An infrared single-pixel computational imaging system and method

By introducing an optical collection and alignment module into the infrared single-pixel imaging system, the problems of light field coupling and photosensitive surface alignment are solved, achieving high signal-to-noise ratio and reliable infrared imaging, and supporting the integration of high-quality two-dimensional heterojunction detectors.

CN123066013APending Publication Date: 2026-11-10XI AN JIAOTONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611393843.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-09
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing infrared single-pixel imaging systems cannot achieve large beam-shrink ratio achromatic optical field coupling and micron-level photosensitive surface alignment when extending to short-wave infrared, resulting in insufficient signal quality and operational reliability, and making it difficult to integrate high-quality two-dimensional heterojunction detectors.

Method used

An optical collection and alignment module is employed, including a first-stage achromatic beam-shrinking mirror group, a long-pass dichroic mirror, and a reflective microscope objective. Combined with a visible light source and a camera, it enables achromatic beam-shrinking of the light field and visible light-assisted observation, ensuring that the light spot is precisely aligned with the photosensitive surface of the two-dimensional heterojunction photodetector.

Benefits of technology

It achieves uniformity of optical field coupling across a wide spectral band and visualized monitoring of micron-scale photosensitive surfaces, significantly improving the signal-to-noise ratio and operational reliability. It enables the effective integration of high-quality two-dimensional heterojunction detectors and enhances the performance of the imaging system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123066013A_ABST
    Figure CN123066013A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of infrared computational imaging technology, specifically relating to an infrared single-pixel computational imaging system and method. The system includes a broadband tunable light source module, a spatial light modulation module, a photoelectric detection module, a data acquisition module, an image reconstruction module, and an optical collection and alignment module, which includes a first-stage achromatic beam-shrinking lens group, a long-pass dichroic mirror, a reflective microscope objective, a visible light source, a beam splitter, and a camera. The beam-shrinking lens group achromaticly shrinks the spatially modulated light field, the dichroic mirror reflects the infrared signal light to the sample and transmits visible light to the camera, and the microscope objective focuses the beam transmitted through the sample onto the photosensitive surface of a two-dimensional heterojunction detector. The alignment position of the light spot is monitored in real time through visible light-assisted observation of the optical path. This invention achieves broadband, high-beam-shrinkage-ratio achromatic light field coupling and real-time visualization monitoring of the micron-level photosensitive surface, effectively improving the signal-to-noise ratio and operational reliability of the infrared single-pixel imaging system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of infrared computational imaging technology, specifically relating to an infrared single-pixel computational imaging system and method. Background Technology

[0002] The short-wave infrared band (1.1μm~2.5μm) has significant research value and application prospects in materials science, semiconductor testing, biomedicine, and cultural relic identification. Compared to visible light, short-wave infrared can reveal the chemical composition and molecular structure information of samples; compared to mid- and far-infrared, it has higher spatial resolution, thus becoming a key window for non-destructive testing and spectroscopic analysis.

[0003] Currently, single-pixel imaging technology is considered an effective way to overcome the limitations of traditional infrared array imaging. A typical existing infrared single-pixel imaging system consists of the following basic components: a broadband tunable light source module (such as a supercontinuum laser combined with a monochromator), a spatial light modulation module (the core component being a digital micromirror device (DMD), an optical collection system (usually composed of ordinary lenses or simple beam-shrinking mirror groups), a photoelectric detection module (often using a single-point detector, such as an InGaAs detector or an emerging two-dimensional material / silicon heterojunction detector), a data acquisition module, and an image reconstruction module (computer). Its working principle can be summarized as follows: the light beam emitted from the light source is expanded and then illuminates the DMD; the DMD modulates and encodes the spatial light intensity of the incident beam; the encoded structured light is then illuminated by the optical collection system onto the sample; the transmitted or reflected light carrying sample information is guided by the collection system and focused onto the photosensitive surface of the single-point detector; the detector converts the light signal into an electrical signal; the data acquisition module sequentially records the light intensity value corresponding to each modulation pattern under the synchronous triggering of DMD pattern switching; finally, the computer reconstructs a two-dimensional image from the measured value vector and the known modulation matrix based on an association algorithm or compressed sensing algorithm. Meanwhile, in order to take advantage of the wide-spectrum response of two-dimensional materials in the existing architecture, some studies have attempted to introduce two-dimensional material / silicon heterojunction detectors into the system.

[0004] However, based on the above working principle and existing system architecture, there is a fundamental drawback when extending to short-wave infrared: if the optimal mechanical stripping method is used to fabricate a high-quality two-dimensional heterojunction detector, its photosensitive surface diameter is only on the micrometer scale (10μm~20μm), while the scale of the emitted light field of the DMD is on the centimeter scale. This makes it impossible to completely collect all the spectral components corresponding to each modulation element, causing phase crosstalk artifacts. As a result, the existing optical collection system cannot achieve a large beam-shrink ratio achromatic light field coupling in a wide spectral range to avoid the focused spot shifting beyond the photosensitive surface size at different wavelengths, nor can it accurately align the light spot with the micrometer-sized photosensitive surface in the invisible infrared band. Operators can only rely on repeated trial and error and estimation, using an infrared power meter or oscilloscope to observe weak signal changes and blindly adjust the displacement stage. This makes it difficult to truly and effectively integrate the optimal high-quality two-dimensional heterojunction detector into an infrared single-pixel imaging system, fundamentally restricting the overall improvement of the system in terms of broadband response capability, signal quality, and operational reliability. Summary of the Invention

[0005] To address the problems existing in the prior art, the purpose of this invention is to provide an infrared single-pixel computational imaging system and method, which realizes real-time visualization monitoring of wide-spectrum, high-beam-contraction-ratio achromatic light field coupling and micron-level photosensitive surface, effectively improving the signal-to-noise ratio and operational reliability of the infrared single-pixel imaging system.

[0006] The technical solution of this invention is: An infrared single-pixel computational imaging system and method includes a broadband tunable light source module, a spatial light modulation module, a photoelectric detection module, a data acquisition module, and an image reconstruction module arranged along the optical path. The spatial light modulation module is used to perform spatial light intensity modulation and encoding on the monochromatic light output by the broadband tunable light source module according to a preset modulation matrix. The photoelectric detection module includes a two-dimensional heterojunction photodetector. An optical collection and alignment module is provided between the spatial light modulation module and the photoelectric detection module; the optical collection and alignment module includes a first-stage achromatic beam shrinking lens group, a long-pass dichroic mirror and a reflective microscope objective arranged sequentially along the optical path, as well as a visible light source, a beam splitter and a camera for auxiliary observation; The first-stage achromatic beam-shrinking mirror group achromaticly shrinks the light field emitted from the spatial light modulation module; the long-pass dichroic mirror reflects the shrunken light field onto the sample surface, while simultaneously transmitting the visible light beam to the camera; the illumination light emitted from the visible light source is introduced into the main optical path through the beam splitter and the long-pass dichroic mirror, and then illuminates the sample surface through the reflective microscope objective. The visible light reflected from the sample surface returns along the original path and is captured by the camera through the beam splitter, used for real-time observation and alignment of the photosensitive surface and the position of the focused spot of the two-dimensional heterojunction photodetector; the reflective microscope objective is also used to collect the light beam transmitted through the sample or reflected by the sample and focus it onto the photosensitive surface of the two-dimensional heterojunction photodetector, converting it into an electrical signal to obtain the final target image through the data acquisition module and the image reconstruction module.

[0007] Preferably, the long-pass dichroic mirror includes a first surface and a second surface. A first optical thin film is disposed on the first surface. The first optical thin film has a reflectivity greater than 95% for the wavelength range of 1.1μm to 2.5μm and a transmittance greater than 90% for the visible light wavelength range of 400 nm to 700 nm. The second surface is disposed with a second optical thin film having a preset splitting ratio of 95:5 for the visible light wavelength range.

[0008] Preferably, the primary achromatic beam-shrinking lens group is composed of a cemented doublet negative lens and a cemented doublet positive lens made of CaF2 material, with a beam-shrinking ratio of 1:5 to 1:20, used to shrink the light field emitted from the spatial light modulation module and reflect it to the sample surface through the long-pass dichroic mirror.

[0009] Preferably, the two-dimensional heterojunction photodetector is a graphene / MoS2 heterojunction structure, with the following stacking order: Si / SiO2 substrate, bottom layer graphene, top layer MoS2; the photosensitive surface diameter of the two-dimensional heterojunction photodetector is 10μm~20μm.

[0010] Preferably, the broadband tunable light source module includes a supercontinuum laser, a laser beam expander system, and a monochromator; the supercontinuum laser is used to emit a supercontinuum laser beam covering the 400 nm to 2.5 μm wavelength range; the laser beam expander system is disposed in the output optical path of the supercontinuum laser and is used to perform spatial filtering and beam expansion collimation on the incident supercontinuum laser beam; the monochromator is disposed in the output optical path of the laser beam expander system and is used to filter the spatially filtered and beam-expanded supercontinuum laser beam into monochromatic light of a preset wavelength.

[0011] An infrared single-pixel computational imaging method includes the following steps: S1. Obtain monochromatic light of a preset wavelength; S2. Spatial light intensity modulation and encoding of monochromatic light are performed according to the preset modulation matrix to obtain the light field carrying the encoded information; S3. The light field is sequentially achromatic-reduced and reflected for focusing; during the focusing process, auxiliary illumination light is introduced, and the auxiliary illumination light is split and reflected into the main light path where the light field is located, illuminating the sample surface. The visible light reflected from the sample surface returns along the original path and is captured by the split beam, so as to adjust and confirm that the light field is correctly focused and aligned to the target position. S4. Acquire the light signal after it has passed through or been reflected by the sample, and convert it into an electrical signal; S5. Acquire the electrical signal in real time and combine it with a preset modulation matrix to obtain the target image; S6. Switch the monochromatic light to multiple preset wavelengths, and repeat steps S3 to S5 to obtain a set of target images at different wavelengths.

[0012] Preferably, before step S3, a reference focal plane calibration and spot alignment step is included: acquiring a real-time image of the target position and determining the reference focal plane position by adjusting its position to a preset focal plane; and emitting visible light of a preset wavelength to the reference focal plane position while activating all optical elements configured in its optical path so that the visible light forms an uncoded complete spot on the reference focal plane, and adjusting the spot to align it with the target position by focusing. Furthermore, in step S4: the number of samplings M is obtained according to the preset imaging resolution and downsampling rate, so as to determine M preset modulation matrices; the incident monochromatic light is spatially modulated and encoded sequentially based on the M preset modulation matrices to obtain M encoded light fields, which are then used to illuminate the sample sequentially; the electrical signal output from the target position is acquired synchronously to obtain a vector y containing M photocurrent measurement values. In step S5: the target image is reconstructed using a compressed sensing algorithm based on the vector y and the modulation matrix sequence.

[0013] Preferably, adjusting the light spot to align it with the target position includes the following steps: By observing the real-time image and spot image of the target location, the target position is adjusted so that the spot is focused and falls into the effective area of ​​the target location after beam contraction. The electrical signal value at the target position is monitored in real time, and the target position is adjusted a second time based on the electrical signal value until the electrical signal value reaches the preset maximum value, thereby completing the alignment of the light spot with the target position.

[0014] Preferably, the modulation matrix sequence is a Hadamard modulation matrix sequence. During each measurement, the current Hadamard matrix and its complementary matrix are loaded sequentially to obtain the corresponding electrical signals, which are then acquired. The difference between the electrical signals corresponding to the current Hadamard matrix and its complementary matrix is ​​used to cancel common-mode noise interference.

[0015] Preferably, the target image is reconstructed using a compressed sensing algorithm, determined according to the following formula: , in, , In the formula, For the target image; It is a sparse transformation matrix; These are the sparse coefficients of the image in the transform domain; For the measurement matrix, , from the modulation matrix sequence The row vectors are determined to correspond to the modulation pattern during each measurement; This refers to the synchronously acquired electrical signal vector; This is the regularization parameter.

[0016] Compared with the prior art, the infrared single-pixel computational imaging system and method of the present invention have the following beneficial effects: This invention utilizes an optical collection and alignment module positioned between the spatial light modulation module and the photoelectric detection module. A single-stage achromatic beam-splitter group achromaticly reduces the large-aperture light field modulated by spatial light, effectively solving the problem of unstable beam placement on a micrometer-scale photosensitive surface caused by chromatic aberration of focused light spots at different wavelengths across a wide spectral range. This ensures consistent light field coupling across the broad spectral range. Simultaneously, leveraging the optical path separation characteristics of a long-pass dichroic mirror (reflecting infrared signal light to the sample and transmitting visible light to the camera), combined with an auxiliary observation optical path consisting of a visible light source, beam splitter, and camera, the visible light on the sample surface can be observed even when the infrared signal light is invisible using a reflective microscope objective. Real-time imaging with light return enables visualized and precise monitoring and alignment of the micron-scale photosensitive surface and infrared focused spot position of the two-dimensional heterojunction photodetector, avoiding signal mismatch and operational uncertainty caused by traditional blind adjustment methods. Ultimately, the sample signal light collected by the reflective microscope objective is precisely focused onto the detector photosensitive surface, significantly improving the system's efficiency in collecting weak infrared signals. This greatly enhances the overall signal-to-noise ratio, broadband imaging consistency, and operational reliability of the infrared single-pixel computational imaging system, enabling the effective integration of the optimal micron-scale photosensitive surface two-dimensional heterojunction detector into the system and fundamentally overcoming the bottleneck that restricts system performance improvement. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the imaging system in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the two-dimensional heterojunction photodetector in an embodiment of the present invention; Figure 3 This is a flowchart of the imaging method in an embodiment of the present invention.

[0018] Explanation of reference numerals in the attached figures: 1-1. Supercontinuum laser; 1-2. Laser beam expander system; 1-3. First reflecting mirror; 1-4. Second reflecting mirror; 1-5. Spatial light modulation module; 1-6. Third reflecting mirror; 1-7. First-order achromatic beam shrinking mirror group; 1-8. Long-pass dichroic mirror; 1-9. Sample; 1-10. Reflective microscope objective; 1-11. Photodetector module; 1-12. Beam splitter; 1-13. Visible light source; 1-14. Camera; 1-15. Two-dimensional heterojunction photodetector; 1-16. Displacement stage; 1-17. Data acquisition module; 1-18. Image reconstruction module; 2-1. Doped silicon substrate; 2-2. Silicon oxide insulating layer; 2-3. Electrode contacts; 2-4. Electrode connection; 2-5. Two-dimensional material A nanosheet; 2-6. Two-dimensional material B nanosheet; 2-7. Heterojunction photosensitive region. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0021] Furthermore, the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are feasible for those skilled in the art. If the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0022] Example 1 See Figure 1 and Figure 2As shown, in order to achieve wide-spectrum, high-beam-reduction-ratio achromatic light field coupling and real-time visualization monitoring of micron-level photosensitive surfaces, and effectively improve the signal-to-noise ratio and operational reliability of infrared single-pixel imaging systems, this embodiment provides an infrared single-pixel computational imaging system and method, including a wide-spectrum tunable light source module, a spatial light modulation module 1-5, a photoelectric detection module 1-11 arranged along the optical path, a data acquisition module 1-17, and an image reconstruction module 1-18. The photoelectric detection module 1-11 includes a two-dimensional heterojunction photodetector 1-15, and also includes an optical collection and alignment module, which is arranged between the spatial light modulation module 1-5 and the photoelectric detection module 1-11, for beam-reduction and focusing of the spatially modulated light field onto the photosensitive surface of the two-dimensional heterojunction photodetector 1-15, and realizing real-time visualization monitoring of the light spot position; wherein, the optical collection and alignment module includes a first-stage achromatic beam-reduction lens group 1-7 and a long beam-reduction lens group 1-15 arranged sequentially along the optical path. The system includes a wavepass dichroic mirror 1-8 and a reflective microscope objective 1-10, as well as a visible light source 1-13, a beam splitter 1-12, and a camera 1-14 for auxiliary observation. A first-stage achromatic beam-shrinking mirror group 1-7 is used to achromaticly shrink the light field emitted from the spatial light modulation module 1-5. A long-pass dichroic mirror 1-8 is used to reflect the shrunken infrared signal beam onto the surface of sample 1-9, while simultaneously transmitting a visible light beam to the camera 1-14. The reflective microscope objective 1-10 is used to collect the beam transmitted through or reflected from sample 1-9 and focus it onto the photosensitive surface of the two-dimensional heterojunction photodetector 1-15. Illumination light emitted from the visible light source 1-13 is introduced into the main optical path via the beam splitter 1-12 and the long-pass dichroic mirror 1-8. Visible light returning from the surface of sample 1-9 is captured by the camera 1-14 via the beam splitter 1-12, used for real-time observation of the photosensitive surface of the two-dimensional heterojunction photodetector 1-15 and the position of the focused spot.

[0023] Following the order of light propagation, each optical element is fixed sequentially on a precision optical platform to ensure the overall mechanical stability and optical path coaxiality of the system.

[0024] See Figure 1As shown, in an embodiment of the present invention, the broadband tunable light source module includes a supercontinuum laser 1-1, a laser beam expander system 1-2, and a monochromator. The laser beam expander system 1-2 includes a spatial light filter and a lens group. The spatial light filter, consisting of an objective lens and a pinhole, is used for spatial filtering and beam expansion / collimation of the incident beam. In this embodiment, the supercontinuum laser 1-1 in the broadband tunable light source module is a laser capable of outputting a continuous spectrum from 400 nm to 2300 nm. The supercontinuum beam output by this laser first enters the spatial light filter, which consists of an objective lens and a pinhole, for spatial filtering to eliminate higher-order modes and stray light in the beam. It is then expanded and collimated into parallel light by the lens group, and finally filtered by the monochromator to provide a power-stable, mode-simple collimated monochromatic laser source. The output wavelength range of the broadband tunable light source module is from 400 nm to 2.5 μm, covering the visible light, near-infrared, and short-wave infrared bands. By controlling the monochromator and filter wheel with software, the imaging wavelength can be arbitrarily selected within this range, thereby achieving multispectral imaging.

[0025] In this embodiment, the supercontinuum laser 1-1 and the laser beam expander system 1-2 can be replaced with a single-mode laser (m). 2 <1.3) Combined with a square diffractive optical element (DOE), the Gaussian beam emitted by the single-mode laser is shaped and homogenized by the DOE to form a square spot with uniform energy distribution, which serves as the incident light field for the spatial light modulation module. Due to the wavelength selectivity of the DOE, when the system switches the imaging wavelength, a DOE matching the target wavelength must be replaced accordingly.

[0026] The collimated monochromatic light emitted from the broadband tunable light source module is sequentially deflected by the first reflector 1-3 and the second reflector 1-4, and incident in a direction perpendicular to the working plane of the digital micromirror device (DMD) in the spatial light modulation module 1-5. This ensures that when the micromirror array of the DMD is in the "on" state, the reflected light propagates strictly along the system's preset principal optical axis. In this embodiment, the spatial light modulation module 1-5 includes a DMD, which is disposed in the output light path of the broadband tunable light source module and is used to spatially modulate the incident light beam. The DMD uses a chip with a pixel size of 10.8 μm (e.g., Texas Instruments DLP4500), with an effective pixel count of 912 × 1140. To ensure imaging quality and control diffraction effects, the wavelength of the beam output from the broadband tunable light source module should be less than one-third of the pixel size of the DMD (i.e., λ < 3.6 μm), thereby ensuring a clear spatial modulation pattern can still be obtained in the short-wave infrared band.

[0027] The reflected light path of the digital micromirror device exits at a 24° deflection angle. The third reflector 1-6 is precisely positioned on this deflection path to redirect the beam direction. After being reflected by the third reflector 1-6, the beam enters the optical collection and alignment module perpendicularly.

[0028] See Figure 1 As shown, in an embodiment of the present invention, in the optical collection and alignment module, the light beam, after being deflected by the third reflecting mirror 1-6, enters perpendicularly into the first-stage achromatic beam-shrinking lens group 1-7. The first-stage achromatic beam-shrinking lens group 1-7 is composed of a cemented doublet negative lens and a cemented doublet positive lens made of CaF2 material, with a beam-shrinking ratio of 1:5 to 1:20. It is used to shrink the light field emitted from the spatial light modulation module 1-5 and reflect it to the surface of the sample 1-9 via a long-pass dichroic mirror 1-8. Specifically, this lens group consists of a cemented doublet negative lens with a focal length of -300mm and a cemented doublet positive lens with a focal length of 30mm, with a distance of 330mm between the two lenses. Both lenses are made of CaF2 material and have a transmittance of over 95% in the 1.1μm~2.5μm infrared band. This lens group shrinks the approximately 20mm×20mm light field emitted from the spatial light modulation module 1-5 to approximately 2mm×2mm, with a beam-shrinking ratio of 1:10. The achromatic design ensures that light of different wavelengths has the same beam reduction ratio, thereby avoiding beam shift caused by chromatic aberration.

[0029] The beam, after being reduced in size, is reflected by long-pass dichroic mirrors 1-8 and then illuminates the surface of sample 1-9. (See also...) Figure 1 As shown, in an embodiment of the present invention, the long-pass dichroic mirror 1-8 includes a first surface and a second surface. A first optical thin film is disposed on the first surface. The first optical thin film has a reflectivity greater than 95% for the 1.1μm~2.5μm wavelength range and a transmittance greater than 90% for the visible light band (400nm~700nm). The second surface is disposed with a second optical thin film having a preset splitting ratio for the visible light band. In this embodiment, the first surface of the long-pass dichroic mirror 1-8 is coated with an optical thin film with high reflectivity (greater than 95%) for the near-infrared to short-wave infrared band (1.1μm~2.5μm) and high transmittance (greater than 90%) for the visible light band (400nm~700nm). The second surface is coated with a thin film having a preset splitting ratio for the visible light band; in this embodiment, a 95:5 splitting ratio (95% transmission, 5% reflection) is used. The function of the long-pass dichroic mirrors 1-8 is to achieve the separation of the common optical path of the infrared signal light and the visible light observation light. That is, the infrared signal light is efficiently reflected to the sample 1-9 direction, while the visible light is transmitted to the camera 1-14 direction, thereby ensuring that the two beams of light work in their respective optical paths without interfering with each other.

[0030] The light beam transmitted through samples 1-9 carries spatial structural information of samples 1-9 and is collected by the reflecting microscope objective 1-10 and focused onto the two-dimensional heterojunction photodetector 1-15 in the photodetector module 1-11. In this embodiment, the reflecting microscope objective 1-10 is a Schwarzschild type reflecting objective with a magnification of 15×, a numerical aperture of 0.4, and a working distance of 25 mm. Under the above parameter conditions, the theoretical Airy disk diameter (estimated based on the diffraction limit) of the reflective microscope objective 1-10 is less than 10 μm, which ensures that the focused spot size is smaller than the photosensitive surface diameter (20 μm) of the two-dimensional heterojunction photodetector 1-15. This allows the light energy to be effectively covered and collected by the photosensitive surface, avoiding the loss of some light energy caused by the focused spot diameter being too large and exceeding the effective area of ​​the photosensitive surface. This fundamentally prevents imaging errors and contrast reduction caused by this. At the same time, the long 25 mm working distance provides sufficient physical space for the clamping and position adjustment of samples 1-9, taking into account both practicality and ease of operation.

[0031] See Figure 2 As shown, a typical two-dimensional heterojunction photodetector 1-15 uses a doped silicon substrate 2-1 as a substrate support, and covers it with a silicon oxide insulating layer 2-2 to achieve electrical isolation. Electrode contacts 2-3 are provided at both ends of the device. The electrical signal is led out to two-dimensional material A nanosheet 2-5 and two-dimensional material B nanosheet 2-6 through electrode connection 2-4. The two are stacked to form a heterojunction photosensitive region 2-7, which is used to absorb incident light and generate photogenerated carriers, thereby realizing the conversion and output of photoelectric signals.

[0032] See Figure 1As shown, in this embodiment of the invention, the two-dimensional heterojunction photodetector 1-15 is a graphene / MoS2 heterojunction structure, with the following stacking order: Si / SiO2 substrate, bottom layer graphene, top layer MoS2; the photosensitive surface diameter of the two-dimensional heterojunction photodetector 1-15 is 10μm~20μm. In this embodiment, the two-dimensional heterojunction photodetector 1-15 adopts a graphene / MoS2 heterojunction structure, with the following stacking order: Si / SiO2 substrate, bottom layer graphene, top layer MoS2. Each two-dimensional material layer is stacked sequentially by mechanical peeling and dry transfer methods, and an encapsulation structure is provided for protection. The photosensitive surface diameter of the two-dimensional heterojunction photodetector 1-15 is 20μm, and the electrode pattern is defined by photolithography. A two-dimensional heterojunction photodetector 1-15 is mounted on a three-axis precision displacement stage 1-16. This stage 1-16 is manually controlled using a standard micrometer head, with a travel of 10 mm × 10 mm in the XY plane, a minimum scale of 10 μm, and a positioning accuracy of 1 μm. The Z-axis (optical axis) uses a differential micrometer head, with a coarse adjustment minimum scale of 10 μm and a fine adjustment minimum scale of 0.5 μm. This high-precision displacement stage 1-16, combined with subsequent alignment methods, ensures that the focused light spot accurately falls into the heterojunction photosensitive region 2-7 formed by the stacking of two-dimensional material A nanosheets 2-5 and two-dimensional material B nanosheets 2-6. This effectively avoids the situation where the light spot deviates from the photosensitive region and illuminates only the non-overlapping areas of two-dimensional material A nanosheets 2-5 or two-dimensional material B nanosheets 2-6. Since non-overlapping areas also generate photocurrent under illumination, this deviation introduces additional noise signals, contaminating the true measurement values ​​and ultimately leading to image distortion in the reconstructed image. The alignment mechanism of this invention physically eliminates this source of noise.

[0033] See Figure 1 As shown, in an embodiment of the present invention, the optical collection and alignment module further includes an auxiliary observation optical path, which includes a visible light source 1-13, a beam splitter 1-12, and a camera 1-14. The visible light source 1-13, an LED white light source, is activated to illuminate the detector plane in alignment mode. The illumination light emitted by the visible light source 1-13 is introduced into the main optical path through the beam splitter 1-12 (50% transmission) and the long-pass dichroic mirror 1-8 (transmission), forming an incident illumination system. The illumination light is then projected onto the surface of the sample 1-9 through a reflective microscope objective 1-10. The visible light reflected from the surface of the sample 1-9 returns along the original path and is captured by the camera 1-14 after passing through the beam splitter 1-12 (50% reflection). The camera 1-14 transmits the acquired bright-field image in real time to the display of the image reconstruction module 1-18 (computer) for real-time observation of the photosensitive surface and the position of the focused spot of the two-dimensional heterojunction photodetector 1-15. In this embodiment, camera 1-14 is a CMOS camera 1-14.

[0034] See Figure 1As shown, in an embodiment of the present invention, the data acquisition module 1-17 includes a transimpedance amplifier and a data acquisition card. The transimpedance amplifier is connected to the output terminal of the two-dimensional heterojunction photodetector 1-15, and is used to amplify and convert the pA to nA level electrical signal output by the two-dimensional heterojunction photodetector 1-15 into a voltage signal. The data acquisition card is connected to the output terminal of the transimpedance amplifier and is provided with an external trigger interface for synchronization with the modulation matrix switching signal of the spatial light modulation module 1-5. The maximum signal acquisition frequency is not less than 20kHz. The acquisition circuit contacts are connected to the pre-fabricated metal electrodes (source and drain) on the two-dimensional heterojunction photodetector 1-15 through gold wires to acquire the electrical signal generated by the two-dimensional heterojunction photodetector 1-15 under illumination conditions. The transimpedance amplifier is connected to the output terminal of the two-dimensional heterojunction photodetector 1-15 and is used to amplify and convert the weak electrical signal (pA to nA level) into a voltage signal. The data acquisition card connects to the output of the transimpedance amplifier for synchronous acquisition of voltage signals. It has an external trigger interface to synchronize with the modulation matrix switching signal of the digital micromirror device (DMVD) to ensure accurate correspondence between the measured values ​​and the modulation matrix. The maximum signal acquisition frequency is above 20kHz to cover the upper limit of the modulation frequency of the DMVD. Image reconstruction modules 1-18 connect to the data acquisition card and are used to reconstruct images of the target object based on the modulation matrix sequence and the acquired electrical signals.

[0035] Example 2 See Figure 3 As shown, this embodiment provides an infrared single-pixel computational imaging method, including the following steps: S1. Turn on the visible light source and adjust the two-dimensional heterojunction photodetector to move it near the focal plane of the reflective microscope objective. The two-dimensional heterojunction image obtained by the camera has the clearest structure, the highest contrast, and the sharpest edges. When the knob is finely adjusted, the image as a whole becomes blurred synchronously and no longer exhibits the "scanning" dynamic of local changes from blurry to clear or from clear to blurry. Record the Z-axis position of the two-dimensional heterojunction image at this time as the reference focal plane position.

[0036] S2. Output visible light with a preset wavelength of 450nm~650nm through a broadband tunable light source module, configure the digital micromirror device to the fully open state (i.e., load a full white pattern) so that the visible light forms a complete light spot on the reference focal plane, and adjust the light spot to align it with the photosensitive surface of the two-dimensional heterojunction photodetector.

[0037] S3. Turn off the visible light source, place the sample, and switch the broadband tunable light source module to a preset imaging wavelength to output monochromatic light to illuminate the sample, focusing it along the optical path onto the photosensitive surface of the two-dimensional heterojunction photodetector. Specifically, turn off the visible light source to ensure no stray light interference. Place the sample, switch the broadband tunable light source module to a preset imaging wavelength, and output pure monochromatic light through a monochromator or filter wheel to illuminate the sample.

[0038] S4. Based on the preset imaging resolution and downsampling rate, the number of sampling times M is obtained. The image reconstruction module generates M preset modulation matrices based on the sparsity of the transform domain. These modulation matrix sequences are then loaded sequentially into the spatial light modulation module for encoding. The incident beam is spatially modulated and encoded to obtain M spatially modulated and encoded light fields. The M spatially modulated and encoded light fields are sequentially compressed by the first-stage achromatic beam-shrinking mirror group and reflected by the long-pass dichroic mirror before illuminating the sample. The light field transmitted through the sample is collected by the reflective microscope objective and focused onto the photosensitive surface of the two-dimensional heterojunction photodetector. Simultaneously, the data acquisition module synchronously acquires the electrical signal output by the two-dimensional heterojunction photodetector to obtain a vector y containing M photocurrent measurement values, which is used for algorithm reconstruction.

[0039] Specifically, the computer generates a set of preset Hadamard modulation pattern sequences. This embodiment employs a differential measurement mode: for each Hadamard matrix Φ, its complementary matrix 1-Φ is simultaneously loaded, and the corresponding measured values ​​are subtracted to obtain the differential measured value, effectively eliminating common-mode noise such as ambient light and detector dark current, thereby improving the signal-to-noise ratio. This sequence is transmitted to the DMD via a data interface, controlling the "on / off" state of each micromirror unit, thereby spatially modulating the incident broadband beam. The modulated structured beam is narrowed to a 2 mm × 2 mm field of view through the aforementioned optical path and illuminates the sample. Its transmitted light is focused onto the photosensitive surface of the two-dimensional heterojunction photodetector by a microscope objective, and the two-dimensional heterojunction photodetector converts the optical signal into an electrical signal. The data acquisition module acquires and performs analog-to-digital conversion on the electrical signal output by the two-dimensional heterojunction photodetector in real time at a frequency synchronized with the DMD pattern refresh, and transmits the converted digital signal to the computer for storage.

[0040] The imaging resolution and downsampling rate are preset by the user. For example, if you want to reconstruct an image with a resolution of 256*192, then 256*192 is the imaging resolution. Theoretically, to achieve this imaging resolution, you should perform 256*192=49152 samplings. However, in order to improve the imaging speed, you usually only use a 20% downsampling rate (i.e., only sample 49152*20%=9830 times). Then, the image with a resolution of 256*192 can be restored through relevant algorithms.

[0041] S5. Using the image reconstruction module, the target image is reconstructed based on the vector y of the M sampled photocurrent measurements and the modulation matrix sequence, using the compressed sensing algorithm. S6. Switch the visible light wavelength output by the broadband tunable light source module to multiple preset wavelengths, and repeat steps S3 to S5 to obtain a set of target images at different wavelengths. Specifically, to obtain multispectral images, repeat steps S3 to S5, switching to different wavelengths each time (e.g., 1200 nm, 1500 nm, 1800 nm, 2100 nm, 2400 nm) to obtain a set of images at different wavelengths. Since the detector has broadband response characteristics in the infrared band, multispectral imaging can be achieved without replacing the detector.

[0042] See Figure 3 As shown, in an embodiment of the present invention, step S2, adjusting the light spot to align it with the photosensitive surface of the two-dimensional heterojunction photodetector, includes the following steps: S2-1, Coarse Alignment Stage: Simultaneously observe the photosensitive surface image and the spot image of the two-dimensional heterojunction photodetector through the camera. Manually adjust the XY axis of the displacement stage to focus the spot into the effective area of ​​the photosensitive surface of the two-dimensional heterojunction photodetector after beam contraction. The judgment criterion is that the spot size is observed to be minimized in the camera image (indicating that it is in the focal plane position) and the edge of the spot is completely within the boundary of the photosensitive surface. This step uses visible light to assist in observing the optical path, and for the first time visualizes the invisible infrared focusing process, quickly eliminating the situation where the spot is completely deviated from the photosensitive surface, laying the foundation for improving the subsequent alignment efficiency.

[0043] S2-2, Fine Alignment Stage: The signal acquisition system of the two-dimensional heterojunction photodetector is activated, and the electrical signal value is monitored in real time. The XY position of the displacement stage is finely adjusted until the electrical signal reaches its maximum value, thus aligning the light spot with the photosensitive surface of the two-dimensional heterojunction photodetector. This stage uses the real-time reading of the electrical signal as an objective criterion, and its physical essence is to achieve optimal coincidence between the center of the light spot and the center of the photosensitive surface. When the center of the light spot and the center of the photosensitive surface are precisely aligned, not only is the signal collection efficiency highest, but more importantly, it ensures that the energy of the entire light spot is confined to the heterojunction photosensitive region (…). Figure 2 Within (2-7), the amount of irradiation on the non-overlapping area was minimized. Figure 2 By eliminating stray light in (2-5 and 2-6), the noise signal generated at the source is minimized, ensuring the signal-to-noise ratio and purity of the measurement signal. This precision alignment strategy has significantly higher alignment accuracy and repeatability than simple visual recognition.

[0044] See Figure 3As shown, in this embodiment of the invention, the modulation matrix sequence in step S4 is a Hadamard modulation matrix sequence. During each measurement, the spatial light modulation module sequentially loads the current Hadamard matrix and its complementary matrix to obtain the corresponding electrical signals, which are then input into the data acquisition module. The data acquisition module uses the difference between the electrical signals corresponding to the current Hadamard matrix and its complementary matrix to cancel common-mode noise interference. At the start of imaging, the image reconstruction module's built-in control software generates a set of preset Hadamard modulation pattern sequences. This embodiment uses a differential measurement mode: for each Hadamard matrix Φ, its complementary matrix 1-Φ is simultaneously loaded, and the corresponding measured values ​​are subtracted to obtain the differential measured value, effectively eliminating common-mode noise such as ambient light and detector dark current, thereby improving the signal-to-noise ratio.

[0045] See Figure 3 As shown, in an embodiment of the present invention, the target image is reconstructed using a compressed sensing algorithm in step S5, and is determined according to the following formula: , in, , In the formula, For the target image; measurement matrix That is, the Hadamard modulation matrix Φ is derived from a preset modulation matrix sequence. The row vectors are determined to correspond to the modulation pattern during each measurement; This refers to the electrical signal vector synchronously acquired by the photodetector. For sparse transformation matrices (such as DCT or wavelets); These are the sparse coefficients of the image in the transform domain; is a regularization parameter used to adjust the trade-off between the sparsity of the reconstructed image and its fit to the measurement data.

[0046] In terms of solution strategy, the image reconstruction module employs the Alternating Directional Multiplier Method (ADMM) to iteratively solve the above optimization problem. ADMM decomposes this complex problem into three sub-problems that are iterated alternately:

[0047] (1) Image update subproblem: While maintaining the current sparse estimate, update the image to better match the measurement data, and use the conjugate gradient method to efficiently solve the linear equation system; (2) Sparse update subproblem: The current reconstructed image is sparsified by using a total variational denoiser as a sparse prior operator to gradually eliminate noise and artifacts during iteration, so that the image edges remain clear while the overall image tends to be smooth. (3) Multiplier update subproblem: Update the Lagrange multipliers to reconcile the differences between the first two steps and ensure that the iterative process converges.

[0048] Building upon the aforementioned framework, the image reconstruction module further introduces a plug-and-play mechanism, allowing for flexible replacement with high-performance denoisers such as BM3D or deep learning networks in the sparse update step. This enables adaptive optimization of reconstruction quality under different noise levels and imaging conditions, without requiring re-derived algorithm formulas, thus exhibiting good scalability and robustness. The algorithm can reconstruct images with a resolution of 128×128 pixels.

[0049] See Figure 3 As shown, in an embodiment of the present invention, in the multispectral imaging of step S6, if a multispectral image needs to be acquired, steps S3 to S5 are repeated, switching different wavelengths (e.g., 1200 nm, 1500 nm, 1800 nm, 2100 nm, 2400 nm) each time to obtain a set of images at different wavelengths. Since the two-dimensional heterojunction photodetector has a broadband response characteristic in the infrared band, multispectral imaging can be achieved without replacing the detector.

[0050] Example 3 This embodiment provides a range of optimized key parameters for the optical collection and alignment module to adapt to different application requirements: Beam-shrinking lens group design: The preferred beam-shrinking ratio for the first-order achromatic beam-shrinking lens group is between 1:5 and 1:20. A larger beam-shrinking ratio results in a more compact optical path, but also increases the difficulty of aberration correction. The diameter of the light field after beam shrinking must be smaller than the entrance pupil diameter of the reflecting microscope objective to ensure the beam fills the effective aperture of the off-axis parabolic mirror.

[0051] The cutoff wavelength of a dichroic mirror: In near-infrared applications, the start (cutoff) wavelength of a long-pass dichroic mirror can be adjusted according to the required wavelength range, including 650nm, 805nm, 950nm, 1000nm, etc.

[0052] Compared with the prior art, the present invention has the following beneficial effects: (i) Broadband multispectral imaging capability This invention constructs a single-pixel imaging system covering the visible, near-infrared, and short-wave infrared bands by integrating a broadband tunable light source module (400nm~2.5μm) with a two-dimensional heterojunction photodetector. The two-dimensional heterojunction photodetector (graphene / MoS2) has intrinsic broadband response characteristics for the above-mentioned bands, and can perform imaging tasks at any preset wavelength without replacing the detector. This achieves efficient integrated multispectral infrared imaging and overcomes the technical defects of traditional infrared detectors, such as single-band operation and cumbersome replacement.

[0053] (ii) High signal-to-noise ratio and noise suppression The present invention adopts a Hadamard basis differential measurement mode, in which a complementary matrix is loaded simultaneously for each modulation matrix, and the corresponding measured values of the two are subtracted to obtain a differential signal. This solution can effectively cancel common-mode interferences such as ambient light fluctuation, detector dark current drift and circuit thermal noise, and significantly improve the signal-to-noise ratio of weak optical signal detection. Combined with the high responsivity of the two-dimensional heterojunction detector under photocurrents at the pA to nA level, the present system can still maintain excellent image quality in low luminous flux imaging scenarios.

[0054] (3) Broadband achromatic beam shrinking and precise spot control In the present invention, the optical collection and alignment module adopts a first-stage achromatic beam shrinking lens group made of CaF2 (composed of a negative lens with a focal length of -300mm and a positive lens of 30mm), which has the advantages of low dispersion and higher transmittance. Its transmittance can be higher than 95% in the range of 200nm to 8μm, and the beam shrinking magnification is consistent at different wavelengths. This design effectively eliminates spot offset and scale change caused by chromatic aberration in applications in the short-wave infrared band (1.1μm~2.5μm), ensures that no repeated adjustment of optical path alignment is required during multi-wavelength switching, and improves the stability and repeatability of the imaging process.

[0055] (4) High-precision real-time alignment monitoring function By introducing an epifluorescence illumination auxiliary observation optical path (LED visible light source, dichroic mirror, beam splitter and camera), the present invention realizes real-time visual monitoring of the positions of the photosensitive surface of the detector and the focused spot in the visible light mode. With the cooperation of a three-axis precision displacement stage (X-Y plane positioning accuracy 1μm, Z-axis fine adjustment accuracy 0.5μm), focal plane positioning and spot alignment can be completed quickly before imaging, which greatly reduces the alignment difficulty caused by invisible light in the infrared band, and improves the system operation convenience and imaging reliability.

[0056] (5) High-quality image reconstruction at sub-sampling rate Combined with the compressed sensing algorithm framework, under the condition that the number of measurements is less than the number of target pixels (M<N), the present invention realizes high-quality image reconstruction with a resolution of 128×128 pixels through sparse constraint optimization (regularization in the wavelet domain or total variation domain). Compared with the point-by-point scanning imaging method, the present system greatly reduces the required number of measurements and acquisition time, and lowers the burden of data storage and transmission; at the same time, it supports multiple solving algorithms such as convex optimization, iterative thresholding, total variation and deep learning, and can flexibly select the optimal reconstruction strategy according to the actual application scenario, with good scalability.

[0057] (6) Compact system structure and high integration level This invention achieves miniaturized integration of the system while ensuring optical performance through a reflective optical path folding design and a rational layout of the DMD spatial light modulator and beam-shrinking lens group. The various functional modules (light source, modulation, collection, detection, acquisition, and reconstruction) work together to form a complete closed-loop imaging link from optical signal input to digital image output, possessing the potential for engineering applications and portable development.

[0058] (vii) Key parameters are adjustable and highly adaptable The beam-shrinking ratio of the beam-shrinking lens group of the present invention can be preferably adjusted within the range of 1:5 to 1:20. The cutoff wavelength of the dichroic mirror can be selected at nodes such as 650nm, 805nm, 950nm, and 1000nm according to the near-infrared application requirements. The key parameters of the system have flexible configuration capabilities and can be adapted to imaging scenarios with different fields of view, different bands, and different resolution requirements, and have good versatility and scalability.

[0059] In summary, this invention solves the matching problem between the spot size and the micrometer-scale photosensitive surface in a wide-spectrum range by using a single-stage achromatic beam-shrinking lens group, ensuring efficient coupling of light field energy. Real-time monitoring of the spot position is achieved through an auxiliary observation optical path consisting of a long-pass dichroic mirror, a beam splitter, a camera, and a visible light source. Furthermore, a two-step alignment method (coarse and fine) not only ensures spatial overlap between the spot and the photosensitive surface but also actively confines the spot within the effective working area of ​​the heterojunction. The synergistic effect of these three technical features fundamentally overcomes the two major technical defects of traditional systems: signal loss due to spot mismatch and measurement noise introduced by the light response in non-overlapping regions. This significantly improves the imaging accuracy and reliability of the system.

[0060] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An infrared single-pixel computational imaging system, comprising a broadband tunable light source module, a spatial light modulation module, a photoelectric detection module, and a data acquisition module and an image reconstruction module arranged along an optical path, wherein the spatial light modulation module is used to spatially modulate and encode the monochromatic light output by the broadband tunable light source module according to a preset modulation matrix, and the photoelectric detection module comprises a two-dimensional heterojunction photodetector; characterized in that: An optical collection and alignment module is provided between the spatial light modulation module and the photoelectric detection module; the optical collection and alignment module includes a first-stage achromatic beam shrinking lens group, a long-pass dichroic mirror and a reflective microscope objective arranged sequentially along the optical path, as well as a visible light source, a beam splitter and a camera for auxiliary observation; The first-stage achromatic beam-shrinking mirror group achromaticly shrinks the light field emitted from the spatial light modulation module; the long-pass dichroic mirror reflects the shrunken light field onto the sample surface, while simultaneously transmitting the visible light beam to the camera; the illumination light emitted from the visible light source is introduced into the main optical path through the beam splitter and the long-pass dichroic mirror, and then illuminates the sample surface through the reflective microscope objective. The visible light reflected from the sample surface returns along the original path and is captured by the camera through the beam splitter, used for real-time observation and alignment of the photosensitive surface and the position of the focused spot of the two-dimensional heterojunction photodetector; the reflective microscope objective is also used to collect the light beam transmitted through the sample or reflected by the sample and focus it onto the photosensitive surface of the two-dimensional heterojunction photodetector, converting it into an electrical signal to obtain the final target image through the data acquisition module and the image reconstruction module.

2. The infrared single-pixel computational imaging system according to claim 1, characterized in that, The long-pass dichroic mirror includes a first surface and a second surface. A first optical thin film is disposed on the first surface. The first optical thin film has a reflectivity greater than 95% for the wavelength range of 1.1μm to 2.5μm and a transmittance greater than 90% for the visible light wavelength range of 400 nm to 700 nm. A second optical thin film with a preset splitting ratio of 95:5 for the visible light wavelength range is disposed on the second surface.

3. The infrared single-pixel computational imaging system according to claim 1, characterized in that, The primary achromatic beam-shrinking lens group consists of a cemented doublet negative lens and a cemented doublet positive lens made of CaF2 material, with a beam-shrinking ratio of 1:5 to 1:

20. It is used to shrink the light field emitted from the spatial light modulation module and reflect it to the sample surface through the long-pass dichroic mirror.

4. The infrared single-pixel computational imaging system according to claim 1, characterized in that, The two-dimensional heterojunction photodetector has a graphene / MoS2 heterojunction structure, and its stacking order is: Si / SiO2 substrate, bottom layer graphene, top layer MoS2; the photosensitive surface diameter of the two-dimensional heterojunction photodetector is 10μm~20μm.

5. The infrared single-pixel computational imaging system according to claim 1, characterized in that, The broadband tunable light source module includes a supercontinuum laser, a laser beam expander system, and a monochromator. The supercontinuum laser emits a supercontinuum laser beam covering a wavelength range of 400 nm to 2.5 μm. The laser beam expander system is positioned in the output optical path of the supercontinuum laser and is used to perform spatial filtering and beam expansion collimation on the incident supercontinuum laser beam. The monochromator is positioned in the output optical path of the laser beam expander system and is used to filter the spatially filtered and beam-expanded supercontinuum laser beam into monochromatic light of a preset wavelength.

6. An infrared single-pixel computational imaging method, characterized in that, Includes the following steps: S1. Obtain monochromatic light of a preset wavelength; S2. Spatial light intensity modulation and encoding of monochromatic light are performed according to the preset modulation matrix to obtain the light field carrying the encoded information; S3. The light field is sequentially achromaticized, beam-shrinked, and reflected for focusing; During the focusing process, auxiliary illumination light is introduced and then split and reflected into the main optical path where the light field is located, illuminating the sample surface. The visible light reflected from the sample surface returns along the original path and is captured by the split beam to adjust and confirm that the light field is correctly focused and aligned to the target position. S4. Acquire the light signal after it has passed through or been reflected by the sample, and convert it into an electrical signal; S5. Acquire the electrical signal in real time and combine it with a preset modulation matrix to obtain the target image; S6. Switch the monochromatic light to multiple preset wavelengths, and repeat steps S3 to S5 to obtain a set of target images at different wavelengths.

7. The infrared single-pixel computational imaging method according to claim 6, characterized in that: Before step S3, a reference focal plane calibration and spot alignment step is also included: acquiring a real-time image of the target position and adjusting its position to a preset focal plane to determine the reference focal plane position; and emitting visible light of a preset wavelength to the reference focal plane position while activating all optical elements in its optical path so that the visible light forms an uncoded complete spot on the reference focal plane, and adjusting the spot to align it with the target position by focusing. Furthermore, in step S4: the number of samplings M is obtained according to the preset imaging resolution and downsampling rate, so as to determine M preset modulation matrices; the incident monochromatic light is spatially modulated and encoded sequentially based on the M preset modulation matrices to obtain M encoded light fields, which are then used to illuminate the sample sequentially; the electrical signal output from the target position is acquired synchronously to obtain a vector y containing M photocurrent measurement values. In step S5: the target image is reconstructed using a compressed sensing algorithm based on the vector y and the modulation matrix sequence.

8. The infrared single-pixel computational imaging method according to claim 7, characterized in that, Adjusting the light spot to align it with the target position includes the following steps: By observing the real-time image and spot image of the target location, the target position is adjusted so that the spot is focused and falls into the effective area of ​​the target location after beam contraction. The electrical signal value at the target position is monitored in real time, and the target position is adjusted a second time based on the electrical signal value until the electrical signal value reaches the preset maximum value, thereby completing the alignment of the light spot with the target position.

9. The infrared single-pixel computational imaging method according to claim 6, characterized in that, The modulation matrix sequence is a Hadamard modulation matrix sequence. During each measurement, the current Hadamard matrix and its complementary matrix are loaded sequentially to obtain the corresponding electrical signals, which are then acquired. The difference between the electrical signals corresponding to the current Hadamard matrix and its complementary matrix is ​​used to cancel common-mode noise interference.

10. The infrared single-pixel computational imaging method according to claim 7, characterized in that, The target image is reconstructed using a compressed sensing algorithm, and the result is determined according to the following formula: , in, , In the formula, For the target image; It is a sparse transformation matrix; These are the sparse coefficients of the image in the transform domain; For the measurement matrix, , from the modulation matrix sequence The row vectors are determined to correspond to the modulation pattern during each measurement; This refers to the synchronously acquired electrical signal vector; This is the regularization parameter.