..., M }; In step S4: In the formula, I k Indicates the first k Zhang Image quilt M A binary mask pattern P Encoded single-pixel measurements.

4. The multi-wavelength single-pixel imaging method based on a graphene detector according to claim 3, characterized in that, In step S5: The optimized neural network mapping function model and the optimized binary mask pattern are obtained by minimizing the loss function as follows: In the formula, This represents a set of weights and bias parameters. The neural network mapping function model defined by parameters, This represents the optimized neural network mapping function model. This represents the optimized binary mask pattern.

5. An imaging system applied to the multi-wavelength single-pixel imaging method based on a graphene detector as described in any one of claims 1 to 4, characterized in that, This includes visible light single-pixel imaging optical paths and terahertz single-pixel imaging optical paths, among which: The visible light single-pixel imaging optical path includes a visible light source (100), a beam expander (200), a digital micromirror device (300), a projection mirror group, a terahertz modulator (510), an off-axis parabolic mirror group, and a graphene detector (700) arranged in sequence. It also includes a single-pixel imaging device that is communicatively connected to the graphene detector (700) and the digital micromirror device (300). The optimized binary mask pattern is used as input to the digital micromirror device (300), and the optimized neural network mapping function model is used by the single-pixel imaging device. In the visible light single-pixel imaging optical path, the visible light source (100) emits a light beam, which is expanded by the beam expander (200) and then propagates to the digital micromirror device (300) to form structured visible light. The structured visible light is projected onto the surface of the terahertz modulator (510) by the projection mirror group and then reflected to the off-axis parabolic mirror group. Then, the structured visible light passes through the off-axis parabolic mirror group and the object to be imaged in the off-axis parabolic mirror group to obtain visible light carrying object information and converges to the graphene detector (700). It is then transmitted to the single-pixel imaging device through the graphene detector (700), where visible light single-pixel imaging is performed by the optimized neural network mapping function model. The terahertz single-pixel imaging optical path can be used to perform terahertz single-pixel imaging at the single-pixel imaging device using an optimized neural network mapping function model.

6. The imaging system according to claim 5, characterized in that, The terahertz single-pixel imaging optical path includes a terahertz source (800) located at the terahertz modulator (510), and also includes the off-axis parabolic mirror group, the graphene detector (700), and the single-pixel imaging device. In the terahertz single-pixel imaging optical path, the terahertz source (800) emits a light beam, which passes through the terahertz modulator (510) and propagates to the off-axis parabolic mirror group. Then, the light beam passes through the off-axis parabolic mirror group and the object to be imaged located in the off-axis parabolic mirror group to obtain a terahertz beam carrying object information and converges to the graphene detector (700). The beam is then transmitted to the single-pixel imaging device through the graphene detector (700), and terahertz single-pixel imaging is performed at the single-pixel imaging device using an optimized neural network mapping function model.

7. The imaging system according to claim 5, characterized in that, The projection lens group includes a first projection lens (410), a reflector (420), and a second projection lens (430) arranged in sequence; the off-axis parabolic mirror group includes a first off-axis parabolic mirror (610) and a second off-axis parabolic mirror (620); the visible light single-pixel imaging optical path also includes a first chopper (520) disposed between the visible light source (100) and the beam expander (200), and the first chopper (520) is communicatively connected to the single-pixel imaging device.

8. The imaging system according to claim 5, characterized in that, The single-pixel imaging device includes a computer (910), a data acquisition unit (920), a lock-in amplifier (930), and a current amplifier (940) connected in sequence. The current amplifier (940) is electrically connected to the graphene detector (700), and the digital micromirror device (300) is electrically connected to the data acquisition unit (920). The optimized neural network mapping function model is stored in the computer (910) and can be called.

9. The imaging system according to claim 5, characterized in that, The fabrication method of the graphene detector (700) includes the following steps: The graphene film is transferred to the substrate layer (710). Photoresist was coated onto a substrate sample with a graphene film, and then exposed and developed using maskless patterning. The developed sample was subjected to plasma etching, and then the photoresist was removed to obtain a patterned graphene layer (720). Photoresist was coated onto a substrate sample with a patterned graphene layer (720), maskless overlay was performed, and development was carried out to complete the positioning photolithographic patterning of the electrode area on the sample. Ti thin film and Au thin film are deposited sequentially in the electrode area of ​​the positioning photolithography pattern, and then the thin film outside the electrode area is peeled off to obtain the first electrode (730) and the second electrode (740). The sample after obtaining the first electrode (730) and the second electrode (740) is annealed to obtain the graphene detector (700).

10. The imaging system according to claim 5 or 9, characterized in that, The graphene detector (700) includes a substrate (710) and a patterned graphene layer (720) disposed on the substrate (710). The two ends of the patterned graphene layer (720) are coated with a first electrode (730) and a second electrode (740). The first electrode (730) and the second electrode (740) are both Ti / Au electrodes.

Images