Structured light detection system and method based on photon-level single-pixel wavefront imaging

Abstract

The invention provides a structured light detection system and method based on photon-level single-pixel wavefront imaging, and belongs to the technical field of wavefront imaging. The problem of low detection sensitivity of a single-pixel wavefront imaging technology in extreme environment communication is solved; comprising a structured light generation module, a light beam transmission module and a structured light detection module, the structured light generation module is used for generating structured light carrying different topological charges, the structured light is remotely transmitted to the structured light detection module through the light beam transmission module, and the structured light detection module is based on a coaxial phase-shifting interference wavefront imaging principle. A long-distance transmitted structured light signal is processed by adopting a single photon detection technology and utilizing a digital micromirror device and a Hadamard mask modulation technology; according to the method, under 50m long-distance transmission, each mask only needs 30 photons on average to complete topological charge identification, the structured light identification accuracy can reach 96% or above, and the method is suitable for structured light detection under ultra-long-distance and extremely-weak light conditions.

Description

Technical Field

[0001] The present invention provides a structured light detection system and method based on photon-level single-pixel wavefront imaging, belonging to the technical field of wavefront sensing and computational imaging. Background Art

[0002] Structured light is a light field with a special spatial structure distribution constructed by specific optical mode and phase control. In recent years, structured light has shown great application potential in the field of optical communication. By regulating the properties of light waves, structured light can be used to achieve efficient transmission and precise measurement of information. For example, based on the regulation of orbital angular momentum, structured light can achieve multi-channel information transmission, greatly improving communication capacity and security. With the increasing demand for precision measurement and efficient transmission, it is particularly important to accurately detect the effective information carried by structured light. Traditional structured light detection technologies mainly include phase modulation, amplitude modulation and polarization modulation. Phase modulation detection technology uses the phase information of light for encoding. This method requires precise interferometers, and when the environmental conditions are unstable or there is large noise interference, the detection accuracy will be significantly reduced, and it is difficult to maintain consistency and stability. The amplitude modulation method is simple to operate, but due to its low signal-to-noise ratio, it is difficult to achieve high-precision detection under extremely weak light conditions. The polarization modulation method has certain advantages in anti-interference, but it has high requirements on the accuracy and stability of the detection equipment and is easily affected by changes in the external environment. Therefore, in complex transmission environments (extremely weak light, long-distance transmission, underwater turbulence and scattering, etc.), traditional structured light detection technology faces huge challenges. Due to the difficulty in capturing enough signal photons, traditional detection systems have difficulty maintaining high signal-to-noise ratio and high accuracy.

[0003] In recent years, single-pixel wavefront imaging has become an emerging structured light detection technology. It uses a single detector to obtain high-quality wavefront image information through time or space encoding, and can effectively identify the key information carried by structured light. It has the advantages of simple system, high stability and strong anti-interference ability. However, in extremely complex environments, due to the limitations of photodetectors, their detection sensitivity is low and the detection accuracy of structured light is limited. With the help of highly sensitive single-photon detectors, single-photon detection technology can detect very weak light signals, or even the arrival of a single photon. Single-photon detection technology has the advantages of high detection sensitivity and high accuracy, and is suitable for light signal detection in extreme environments, but it cannot identify the effective information carried by structured light. Therefore, it is necessary to combine single-pixel wavefront imaging and single-photon detection technology to develop a structured light detection system based on photon-level single-pixel wavefront imaging to achieve high-sensitivity detection of structured light under extremely weak light conditions. Summary of the invention

[0004] In view of the problem of low detection sensitivity faced by single-pixel wavefront imaging technology in extreme environment communications, the present invention proposes a structured light detection system and method based on photon-level single-pixel wavefront imaging.

[0005] The technical solution adopted by the present invention is: a structured light detection system based on photon-level single-pixel wavefront imaging, including a structured light generation module, a light beam transmission module and a structured light detection module. The structured light generation module is used to generate structured light carrying different topological charges, and the structured light is transmitted to the structured light detection module over a long distance through the light beam transmission module. The structured light detection module is based on the principle of coaxial phase-shifting interference wavefront imaging, and processes the structured light signal transmitted over a long distance by adopting single-photon detection technology, using digital micromirror devices and Hadamard mask modulation technology.

[0006] The structured light generating module comprises a laser, wherein the laser emitted by the laser is collimated and expanded by a beam expander, and then reflected by a reflector, and the reflected light is polarized by a polarizer to become linearly polarized light, and then a half-wave plate is used to adjust the polarization direction of the light beam so that the light beam is consistent with the arrangement direction of liquid crystal molecules in a spatial light modulator, and a first computer is used to load holograms with different topological charges on the spatial light modulator to regulate the phase of the laser, so as to generate structured light carrying different topological charges, and a 4f system composed of a first convex lens, an aperture and a second convex lens is used to select a first-order diffraction light beam.

[0007] The structured light detection module includes a digital micromirror device, a single-photon detector, a time-correlated single-photon counter and a second computer. A telescope is used to focus the structured light wavefront signal onto the digital micromirror device, and a binary grating-Hadamard mask based on a checkerboard reference pattern is loaded on the digital micromirror device to modulate the wavefront signal to achieve structured light coaxial phase-shifting interference; a third convex lens is used to perform Fourier transform on the first-order diffracted light after interference, and the zero-frequency component of the pinhole filter at the spectrum plane is detected by the single-photon detector; the photon count of each mask of the single-photon detector per unit time is recorded by a time-correlated single-photon counter and transmitted to the second computer, the second computer uses a second-order correlation algorithm to reconstruct the structured light wavefront image, improves the imaging quality by a phase denoising neural network, and uses an orbital angular momentum recognition neural network to recognize the topological charge carried in the structured light.

[0008] The aperture r of the pinhole needs to meet the following conditions:

[0009]

[0010] In the above formula: λ is the wavelength of the light beam, f is the focal length of the third convex lens, and d is the diameter of the light beam.

[0011] The structured light generated by the structured light generation module is a Laguerre-Gaussian beam, a Bessel-Gaussian beam or a Hermite-Gaussian beam carrying different topological charges.

[0012] A structured light detection method based on photon-level single-pixel wavefront imaging, using a structured light detection system based on photon-level single-pixel wavefront imaging, includes the following steps:

[0013] Step 1: Generation and emission of structured light wavefront signals;

[0014] Step 2: The structured light wavefront signal is transmitted over long distances in the beam transmission module;

[0015] Step 3: Wavefront signal reception and imaging process:

[0016] Step 3.1, using a telescope to focus the structured light wavefront signal onto a digital micromirror device;

[0017] Step 3.2, loading a binary grating-Hadamard mask based on a checkerboard reference pattern on the digital micromirror device to modulate the wavefront signal to achieve structured light coaxial phase-shifting interference;

[0018] The steps for generating a binary grating-Hadamard mask based on a checkerboard reference pattern loaded on a digital micromirror device are as follows:

[0019] Step 3.2.1, divide the mask into a signal area and a reference area in a checkerboard pattern, set the target wavefront imaging resolution to N×N, use 2×2 checkerboard grids to form a superpixel, in the four-step phase shift method, each checkerboard grid corresponds to 4×4 digital micromirror device pixels, and the total size of the loaded mask corresponds to 8N×8N digital micromirror device pixels;

[0020] Step 3.2.2: Use Hadamard mask to encode the wavefront signal in the signal region. The number of Hadamard masks in the full sampling method is N. 2 Zhang, the total number of masks loaded by the four-step phase shift method is 4N 2 open;

[0021] Step 3.2.3, load a binary grating with a period of 4 pixels in the checkerboard signal area and the reference area respectively. The phase of the binary grating loaded in the corresponding signal area of ​​each Hadamard mask is The reference area remains unchanged, and the phase of the binary grating is shifted by 0, π / 2, π, and 3π / 2 respectively, and the corresponding detection values ​​are

[0022] Step 3.3, using a third convex lens to perform Fourier transform on the first-order diffracted light after interference, and the zero-frequency component filtered by the pinhole at the spectrum plane is detected by a single-photon detector;

[0023] Step 3.4, using a time-correlated single-photon counter to record the photon count of the single-photon detector per unit time of each mask and transmit it to a second computer;

[0024] Step 4: Photon-level single-pixel wavefront imaging process:

[0025] Step 4.1, the complex coefficient C can be calculated from the four-step phase shift n :

[0026]

[0027] Where n = 1, 2, 3…N 2 , represents the number of Hadamard masks; i represents the imaginary unit;

[0028] Step 4.2, reconstruct the wavefront W(x,y) of the structured light through the second-order correlation algorithm:

[0029]

[0030] In the formula, H n (x,y) represents the nth Hadamard mask;

[0031] Step 4.3, calculate the phase of the structured light wavefront from the reconstructed wavefront W(x,y):

[0032]

[0033] Step 5: Image denoising and orbital angular momentum identification process.

[0034] The specific implementation process of step 1 is as follows:

[0035] Step 1.1, the laser emitted by the laser is collimated and expanded by a beam expander, and then reflected by a reflector and polarized by a polarizer into linearly polarized light;

[0036] Step 1.2, using a half-wave plate to adjust the polarization direction of the light beam to be consistent with the arrangement direction of the liquid crystal molecules in the spatial light modulator;

[0037] Step 1.3, using the first computer to load holograms with different topological charges on the spatial light modulator to adjust the phase of the laser and generate structured light with different topological charges;

[0038] Step 1.4, use the first convex lens, the aperture, and the second convex lens to form a 4f system to select the first-order diffraction beam.

[0039] The specific implementation process of step 5 is as follows:

[0040] Step 5.1, use the phase denoising neural network to denoise the wavefront phase image reconstructed based on the second-order correlation algorithm:

[0041]

[0042] Where D is the phase denoising neural network, S is the denoised phase image, θ D are the parameters of the neural network;

[0043] Step 5.2, use the orbital angular momentum identification neural network to identify the topological charge carried in the phase image:

[0044] (l,p)=R(S);

[0045] Where R is the orbital angular momentum identification neural network, S is the denoised phase image, l is the topological charge, and p is the radial mode number.

[0046] The beneficial effects of the present invention compared with the prior art are as follows:

[0047] 1. The present invention modulates the structured light wavefront signal through a binary grating-Hadamard mask based on a checkerboard reference pattern, realizes the coaxial phase-shift interference of the wavefront signal, and reconstructs the wavefront image with high precision in combination with a second-order correlation algorithm, thus realizing photon-level high-sensitivity structured light detection in an extreme scattering environment and 50m long-distance transmission;

[0048] 2. The present invention adopts a phase denoising neural network to effectively remove the noise introduced during the transmission process, significantly improving the image signal-to-noise ratio by more than 5dB;

[0049] 3. The orbital angular momentum recognition neural network based on structured light prior information in the present invention accurately recognizes the topological charge in the phase image. On average, only 30 photons are needed per mask to complete the recognition of the topological charge, and the recognition accuracy can reach more than 96%, which significantly improves the ability to analyze the information carried in the structured light. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The present invention will be further described below in conjunction with the accompanying drawings:

[0051] Figure 1 is a system structure diagram of the present invention;

[0052] Figure 2 It is a flow chart of the binary grating-Hadamard mask construction method of the present invention;

[0053] Figure 3 The wavefront phase imaging result diagram of different topological charges of the present invention;

[0054] Figure 4 This is a wavefront phase imaging result diagram under different scattering conditions of the present invention;

[0055] Figure 5It is a phase denoising neural network structure diagram of the present invention;

[0056] Figure 6 This is a graph showing the phase denoising results of the present invention under different average photon numbers;

[0057] Figure 7 This is a structural diagram of a neural network for orbital angular momentum identification according to the present invention;

[0058] Figure 8 It is a line graph of the topological charge identification accuracy under different average photon numbers of the present invention;

[0059] In the figure: 1 is a structured light generating module, 2 is a light beam transmitting module, 3 is a structured light detecting module, 101 is a laser, 102 is a beam expander, 103 is a reflector, 104 is a polarizer, 105 is a half-wave plate, 106 is a spatial light modulator, 1071 is a first convex lens, 1072 is a second convex lens, 108 is an aperture, 109 is a first computer, 301 is a telescope, 302 is a digital micromirror device, 303 is a third convex lens, 304 is a pinhole, 305 is a single photon detector, 306 is a time-correlated single photon counter, and 307 is a second computer. DETAILED DESCRIPTION

[0060] like Figures 1 to 8 As shown, the present invention provides a structured light detection system based on photon-level single-pixel wavefront imaging, including a structured light generating module 1, a light beam transmission module 2 and a structured light detection module 3, wherein the structured light generating module 1 is used to generate structured light carrying different topological charges, and the structured light is transmitted over a long distance to the structured light detection module 3 through the light beam transmission module 2. The structured light detection module 3 is based on the principle of coaxial phase-shifting interference wavefront imaging, and processes the structured light signal transmitted over a long distance by adopting single-photon detection technology, utilizing digital micromirror devices and Hadamard mask modulation technology.

[0061] Specifically, Figure 1 As shown, the structured light generating module 1 includes a laser 101. The laser emitted by the laser 101 is collimated and expanded by a beam expander 102, and then reflected by a reflector 103. The reflected light is polarized by a polarizer 104 and becomes linearly polarized light. A half-wave plate 105 is used to adjust the polarization direction of the light beam so that it is consistent with the arrangement direction of liquid crystal molecules in a spatial light modulator 106. A first computer 109 is used to load holograms with different topological charges on the spatial light modulator 106 to regulate the phase of the laser and generate structured light with different topological charges. A 4f system composed of a first convex lens 1071, an aperture 108 and a second convex lens 1072 is used to select a first-order diffraction beam.

[0062] The structured light may be a Laguerre-Gaussian beam, a Bessel-Gaussian beam or a Hermite-Gaussian beam carrying different topological charges.

[0063] The structured light detection module 3 includes a digital micromirror device 302, a single photon detector 305, a time-correlated single photon counter 306, and a second computer 307. The telescope 301 is used to focus the structured light wavefront signal to the digital micromirror device 302, and a binary grating-Hadamard mask based on a checkerboard reference pattern is loaded on the digital micromirror device 302 to modulate the wavefront signal to achieve structured light coaxial phase-shift interference; the third convex lens 303 is used to perform Fourier transform on the first-order diffracted light after interference, and the zero-frequency component filtered by the pinhole 304 at the spectrum plane is detected by the single photon detector 305; the time-correlated single photon counter 306 records the photon count of each mask of the single photon detector 305 in unit time and transmits it to the second computer 307, and the second-order correlation algorithm is used to reconstruct the structured light wavefront image. Finally, the imaging quality is improved by the phase denoising neural network, and the orbital angular momentum recognition neural network is used to identify the topological charge carried in the structured light.

[0064] The aperture r of the pinhole 304 needs to meet the following conditions:

[0065]

[0066] In the above formula: λ is the wavelength of the light beam, f is the focal length of the third convex lens, and d is the diameter of the light beam.

[0067] The present invention proposes a structured light detection method based on photon-level single-pixel wavefront imaging, based on the above detection system, comprising the following steps:

[0068] Step 1: Structured light wavefront signal generation and emission process:

[0069] Step 1.1, the laser emitted by the laser 101 is collimated and expanded by the beam expander 102, and then reflected by the reflector 103 and polarized by the polarizer 104 to become linearly polarized light;

[0070] Step 1.2, using the half-wave plate 105 to adjust the polarization direction of the light beam to be consistent with the arrangement direction of the liquid crystal molecules in the spatial light modulator 106;

[0071] Step 1.3, using the first computer 109 to load holograms with different topological charges on the spatial light modulator 106 to adjust the phase of the laser, thereby generating structured light with different topological charges;

[0072] Step 1.4, using the first convex lens 1071, the aperture 108, and the second convex lens 1072 to form a 4f system to select the first-order diffraction beam.

[0073] Step 2: The structured light wavefront signal is transmitted over long distances in the beam transmission module 2.

[0074] Step 3: Wavefront signal reception and imaging process:

[0075] Step 3.1, using the telescope 301 to focus the structured light wavefront signal onto the digital micromirror device 302;

[0076] Step 3.2, loading a binary grating-Hadamard mask based on a checkerboard reference pattern on the digital micromirror device 302 to modulate the wavefront signal to achieve structured light coaxial phase-shifting interference;

[0077] The steps of generating a binary grating-Hadamard mask based on a checkerboard reference pattern loaded on the digital micromirror device 302 are as follows:

[0078] Step 3.2.1, divide the mask into a signal area and a reference area in a checkerboard pattern, set the target wavefront imaging resolution to N×N, use 2×2 checkerboard grids to form a superpixel, in the four-step phase shift method, each checkerboard grid corresponds to 4×4 digital micromirror devices 302 pixels, and the total size of the loaded mask corresponds to 8N×8N digital micromirror devices 302 pixels;

[0079] In this embodiment, the target wavefront imaging resolution is set to 64×64, and a superpixel is formed by using 2×2 checkerboard grids. In the four-step phase shift method, each checkerboard grid corresponds to 4×4 digital micromirror devices 302 pixels, and the total size of the loaded mask is 512×512 digital micromirror devices 302 pixels;

[0080] Step 3.2.2: Use Hadamard mask to encode the wavefront signal in the signal region. The number of Hadamard masks in the full sampling method is N. 2 Zhang, the total number of masks loaded by the four-step phase shift method is 4N 2 open;

[0081] In this embodiment, the number of Hadamard masks in the full sampling method is 4096, and the total number of masks loaded in the four-step phase shift method is 16384;

[0082] Step 3.2.3, load a binary grating with a period of 4 pixels in the checkerboard signal area and the reference area respectively. The phase of the binary grating loaded in the corresponding signal area of ​​each Hadamard mask is The reference area remains unchanged, and the phase of the binary grating is shifted by 0, π / 2, π, and 3π / 2 respectively, and the corresponding detection values ​​are

[0083] Step 3.3, using the third convex lens 303 to perform Fourier transform on the first-order diffracted light after interference, and the zero-frequency component filtered by the pinhole 304 at the spectrum plane is detected by the single-photon detector 305;

[0084] Step 3.4: Use the time-correlated single photon counter 306 to record the photon count of the single photon detector 305 per unit time of each mask and transmit it to the second computer 306.

[0085] Step 4: Photon-level single-pixel wavefront imaging process:

[0086] Step 4.1, the complex coefficient C can be calculated from the four-step phase shift n :

[0087]

[0088] Where n = 1, 2, 3…N 2 , represents the number of Hadamard masks; i represents the imaginary unit;

[0089] Step 4.2, reconstruct the wavefront W(x,y) of the structured light through the second-order correlation algorithm:

[0090]

[0091] In the formula, H n (x,y) represents the nth Hadamard mask;

[0092] Step 4.3, calculate the phase of the structured light wavefront from the reconstructed wavefront W(x,y):

[0093]

[0094] Step 5: Image denoising and orbital angular momentum identification process;

[0095] Step 5.1, use the phase denoising neural network to denoise the wavefront phase image reconstructed based on the second-order correlation algorithm:

[0096]

[0097] Where D is the phase denoising neural network, S is the denoised phase image, θ Dis the parameter of the neural network; the structure of the phase denoising neural network is designed based on the U-Net framework and combined with the Masked Self-Attention mechanism. Masked Self-Attention uses a mask to prevent each position from selecting the input information behind it. The encoder part of the network extracts features from the input phase image through a multi-layer convolutional neural network and gradually reduces the spatial dimension through downsampling operations. A masked self-attention module is added after each layer of convolution to focus on the areas where drastic changes occur in the phase image and identify pixels that may contain noise, thereby improving the denoising effect. The decoder part restores the resolution of the image through upsampling and transmits fine-grained feature information with the corresponding encoding layer through jump connections to ensure that the spatial information of the image is preserved. The network finally outputs the denoised phase image, and its loss function uses the mean square error (MSE). By minimizing the difference between the denoised image and the original image, it ensures that the detailed features of the phase are retained while effectively removing the noise.

[0098] Step 5.2, use the orbital angular momentum identification neural network to identify the topological charge carried in the phase image:

[0099] (l,p)=R(S);

[0100] Where R is the orbital angular momentum recognition neural network, S is the denoised phase image, l is the topological charge, and p is the radial mode number. The structure of the orbital angular momentum recognition neural network is optimized based on the traditional convolutional neural network (CNN), and the design focuses on identifying the topological charge carried by structured light. The network extracts features from the complex ripple structure in the input phase image through multiple layers of deep convolutional layers, especially for periodic changes and topological structures in the image. In order to improve the ability to capture local and global features, the network combines multi-scale convolution operations to ensure that local details and global change trends of the phase image can be extracted simultaneously. In addition, a global average pooling layer is introduced in the network design to reduce the risk of overfitting and improve the robustness of classification. In order to enhance the performance of the network when processing complex topological charges, the loss function adopts a weighted cross entropy loss. By assigning weights to different categories during the training process, the class imbalance problem is effectively balanced and the classification accuracy is improved. Experimental results show that the network can still maintain a topological charge recognition accuracy of more than 96% under low signal intensity conditions with an average of only 30 photons per mask. It has strong noise resistance and robustness, and is particularly suitable for structured light signal analysis and classification in extremely weak light environments.

[0101] The effectiveness of the present invention is verified by experiments below:

[0102] In order to verify the effectiveness of the present invention, the performance of the system under different conditions was tested. First, Laguerre-Gaussian structured light holograms with different topological charges were loaded onto the spatial light modulator 106 in an atmospheric environment, and the orbital angular momentum was set to +1 to +7. Then, a set of binary grating-Hadamard masks was loaded onto the digital micromirror device 302 to modulate the structured light wavefront signal, and the modulated signal was detected by the single photon detector 305, and the wavefront image was reconstructed using a second-order correlation algorithm. Figure 3 Phase reconstruction images of Laguerre-Gaussian beams with different topological charges are shown. It can be observed that the system can accurately reconstruct the wavefront phase information.

[0103] By changing the water turbidity to 0.1NTU, 5NTU, 20NTU, 40NTU, 60NTU, and 90NTU, the transmission of structured light in different scattering underwater environments was simulated. Figure 4 As shown in the figure, when the structured light encounters scattering and obstruction interference during transmission, obvious noise appears in the reconstructed wavefront phase image, and the intensity of the noise increases with the increase of the scattering degree.

[0104] In order to effectively remove these noises, this system proposes a phase denoising neural network. Figure 5 As shown in , the neural network uses the prior information of the structured light phase to suppress noise. Figure 6 As shown, by calculating the peak signal-to-noise ratio, under the condition that the average number of photons per mask is 15, the peak signal-to-noise ratio before denoising is 9.06dB, and the peak signal-to-noise ratio after denoising is 15.84dB, and the image quality is improved by 6.78dB; under the condition that the average number of photons per mask is 30, the peak signal-to-noise ratio before denoising is 14.28dB, and the peak signal-to-noise ratio after denoising is 20.26dB, and the image quality is improved by 6.34dB; under the condition that the average number of photons per mask is 50, the peak signal-to-noise ratio before denoising is 16.38dB, and the peak signal-to-noise ratio after denoising is 22.53dB, and the image quality is improved by 6.15dB; after comprehensive calculation, the image quality is improved by more than 5dB on average after being processed by the phase denoising neural network. The experimental results show that the phase denoising neural network of the present invention significantly reduces the noise interference in the image.

[0105] In order to verify the capability of this system in orbital angular momentum identification, we conducted orbital angular momentum identification experiments under different average photon numbers. Figure 7 As shown in Figure 1, the orbital angular momentum recognition neural network is used to analyze the reconstructed wavefront signal and identify the topological charge carried in the structured light. The experimental results show that the recognition neural network can stably and accurately identify the orbital angular momentum information carried in the structured light under different conditions. Figure 8As shown in Table 1, under the condition of 15 photons per mask on average, the recognition accuracy of orbital angular momentum is 84.3%; under the condition of 30 photons per mask on average, the recognition accuracy of orbital angular momentum exceeds 96%, reaching a qualified level, which is sufficient to meet most practical application requirements; under the conditions of 50 and 200 photons per mask on average, the recognition accuracy of orbital angular momentum is 97.5 and 98.1% respectively, and the recognition accuracy is further improved; when the average number of photons per mask is 1000, the recognition accuracy of orbital angular momentum reaches 99.6%, showing excellent recognition performance. The present invention has high sensitivity and high robustness in orbital angular momentum recognition, which provides an effective solution for parsing and identifying structured light field features in complex environments.

[0106] Average number of photons 15 30 50 200 1000 Correct identification 4215 4807 4877 4905 4981 Identifying Errors 785 193 123 95 19 Recognition accuracy 84.3％ 96.1％ 97.5％ 98.1％ 99.6％

[0107] Table 1 Topological charge identification accuracy of the present invention under different average photon numbers.

[0108] The detection system proposed in the present invention is based on the principle of coaxial phase-shifting interference wavefront imaging. By adopting single-photon detection technology, it significantly improves the detection capability in extremely weak light environments, and uses digital micromirror devices and Hadamard mask modulation technology to effectively process long-distance transmitted structured light signals. Combining the phase denoising neural network and the orbital angular momentum identification neural network, it not only enhances the system's anti-interference ability in complex environments, but also improves the accuracy and stability of structured light detection. The introduction of this system provides a new idea for solving the shortcomings of existing structured light detection technology and has broad application prospects.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A structured light detection system based on photon-level single-pixel wavefront imaging, characterized in that: The invention comprises a structured light generating module (1), a light beam transmission module (2) and a structured light detection module (3); the structured light generating module (1) is used to generate structured light carrying different topological charges; the structured light is transmitted over a long distance to the structured light detection module (3) through the light beam transmission module (2); the structured light detection module (3) is based on the coaxial phase-shifting interference wavefront imaging principle, adopts single-photon detection technology, utilizes a digital micromirror device (302) and Hadamard mask modulation technology, and processes the structured light signal transmitted over a long distance.

2. A structured light detection system based on photon-level single-pixel wavefront imaging according to claim 1, characterized in that: The structured light generating module (1) comprises a laser (101). The laser light emitted by the laser (101) is collimated and expanded by a beam expander (102), and then reflected by a reflector (103). The reflected light is polarized by a polarizer (104) and becomes linearly polarized light. The polarization direction of the light beam is adjusted by a half-wave plate (105) so that the light beam is consistent with the arrangement direction of liquid crystal molecules in a spatial light modulator (106). A first computer (109) is used to load holograms with different topological charges on the spatial light modulator (106) to control the phase of the laser light, so as to generate structured light with different topological charges. A 4f system composed of a first convex lens (1071), an aperture (108) and a second convex lens (1072) is used to select a first-order diffracted light beam.

3. A structured light detection system based on photon-level single-pixel wavefront imaging according to claim 1 or 2, characterized in that: The structured light detection module (3) comprises a digital micromirror device (302), a single-photon detector (305), a time-correlated single-photon counter (306) and a second computer (307); a telescope (301) is used to focus a structured light wavefront signal onto the digital micromirror device (302); a binary grating-Hadamard mask based on a checkerboard reference pattern is loaded on the digital micromirror device (302) to modulate the wavefront signal, thereby realizing structured light coaxial phase-shift interference; and a third convex lens (303) is used to focus the first-order diffracted light after interference. Fourier transform is performed, and the zero-frequency component filtered by the pinhole (304) at the spectrum plane is detected by the single-photon detector (305); the photon count of each mask of the single-photon detector (305) per unit time is recorded by a time-correlated single-photon counter (306) and transmitted to a second computer (307), and the second computer (307) reconstructs the structured light wavefront image using a second-order correlation algorithm, improves the imaging quality through a phase denoising neural network, and uses an orbital angular momentum recognition neural network to recognize the topological charge carried in the structured light.

4. The structured light detection system based on photon-level single-pixel wavefront imaging according to claim 3, characterized in that: The aperture r of the pinhole (304) needs to satisfy the following conditions: In the above formula: λ is the wavelength of the light beam, f is the focal length of the third convex lens, and d is the diameter of the light beam.

5. The structured light detection system based on photon-level single-pixel wavefront imaging according to claim 2, characterized in that: The structured light generated by the structured light generating module (1) is a Laguerre-Gaussian beam, a Bessel-Gaussian beam or a Hermite-Gaussian beam carrying different topological charges.

6. A structured light detection method based on photon-level single-pixel wavefront imaging, characterized in that: The structured light detection system based on photon-level single-pixel wavefront imaging as described in any one of claims 1 to 5 comprises the following steps: Step 1: Generation and emission of structured light wavefront signals; Step 2: The structured light wavefront signal is transmitted over a long distance in the beam transmission module (2); Step 3: Wavefront signal reception and imaging process: Step 3.1, using a telescope (301) to focus the structured light wavefront signal onto a digital micromirror device (302); Step 3.2, loading a binary grating-Hadamard mask based on a checkerboard reference pattern on the digital micromirror device (302) to modulate the wavefront signal to achieve structured light coaxial phase-shifting interference; The steps of generating a binary grating-Hadamard mask based on a checkerboard reference pattern are loaded on the digital micromirror device (302): Step 3.2.1, dividing the mask into a signal area and a reference area in a checkerboard pattern, setting the target wavefront imaging resolution to N×N, using 2×2 checkerboard grids to form a superpixel, in the four-step phase shift method, each checkerboard grid corresponds to 4×4 digital micromirror device (302) pixels, and the total size of the loaded mask corresponds to 8N×8N digital micromirror device (302) pixels; Step 3.2.2: Use Hadamard mask to encode the wavefront signal in the signal region. The number of Hadamard masks in the full sampling method is N. 2 Zhang, the total number of masks loaded by the four-step phase shift method is 4N 2 open; Step 3.2.3, load a binary grating with a period of 4 pixels in the checkerboard signal area and the reference area respectively. The phase of the binary grating loaded in the corresponding signal area of ​​each Hadamard mask is The reference area remains unchanged, and the phase of the binary grating is shifted by 0, π / 2, π, and 3π / 2, respectively. The corresponding detection values ​​are Step 3.3, using the third convex lens (303) to perform Fourier transform on the first-order diffracted light after interference, and the zero-frequency component filtered by the pinhole (304) at the spectrum plane is detected by the single photon detector (305); Step 3.4, using a time-correlated single-photon counter (306) to record the photon count of the single-photon detector (305) per unit time of each mask and transmit it to a second computer (306); Step 4: Photon-level single-pixel wavefront imaging process: Step 4.1, the complex coefficient C can be calculated from the four-step phase shift n : Where n = 1, 2, 3…N 2 , represents the number of Hadamard masks; i represents the imaginary unit; Step 4.2, reconstruct the wavefront W(x,y) of the structured light through the second-order correlation algorithm: In the formula, H n (x,y) represents the nth Hadamard mask; Step 4.3, calculate the phase of the structured light wavefront from the reconstructed wavefront W(x,y): Step 5: Image denoising and orbital angular momentum identification process.

7. The structured light detection method based on photon-level single-pixel wavefront imaging according to claim 6, characterized in that: The specific implementation process of step 1 is as follows: Step 1.1, the laser light emitted by the laser (101) is collimated and expanded by the beam expander (102), and then reflected by the reflector (103) and polarized by the polarizer (104) into linearly polarized light; Step 1.2, using a half-wave plate (105) to adjust the polarization direction of the light beam to be consistent with the arrangement direction of the liquid crystal molecules in the spatial light modulator (106); Step 1.3, using the first computer (109) to load holograms with different topological charges on the spatial light modulator (106) to adjust the phase of the laser and generate structured light with different topological charges; Step 1.4, using the first convex lens (1071), the aperture (108), and the second convex lens (1072) to form a 4f system to select the first-order diffraction beam.

8. The structured light detection method based on photon-level single-pixel wavefront imaging according to claim 6, characterized in that: The specific implementation process of step 5 is as follows: Step 5.1, use the phase denoising neural network to denoise the wavefront phase image reconstructed based on the second-order correlation algorithm: Where D is the phase denoising neural network, S is the denoised phase image, θ D are the parameters of the neural network; Step 5.2, use the orbital angular momentum identification neural network to identify the topological charge carried in the phase image: (l,p)=R(S); Where R is the orbital angular momentum identification neural network, S is the denoised phase image, l is the topological charge, and p is the radial mode number.

Images