Array light source based single-pixel imaging method and device

Abstract

The application discloses a kind of single-pixel imaging method and device based on array light source, including obtaining the illumination light field of array light beam and the back light intensity detection value of the reflection back after array light beam irradiation on target object;Based on back light intensity detection value and illumination light field, reconstruct target image as rough target image;Based on rough target image and back light intensity detection value, reconstruct target image using depth neural network without training, output final target image.The present application realizes high efficiency imaging and long distance target detection based on array light source, while removing image noise, artifact and periodicity by introducing depth neural network, further improves the imaging quality and imaging efficiency of single-pixel imaging.

Description

Technical Field

[0001] This invention relates primarily to the field of target detection imaging technology, and in particular to a single-pixel imaging method and apparatus based on an array light source. Background Technology

[0002] Optical imaging, as a crucial means for humans to perceive the world, plays a vital role in theoretical research, technological development, and daily life. Traditional optical detection imaging, such as cameras, typically uses array detectors made of silicon-based materials, such as CCDs and CMOS, to directly image objects. The image and object satisfy a point-to-point mapping relationship. However, its imaging band is narrow, generally responding to the visible light band, with limited application in the non-visible light band and high cost. Furthermore, its detection energy is dispersed, and its detection sensitivity is low, limiting its application in long-distance target detection. In contrast, single-pixel imaging, as a computational imaging technology, primarily acquires information through single-pixel detectors made of materials such as germanium and silicon. It features a wide response band and has been widely used in non-visible light band imaging, such as infrared imaging, terahertz imaging, and X-ray imaging. It also offers advantages such as concentrated detection energy, high sensitivity, and fast response, making it applicable to fields such as long-range remote sensing and single-pixel laser imaging radar, demonstrating significant practical application value.

[0003] The basic principle of single-pixel imaging technology is that laser light is modulated into a structured light field with a specific spatial intensity distribution by a spatial light modulator, which then illuminates the target. The reflected light intensity from the target is received by a single-pixel detector. After multiple samplings, different reconstruction algorithms are used to obtain an image of the target object. In single-pixel imaging technology, the illumination light field is usually generated by devices such as spatial light modulators, LED arrays, and laser phased arrays. Among them, the refresh frequency of spatial light modulators is up to 22.4 kHz, the refresh frequency of LED arrays is up to MHz (relatively low), while the refresh frequency of laser phased arrays can reach GHz (GHz). In addition, the former two have relatively high power loss and short transmission distance, while laser phased arrays can achieve high power output. Therefore, using a laser phased array as an illumination source can achieve high-efficiency imaging and long-distance target detection. However, the regular arrangement of the array light source will lead to the periodicity of the reconstructed image, which seriously affects the imaging quality. Therefore, there is an urgent need for a single-pixel imaging method that uses an array light source as an illumination source to improve both imaging quality and imaging efficiency. Summary of the Invention

[0004] To address the problems of short detection distance, low imaging efficiency, and poor imaging quality in current single-pixel imaging methods, this invention provides a single-pixel imaging method and apparatus based on an array light source. This invention achieves high-efficiency imaging and long-range target detection based on an array light source, and further improves the imaging quality and efficiency of single-pixel imaging by introducing a deep neural network to remove image noise, artifacts, and periodicity.

[0005] To solve the above problems, the technical solution of the present invention is as follows:

[0006] On one hand, the present invention provides a single-pixel imaging method based on an array light source, comprising:

[0007] Acquire the illumination field of the array beam and the intensity of the reflected light after the array beam illuminates the target object;

[0008] The target image is reconstructed based on the backlight intensity detection value and the illumination light field, serving as a rough target image;

[0009] Based on the rough target image and the backlight intensity detection value, the target image is reconstructed using a deep neural network to output the final target image.

[0010] The array beam is composed of a multi-beam array, and each beam is randomly phase modulated by a phase modulator, which has the characteristics of fast modulation frequency, enabling rapid refresh and high-speed sampling of the illumination field, and effectively improving imaging efficiency.

[0011] Furthermore, the target image is reconstructed using a differential ghost imaging algorithm based on the backlight intensity detection value and the illumination light field, including:

[0012]

[0013] Among them O DGI For a rough target image, P n I represents the illumination field generated by the array beam for each frame. n R represents the detected backlight intensity value corresponding to the illumination field illuminating the object in each frame. n This represents the reference signal detection value, which is the intensity of the illumination light field in each frame that is directly recorded. 、 、 <r>These represent the ensemble averages of the illumination field, the backlight intensity detection value, and the reference signal detection value, respectively.

[0014] Furthermore, based on the coarse target image and the backlight intensity detection value, the target image is reconstructed using a deep neural network that does not require training, including:

[0015] The rough target image and the backlight intensity detection value are input into a deep neural network to obtain the current target image;

[0016] Based on the single-pixel imaging mathematical model, the current target image is convolved with the illumination light field to obtain the estimated value of the backlight intensity;

[0017] The loss function of the deep neural network is constructed based on the backlight intensity detection value and the backlight intensity estimate value. The parameters of the deep neural network are continuously optimized and iterated until the convergence condition is met. The target image output by the deep neural network when the convergence condition is met is the final target image.

[0018] Furthermore, the root mean square error of the backlight intensity detection value and the backlight intensity estimate value is used as the loss function of the deep neural network.

[0019] On one hand, the present invention provides a single-pixel imaging device based on an array light source, comprising:

[0020] The input module is used to acquire the illumination field of the array beam and the intensity of the reflected light after the array beam illuminates the target object.

[0021] A coarse target image reconstruction module is used to reconstruct the target image based on the backlight intensity detection value and the illumination light field, as a coarse target image;

[0022] The optimization module is used to reconstruct the target image based on the rough target image and the backlight intensity detection value using a deep neural network that does not require training, and output the final target image.

[0023] On the other hand, the present invention provides a single-pixel imaging system based on an array light source, comprising:

[0024] An array beam generating device is used to generate an array beam and illuminate a target object with the array beam; the array beam is composed of a multi-channel unit beam array, wherein each channel unit beam is randomly phase modulated by a phase modulator.

[0025] The acquisition unit is used to acquire the illumination field of the array beam and the intensity of the reflected light after the array beam illuminates the target object.

[0026] A coarse target image reconstruction module is used to reconstruct the target image based on the backlight intensity detection value and the illumination light field, as a coarse target image;

[0027] The optimization module is used to reconstruct the target image based on the rough target image and the backlight intensity detection value using a deep neural network, and output the final target image.

[0028] On the other hand, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described single-pixel imaging method based on an array light source.

[0029] On the other hand, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described single-pixel imaging method based on an array light source.

[0030] Compared with the prior art, the technical effects of the present invention are as follows:

[0031] This invention provides a single-pixel imaging method based on an array light source. The array beam is composed of a multi-channel unit beam array, where each unit beam undergoes random phase modulation via a phase modulator. Compared with existing spatial light modulation devices, the imaging system of this invention features a high modulation frequency, enabling rapid refresh and high-speed sampling of the illumination light field, effectively improving imaging efficiency.

[0032] This invention reconstructs the target image based on a coarse target image and backlight intensity detection values ​​using a deep neural network that requires no training, outputting the final target image. Compared with traditional reconstruction algorithms, it can efficiently extract image features and information, effectively remove noise, artifacts, and periodicity, and improve the quality of the reconstructed image.

[0033] In summary, this invention has the advantages of high imaging quality, high efficiency, high transmission power, and long detection range, and can play an important role in the field of long-distance target detection. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0035] Figure 1 A flowchart illustrating a single-pixel imaging method based on an array light source, as provided in one embodiment;

[0036] Figure 2 This is a schematic diagram of a single-pixel imaging system based on an array light source, provided as an embodiment.

[0037] Figure 3 This is a schematic diagram of an array of beams arranged in hexagons in one embodiment;

[0038] Figure 4 This is an illumination light field diagram from one embodiment;

[0039] Figure 5 This is a target image in one embodiment;

[0040] Figure 6 In one embodiment, different algorithms are used to obtain reconstructed target images, where (a) is the target object, which is a binary image (three slits), (a1) is the target object image obtained by the traditional differential ghost imaging algorithm, (a2) is the target object image obtained by the traditional compressed sensing algorithm, and (a3) ​​is the target object image reconstructed by the single-pixel imaging method based on array light source proposed in this embodiment.

[0041] Figure 7 This is a schematic diagram illustrating the principle framework of a single-pixel imaging method based on an array light source in one embodiment. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the spirit of the disclosed content will be clearly explained below with reference to the accompanying drawings and detailed description. Any person skilled in the art, after understanding the embodiments of the present invention, can make changes and modifications based on the techniques taught in the present invention without departing from the spirit and scope of the present invention. The illustrative embodiments and descriptions of the present invention are used to explain the present invention, but are not intended to limit the present invention.

[0043] Reference Figure 1 One embodiment provides a single-pixel imaging method based on an array light source, comprising:

[0044] Acquire the illumination field of the array beam and the intensity of the reflected light after the array beam illuminates the target object.

[0045] The target image is reconstructed based on the backlight intensity detection value and the illumination light field, serving as a rough target image;

[0046] Based on the rough target image and the backlight intensity detection value, the target image is reconstructed using a deep neural network to output the final target image.

[0047] The array beam is composed of a multi-beam array, and each beam is randomly phase modulated by a phase modulator, which has the characteristics of fast modulation frequency, enabling rapid refresh and high-speed sampling of the illumination field, and effectively improving imaging efficiency.

[0048] One embodiment proposes a method for reconstructing a target image based on backlight intensity detection values ​​and illumination light field. Specifically, the target image is reconstructed using a differential ghost imaging algorithm based on backlight intensity detection values ​​and illumination light field, including:

[0049]

[0050] Among them O DGI For a rough target image, P n I represents the illumination field generated by the array beam for each frame. n R represents the detected backlight intensity value corresponding to the illumination field illuminating the object in each frame. n This represents the reference signal detection value, which is the intensity of the illumination light field in each frame that is directly recorded.< / r> 、 、 <r>These represent the ensemble averages of the illumination field, the backlight intensity detection value, and the reference signal detection value, respectively.

[0051] One embodiment proposes reconstructing the target image based on a coarse target image and backlight intensity detection values ​​using a deep neural network that does not require training, including:

[0052] The rough target image and the backlight intensity detection value are input into a deep neural network to obtain the current target image;

[0053] Based on the single-pixel imaging mathematical model, the current target image is convolved with the illumination light field to obtain the estimated value of the backlight intensity;

[0054] The loss function of the deep neural network is constructed based on the backlight intensity detection value and the backlight intensity estimate value. The parameters of the deep neural network are continuously optimized and iterated until the convergence condition is met. The target image output by the deep neural network when the convergence condition is met is the final target image.

[0055] Furthermore, the root mean square error of the backlight intensity detection value and the backlight intensity estimate value is used as the loss function of the deep neural network. The goal is to reduce the loss function, thereby guiding the continuous optimization of the deep neural network parameters. As the number of iterations increases, the output image of the iteratively updated deep neural network is less affected by noise and periodicity, and the quality of the target image output by the deep neural network increases.

[0056] It is understood that the convergence condition in this invention is not limited in its setting method, and those skilled in the art can make reasonable settings based on common knowledge, conventional technical means, or other existing technologies. This is not general; the convergence condition can be set as follows: Among them I n I represents the detected backlight intensity value corresponding to the object after each frame of illumination light field illuminates it. e This represents the estimated backlight intensity after the object is illuminated by the lighting field in each frame, where θ is a set threshold. Alternatively, a maximum number of iterations can be set; the iteration ends when the maximum number of iterations is reached.

[0057] In one embodiment, a single-pixel imaging device based on an array light source is provided, comprising:

[0058] The input module is used to acquire the illumination field of the array beam and the intensity of the reflected light after the array beam illuminates the target object.

[0059] A coarse target image reconstruction module is used to reconstruct the target image based on the backlight intensity detection value and the illumination light field, as a coarse target image;

[0060] The optimization module is used to reconstruct the target image based on the rough target image and the backlight intensity detection value using a deep neural network, and output the final target image.

[0061] The implementation methods of the above modules and the construction of the model can all adopt the methods described in any of the foregoing embodiments, and will not be repeated here.

[0062] One embodiment provides a single-pixel imaging system based on an array light source, comprising:

[0063] An array beam generating device is used to generate an array beam and illuminate a target object with the array beam; the array beam is composed of a multi-channel unit beam array, wherein each channel unit beam is randomly phase modulated by a phase modulator.

[0064] The acquisition unit is used to acquire the illumination field of the array beam and the intensity of the reflected light after the array beam illuminates the target object.

[0065] A coarse target image reconstruction module is used to reconstruct the target image based on the backlight intensity detection value and the illumination light field, as a coarse target image;

[0066] The optimization module is used to reconstruct the target image based on the rough target image and the backlight intensity detection value using a deep neural network, and output the final target image.

[0067] The specific structural type of the array beam generating device is not limited, and those skilled in the art can make a reasonable selection based on common knowledge, conventional technical means, or other existing technologies.

[0068] Without loss of generality, refer to Figure 2 An array beam generating device in one embodiment includes a laser 1, a collimating and beam expanding system 2, a beam splitter 3, a phase modulator 4, and a mask 5. The wavelength of the laser 1 is not limited. For example, in one embodiment, the laser 1 operates in the non-visible light band with a wavelength of 1064 nm. The collimating and beam expanding system 2 consists of collimating and beam expanding lenses, which collimate and expand the light emitted by the laser 1. The beam splitter 3 divides the single laser beam into multiple unit beams; the specific number of beams depends on specific requirements. The phase modulator 4 is a lithium niobate electro-optic phase modulation crystal, used to load a random phase, with a modulation frequency up to GHz. The mask 5 is arranged in an array; the array arrangement of the mask 5 determines the arrangement of the array beams. The specific array arrangement of the mask 5 is not limited and depends on the required array beam. Figure 3 This is a schematic diagram of a hexagonal array of beams in one embodiment. The mask 5 is a hexagonal array containing 19 beams, with a beam aperture of 6mm, a beam spacing of 8mm, and a duty cycle of 0.75. After passing through the mask 5, the multiple unit beams become 19 hexagonal array illumination sources, where the frequency, phase, and polarization of each beam are identical. A random voltage is applied to the phase modulator 4 on the corresponding transmission path of each unit beam via a data acquisition card or circuit control board. Each phase modulator 4 performs random phase modulation on the corresponding unit beam, adjusting the phase of the emitted light from different apertures to ensure that the phases of the emitted light from different apertures are all different. The controller 10 controls each phase modulator 4 to perform random phase modulation on each unit beam, ensuring that the phases of the emitted light from different apertures are all different.

[0069] The acquisition unit includes a projection lens 6, a beam splitter 7, a camera 8, and a single-pixel detector 12. The array beam output from the array beam generating device passes sequentially through the projection lens 6 and the beam splitter 7. After being converged by the projection lens 6, it is split into two paths by the beam splitter 7. One path serves as a signal beam, at which a target object is placed. The signal beam illuminates the target object 11, and the single-pixel detector 12 records the intensity of the reflected light after the array beam illuminates the target object 11. The other path serves as a reference beam, where the illumination field of the array beam is recorded by the CCD camera and converted into an electrical signal by the data acquisition card 13, which is then output to the computer 9 for image reconstruction. The light field recording by the camera 8 and the intensity collection process by the single-pixel detector 12 must be synchronized.

[0070] A single-pixel imaging method based on an array light source, implemented by computer 9, includes:

[0071] The illumination field of the array beam from camera 8 and the intensity of the reflected light after the array beam illuminates the target object are collected by single-pixel detector 12.

[0072] Based on the backlight intensity detection value and the illumination light field, the differential ghost imaging algorithm is used to reconstruct the target image at a low sampling rate, which serves as a coarse target image.

[0073]

[0074] Among them O DGI For a rough target image, P n I represents the illumination field generated by the array beam for each frame. n R represents the detected backlight intensity value corresponding to the illumination field illuminating the object in each frame. n This represents the reference signal detection value, which is the intensity of the illumination light field in each frame that is directly recorded.< / r> 、 、 <r>These represent the ensemble averages of the illumination field, the backlight intensity detection value, and the reference signal detection value, respectively.

[0075] Based on a rough target image and backlight intensity detection values, a deep neural network is used to reconstruct the target image, outputting the final target image, including:

[0076] The rough target image and the backlight intensity detection value are input into a deep neural network to obtain the current target image O. D ;

[0077] Based on the single-pixel imaging mathematical model, the current target image is convolved with the illumination light field to obtain the estimated backlight intensity value I. e = < O D ,P n >;

[0078] The root mean square error (RMSE) of the backlight intensity detection value and the backlight intensity estimate value is used as the loss function of the deep neural network. The goal is to reduce the loss function, guiding the deep neural network parameters to continuously optimize and iterate. As the number of iterations increases, the output image is less affected by noise and periodicity, and the image quality improves. When the convergence condition is met, the target image output by the deep neural network that satisfies the convergence condition is taken as the final target image. The convergence condition is set as follows: I n Ie represents the detected backlight intensity value after each frame of illumination light field illuminates the object, θ represents the estimated backlight intensity value after each frame of illumination light field illuminates the object, and θ is the set threshold.

[0079] In one embodiment, the structure of a single-pixel imaging system based on an array light source is as follows: Figure 2 As shown, the target object 11 is a binary image (three slits), as follows: Figure 5 As shown. The single-pixel detector used is a germanium gain-adjustable detector (Thorlabs PDA-50B2). The camera used is a CCD camera (Allied Vision Stingray F-125). The illumination light field recording process and the backlight intensity detection process must be completely synchronized. Figure 3 This is a schematic diagram of a hexagonal array of beams. Figure 4 This is the illumination light field distribution recorded in this embodiment. For example... Figure 7 As shown in the diagram, the principle framework of the single-pixel imaging method based on array light source used in this embodiment is as follows: First, under low sampling rate conditions, the image is roughly reconstructed using the differential ghost imaging algorithm. Then, the reflected light intensity detection value after the array light beam collected by the single-pixel detector 12 illuminates the target object is loaded together with the image into a deep neural network to obtain an optimized output image. The estimated value of the reflected light intensity is obtained according to the single-pixel imaging mathematical model. Then, the root mean square error between the reflected light intensity value and the estimated value of the reflected light intensity value is calculated as the loss function of the network to guide the optimization of network parameters, reduce errors, and obtain high-quality image output.

[0080] The deep neural network used in this embodiment is the U-net network architecture, consisting of an encoder, skip connections, and a decoder. The encoder contains three downsampling layers that progressively extract image features, while the decoder contains three upsampling layers to recover image information and resolution, connected by two convolutional layers. Skip connections map image information from the input layer to the output layer, effectively mitigating the vanishing gradient problem. Specifically, each downsampling layer includes two 5×5 convolutional kernels with a stride of 1, followed by a max-pooling layer, a batch normalization layer, and a ReLU activation function. Each upsampling layer contains one deconvolutional layer and two convolutional layers with 5×5 kernels and a stride of 1, followed by batch normalization and a ReLU activation function. Furthermore, the process is optimized using the Adam optimizer with parameters set to beta1 = 0.9, beta2 = 0.999, epsilon = 1e-9, and an initial learning rate of 0.05. The final output is a high-quality image with a resolution of 64×64. Image features are extracted layer by layer through multiple downsampling layers, and then image resolution is restored through upsampling layers, with feature information passed through connecting layers in between. Using the single-pixel imaging method based on an array light source provided by this invention, an image of the target object is obtained.

[0081] Figure 6 This document describes the reconstructed target images obtained using different algorithms in one embodiment. (a) represents the target object, which is a binary image (three slits). (a1) shows the target object image obtained using the traditional differential ghost imaging algorithm. (a2) shows the target object image obtained using the traditional compressed sensing algorithm. (a3) ​​shows the target object image reconstructed using the single-pixel imaging method based on an array light source proposed in this embodiment. In this embodiment, the image resolution is 128*128, the number of samples is 256, and the sampling rate is approximately 1.6%. It can be seen that the images reconstructed by the traditional differential ghost imaging algorithm and the compressed sensing algorithm have obvious periodicity and noise, resulting in poor image quality. In contrast, the target object image reconstructed by the single-pixel imaging method based on an array light source proposed in this embodiment has no periodicity or obvious noise, and the image contours and details are clearly visible, exhibiting high image quality. Figure 6 (a) Close. Therefore, the efficient single-pixel imaging method based on array light source and deep neural network proposed in this invention can be effectively applied to the field of single-pixel imaging technology, with good imaging quality and high imaging efficiency.

[0082] On the other hand, the present invention provides a computer device including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the single-pixel imaging method based on an array light source provided in any of the above embodiments. The computer device may be a server. The computer device includes a processor, a memory, a network interface, and a database connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device stores sample data. The network interface of the computer device is used for communication with external terminals via a network connection.

[0083] On the other hand, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the single-pixel imaging method based on an array light source provided in any of the above embodiments.

[0084] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0085] Matters not covered in this invention are common knowledge.

[0086] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0087] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

[0088] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.< / r>

Claims

1. A single-pixel imaging method based on an array light source, characterized in that, include: The illumination field of the array beam and the intensity of the reflected light after the array beam illuminates the target object are obtained. The array beam is composed of a multi-channel unit beam array, and each channel unit beam is randomly phase modulated by a phase modulator. The target image is reconstructed based on the backlight intensity detection value and the illumination light field, serving as a rough target image; Based on a rough target image and backlight intensity detection values, a deep neural network is used to reconstruct the target image, outputting the final target image, including: The rough target image and the backlight intensity detection value are input into the deep neural network to obtain the current target image; Based on the single-pixel imaging mathematical model, the current target image is convolved with the illumination light field to obtain the estimated value of the backlight intensity; The loss function of the deep neural network is constructed based on the backlight intensity detection value and the backlight intensity estimate value. The parameters of the deep neural network are continuously optimized and iterated until the convergence condition is met. The target image output by the deep neural network when the convergence condition is met is the final target image.

2. The single-pixel imaging method based on an array light source according to claim 1, characterized in that, The target image is reconstructed using a differential ghost imaging algorithm based on the backlight intensity detection value and the illumination light field, including: in For a rough target image, This represents the illumination field generated by the array beam for each frame. This represents the detected backlight intensity value after each frame of illumination light field illuminates the object. This represents the reference signal detection value, which is the intensity of the illumination light field in each frame that is directly recorded. , , These represent the ensemble averages of the illumination field, the backlight intensity detection value, and the reference signal detection value, respectively.

3. The single-pixel imaging method based on an array light source according to claim 1, characterized in that, The root mean square error of the backlight intensity detection value and the backlight intensity estimate value is used as the loss function of the deep neural network.

4. The single-pixel imaging method based on an array light source according to claim 3, characterized in that, The convergence condition is: ,in This represents the detected backlight intensity value after each frame of illumination light field illuminates the object. This represents the estimated backlight intensity corresponding to the object after each frame of illumination light field illuminates it. The threshold value is set.

5. The single-pixel imaging method based on an array light source according to any one of claims 1 to 4, characterized in that, The convergence condition is: when the number of iterations reaches the set maximum value.

6. A single-pixel imaging device based on an array light source, characterized in that, include: The input module is used to acquire the illumination field of the array beam and the intensity of the reflected light after the array beam illuminates the target object. The array beam is composed of a multi-channel unit beam array, and each channel unit beam is randomly phase modulated by a phase modulator. A coarse target image reconstruction module is used to reconstruct the target image based on the backlight intensity detection value and the illumination light field, as a coarse target image; The optimization module is used to reconstruct the target image based on the coarse target image and the backlight intensity detection value using a deep neural network that does not require training, and outputs the final target image, including: The rough target image and the backlight intensity detection value are input into the deep neural network to obtain the current target image; Based on the single-pixel imaging mathematical model, the current target image is convolved with the illumination light field to obtain the estimated value of the backlight intensity; The loss function of the deep neural network is constructed based on the backlight intensity detection value and the backlight intensity estimate value. The parameters of the deep neural network are continuously optimized and iterated until the convergence condition is met. The target image output by the deep neural network when the convergence condition is met is the final target image.

7. A single-pixel imaging system based on an array light source, characterized in that, include: An array beam generating device is used to generate an array beam and illuminate a target object with the array beam; the array beam is composed of a multi-channel unit beam array, wherein each channel unit beam is randomly phase modulated by a phase modulator. The acquisition unit is used to acquire the illumination field of the array beam and the intensity of the reflected light after the array beam illuminates the target object. A coarse target image reconstruction module is used to reconstruct the target image based on the backlight intensity detection value and the illumination light field, as a coarse target image; The optimization module is used to reconstruct the target image based on the rough target image and the backlight intensity detection value using a deep neural network, and output the final target image, including: The rough target image and the backlight intensity detection value are input into the deep neural network to obtain the current target image; Based on the single-pixel imaging mathematical model, the current target image is convolved with the illumination light field to obtain the estimated value of the backlight intensity; The loss function of the deep neural network is constructed based on the backlight intensity detection value and the backlight intensity estimate value. The parameters of the deep neural network are continuously optimized and iterated until the convergence condition is met. The target image output by the deep neural network when the convergence condition is met is the final target image.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the single-pixel imaging method based on an array light source as described in any one of claims 1 to 4.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the single-pixel imaging method based on an array light source as described in any one of claims 1 to 4.

Images