Edge-lens-array-based transparent tomographic imaging method and sensing device

Abstract

The present application relates to the field of optical sensing technology, and provides a transparent tomographic imaging method based on an edge convex lens array, which receives an incident light source at a first end face to generate a first light signal, and the first light signal is transmitted along the first end face to a circumferential side face. The first light signal is encoded by a convex lens array structure arranged at the circumferential side face, and at least one group of second light signals is obtained after encoding. The second light signal is sensed by at least one optical detection module. The optical detection module transmits the second light signal to a signal reconstruction module for tomographic reconstruction to obtain a third light signal. The present application also discloses a sensing device. The method and the sensing device can transfer the sampling and encoding process of the light signal from a plane to an edge region of an imaging plane, realize angle-related optical encoding of the in-plane propagating light signal, and obtain an imaging result through tomographic reconstruction. Thus, the technical problems of limited transparency and limited resolution in the existing transparent imaging scheme are overcome.

Description

Technical Field

[0001] This invention relates to the field of optical sensing technology, and in particular to a transparent tomographic imaging method and sensing device based on an edge convex lens array. Background Technology

[0002] Current optical imaging systems generally employ charge-coupled devices (CCDs) or complementary metal-oxide-semiconductor (CMOS) as core pixelated photodetectors to achieve direct sampling and photoelectric conversion of light signals. CCD detectors, with their advantages of high sensitivity, low noise, and high consistency, still hold a significant position in high-end imaging. CMOS detectors, on the other hand, leverage their high integration, low power consumption, fast response speed, and controllable cost to rapidly gain popularity in mass-market, miniaturized imaging scenarios such as consumer electronics and industrial inspection. The two complement each other, covering the vast majority of optical imaging application needs. The core working principle of pixelated photodetectors is to spatially discretely sample the continuous light signal transmitted by the optical system through arrayed pixel units, simultaneously converting the photon signal into a measurable and processable electrical signal. Subsequent signal amplification and analog-to-digital conversion then output a digital image signal, providing the foundation for subsequent image enhancement, feature extraction, and intelligent analysis. They are widely used in many key fields such as aerospace, biomedicine, industrial inspection, consumer electronics, and security monitoring.

[0003] CN121410969A discloses an adaptive tomographic grating dispersion correction method and system based on deep learning, comprising the following steps: receiving an incident light beam, diffracting the incident light beam through a tomographic grating to obtain a multi-level tomographic diffracted beam; guiding the multi-level diffracted beam to a planar straight-fringe blazed grating, and compensating for the dispersion angle of the multi-level diffracted beam through the planar straight-fringe blazed grating to obtain a corrected beam; collimating the corrected beam using a binocular lens group and forming an image-side telecentric optical path to obtain an image-side telecentric beam; performing spectral beam splitting and aberration pre-correction on the image-side telecentric beam through a microlens array, followed by image acquisition to obtain a pre-corrected multispectral image; and performing real-time fusion processing on the pre-corrected multispectral image based on a deep learning algorithm, dynamically adjusting the deflection angle of the microlens array, and outputting a dispersion-corrected tomographic microscopic image.

[0004] CN115685538A discloses a tomographic grating dispersion correction system and method. The system includes a microscope objective, a tomographic grating, a planar straight-fringe blazed grating, a tube lens, and a CCD camera. The tomographic grating is placed at the aperture stop of the microscope objective. The planar straight-fringe blazed grating is placed behind the tomographic grating. The tube lens is placed behind the planar straight-fringe blazed grating. The CCD camera is placed behind the tube lens. The system ensures that the light beams incident on the planar straight-fringe blazed grating do not overlap and that the tube lens, blazed grating, and CCD camera form an image-side telecentric optical path.

[0005] Existing imaging units need to absorb or convert light signals during operation, which inevitably obstructs or cuts off the optical path, making it difficult to use as a transparent imaging layer in the optical path. This makes them unsuitable for applications such as cascading within the optical path, multi-layer superposition imaging, or in-situ observation. Summary of the Invention

[0006] Long-term practice has revealed that existing technologies that use pixelated photodetectors such as charge-coupled devices (CCDs) or complementary metal-oxide-semiconductors (CMOSs) to directly sample light signals inevitably cause the imaging units to absorb or convert the light signals during operation, resulting in occlusion or truncation of the optical path. This makes it difficult to use as a transparent imaging layer in the optical path and unsuitable for applications such as cascaded imaging, multi-layer superposition imaging, or in-situ observation.

[0007] In view of this, the present invention provides a transparent tomography method based on an edge-convex lens array, comprising, Step S1: The incident light source is received from the first end face, and a first light signal is generated. The first light signal is transmitted along the first end face to the circumferential side. Step S2: The first optical signal is encoded by a convex lens array structure arranged on the circumferential side, and at least one set of second optical signals is obtained after encoding. The second optical signals are sensed by at least one optical detection module. In step S3, the optical detection module transmits the second optical signal to the signal reconstruction module for tomographic reconstruction to obtain the third optical signal.

[0008] Preferably, the signal reconstruction module establishes a mapping relationship between the second optical signal and the third optical signal.

[0009] Preferably, the mapping relationship is obtained through supervised learning training.

[0010] Preferably, the mapping relationship is trained through supervised learning, in which the second optical signal is used as the input of the neural network and the third optical signal is used as the output of the neural network; the first optical signal is used as the label for training, the loss function is the error between the third optical signal and the first optical signal, and the training is performed by a convolutional neural network backpropagation algorithm or linear inverse calculation. When the error or the number of training rounds reaches a preset value, the neural network is output as the mapping relationship.

[0011] Preferably, in step S1, after receiving the first optical signal from the first end face, the optical signal is transmitted to the circumferential side face along the parallel direction of the first end face by means of scattering or photoluminescence.

[0012] Preferably, in step S1, after the first optical signal passes through the first end face, the first optical signal of proportion a continues to pass through the second end face along the optical path by adjusting the setting.

[0013] The present invention also discloses a sensing device for performing the above-described transparent tomography method based on an edge convex lens array, the sensing device comprising a first end face, a second end face opposite to the first end face, at least one optical detection module, and a signal reconstruction module; The first end face and the second end face are connected by a circumferential side surface; A convex lens array structure is formed on the circumferential side surface; The optical detection module is fixedly mounted outside the convex lens array structure and can be connected to the signal reconstruction module.

[0014] Preferably, the first end face and the second end face are connected by the circumferential side to form a column structure, and multiple column structures can be fixedly arranged in parallel. The column structure includes a prism structure or a cylindrical structure.

[0015] Preferably, the prism structure includes a regular polyhedral prism, and the optical detection module is disposed on at least one of the circumferential side surfaces of the regular polyhedral prism.

[0016] Preferably, the optical detection module includes a fiber optic array or an image sensor.

[0017] This invention provides a transparent tomographic imaging method based on an edge-convex lens array. Steps S1-S3 involve receiving a first optical signal from a first end face, which then transmits to the circumferential side. The first optical signal is encoded by a convex lens array structure arranged on the circumferential side, resulting in at least one set of second optical signals. At least one optical detection module senses these second optical signals. The optical detection module transmits the second optical signals to a signal reconstruction module for tomographic reconstruction, yielding a third optical signal. This invention also discloses a sensing device. The transparent tomographic imaging method and sensing device based on an edge-convex lens array can transfer the sampling and encoding process of the optical signal from the imaging plane to the edge region of the imaging plane, achieving angle-dependent optical encoding of the in-plane propagating optical signal, and combining this with the tomographic reconstruction method to obtain the imaging result. This eliminates the need for a pixelated detection structure in the imaging area, utilizes a transparent imaging plane to avoid substantial obstruction of the optical path, and ensures that the imaging performance is determined by the edge encoding structure of the convex lens array and the optical detection module, avoiding the discrete sampling limitations imposed by pinholes or fiber optic arrays. This overcomes the technical problems of limited transparency, limited resolution, or complex structure in existing transparent imaging schemes. It can be used in conjunction with various tomographic reconstruction or data-driven methods and is suitable for large field of view, multi-layer stacking, or in-optical-path imaging applications. Attached Figure Description

[0018] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a schematic diagram of the structure of a sensing device according to one embodiment of the present invention. Figure 2 This is a schematic diagram of the optical path for mapping and reconstructing the encoded optical signal to the original imaging optical signal in one embodiment of the present invention. Figure 3 This is a three-dimensional schematic diagram of a sensing device according to one embodiment of the present invention.

[0019] Explanation of reference numerals in the attached figures Detailed Implementation

[0020] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] To address the challenges of existing imaging units that inevitably obstruct or interrupt the optical path during operation due to light signal absorption or conversion, making them unsuitable as transparent imaging layers and for applications such as cascaded imaging, multi-layer stacking, or in-situ observation, this invention proposes a transparent tomographic imaging method based on an edge-convex lens array. Figure 1-3 As shown, the transparent tomography method based on an edge-convex lens array includes: Step S1: The first end face 1 receives the incident light source and generates a first light signal. After passing through the first end face 1, the first light signal is transmitted to the circumferential side face 3. Step S2: The first optical signal is encoded by the convex lens array structure 4 arranged on the circumferential side 3, and at least one set of second optical signals is obtained after encoding. The second optical signals are sensed by at least one optical detection module 5. In step S3, the optical detection module 5 transmits the second optical signal to the signal reconstruction module 6 for tomographic reconstruction to obtain the third optical signal.

[0022] This invention provides a transparent tomographic imaging method based on an edge-convex lens array. This method shifts the sampling and encoding process of optical signals from the imaging plane to the edge region of the imaging plane, achieving angle-related optical encoding of the in-plane propagating optical signals, and combining this with tomographic reconstruction methods to obtain the imaging results. This invention eliminates the need for pixelated detection structures within the imaging region, and by utilizing a transparent imaging plane, it avoids substantial obstruction of the optical path. Its imaging performance is jointly determined by the edge encoding structure of the convex lens array and the optical detection module, avoiding the discrete sampling limitations caused by pinholes or fiber optic arrays. This effectively solves the technical problems of insufficient transparency, limited resolution, or complex structures in existing transparent imaging schemes.

[0023] To establish a clear mapping relationship between the second and third optical signals, the encoded detection signal needs to be quickly converted into the original imaging optical signal, which can significantly improve the accuracy and real-time performance of signal reconstruction. In a more preferred embodiment of the invention, the signal reconstruction module 6 establishes a mapping relationship between the second and third optical signals. The signal reconstruction module 6 pre-stores or calibrates the encoding parameters, optical path transmission matrix, and system response function of the convex lens array structure 4 in real time. Based on the light intensity distribution, angle information, and position information of the second optical signal collected by the optical detection module 5, and combined with a preset tomographic reconstruction algorithm, a forward mapping and inverse solution relationship between the second optical signal and the third optical signal to be reconstructed is established. Through iterative optimization, back-projection reconstruction, or model fitting, the third optical signal is inverted from the second optical signal, realizing the mapping and reconstruction from the encoded optical signal to the original imaging optical signal. In practical optimization applications, if the incident light source is a point light source, such as Figure 1As shown. The first end face 1 receives the first optical signal and, after changing the optical path of the first optical signal, it can be emitted from the convex lens array structure 4 on the circumferential side 3. The first end face 1 is made of a transparent material and can couple the incident light source into the transparent plane through scattering or photoluminescence. To ensure high transmittance of the optical signal, so that the first optical signal passes through the first end face 1 without significant attenuation or stray light interference, ensuring the integrity of the optical signal transmission, the material includes polymethyl methacrylate (PMMA) or a photoluminescent transparent material prepared by adding a fluorescent agent to a transparent PMMA substrate. The photoluminescent transparent material can absorb the energy of the first optical signal and release a wavelength-adapted fluorescent signal, indirectly enhancing the intensity of the optical signal, making the intensity of the second optical signal easier for the optical detection module 5 to sense, and reducing the sensitivity requirements of the detection module. Experiments conducted in this invention show that even if the incident light source consists of multiple points or patterns, the first optical signal can still be resolved using the above method, ensuring the effectiveness of imaging in multi-light source and complex pattern scenes.

[0024] To improve the transmittance of the first end face 1 by applying an anti-reflection coating, and to reduce the loss of the optical signal during transmission, the first optical signal can maintain signal integrity and achieve effective transmission over a longer distance after passing through the first end face 1. For example, after passing through the first end face 1, part of the first optical signal can continue through the second end face 2 and enter the photosensitive end face of another parallel sensing device, or enter other media, ensuring the integrity of the optical signal and the consistency of the optical path. In a more preferred embodiment of the invention, in step S1, after the first optical signal passes through the first end face 1, a proportion 'a' of the first optical signal is adjusted to continue passing through the second end face 2 along the optical path. In practical applications, the proportion 'a' is controlled by adjusting the scattering roughness or photoluminescence doping concentration of the first end face 1 and the second end face 2. For example, in practical optical sensor applications, in order to acquire only the information of the optical signal without affecting the transmission of the optical path and the integrity of the image, the transmittance can be further improved by polishing and applying an anti-reflection coating to the first end face 1 and the second end face 2, minimizing the reflection loss of the optical signal at the end face. Furthermore, by setting the material to a photoluminescent material, the acquisition of light signal information is not affected, while the light signal can be transmitted normally. That is, by precisely controlling the size of the ratio 'a', the first light signal of ratio 'a' can continue to pass through the second end face 2 along the light path, while the remaining light signal is transmitted to the circumferential side face 3. This ensures both the integrity and efficiency of the light signal transmission, providing a high-quality original light signal foundation for subsequent edge coding and tomographic reconstruction, and also enables the transmission of light signals over longer distances.

[0025] To establish the information contained in the first optical signal of a more complex incident light source, especially for multi-point light sources or light sources including images, in a more preferred embodiment of the invention, the mapping relationship is obtained through supervised learning training, such as... Figure 2 As shown. Step S3 includes at least the following: Step S31: Import the supervised learning-trained mapping relationship into signal reconstruction module 6 to complete model initialization; In step S32, the signal reconstruction module 6 receives the second optical signal transmitted by the optical detection module 5, and analyzes and extracts the coding features, position information and light intensity data of the optical signal; Step S33: Based on the trained mapping relationship, perform inverse inversion operation on the extracted second optical signal parameters to deduce the corresponding original imaging optical signal, i.e., the third optical signal; Step S34: Perform precision calibration on the inverted third optical signal to eliminate minor errors generated during encoding and transmission, ensuring that the reconstructed optical signal can truly reflect the characteristics of the original incident optical signal, and achieve accurate mapping and complete reconstruction from the encoded optical signal to the original imaging optical signal.

[0026] To further supervise training by using the first optical signal as a label, it can be ensured that the mapping relationship can accurately associate the encoded second optical signal with the original optical signal. In a more preferred embodiment of the invention, the mapping relationship is trained through supervised learning, using the second optical signal as the input to the neural network and the third optical signal as the output of the neural network; training is performed using the first optical signal as the label, with the loss function being the error between the third and first optical signals; training is conducted using algorithms including backpropagation of convolutional neural networks or linear inverse calculation; when the error or the number of training epochs reaches a preset value, the neural network is output as the mapping relationship. For example, a neural network model is built, training parameters are defined, the second optical signal (i.e., the encoded optical signal) is used as the input to the neural network, and the third optical signal (i.e., the original imaging optical signal to be reconstructed) is used as the output of the neural network. The first optical signal (i.e., the original incident optical signal) is used as the training label, and the loss function is set to the error between the third and first optical signals. For example, this error value is set to 0.001, and a loss function, including MSE or MAE, is used to measure it. The model is trained using a convolutional neural network backpropagation algorithm or a linear inverse method, continuously iterating and optimizing the neural network parameters to reduce the error value of the loss function. The training process is monitored in real time. When the error of the loss function reaches or falls below a preset threshold of 0.001, or when the number of training epochs reaches a preset value, training stops, and the neural network model at this point is output. This model represents the mapping relationship between the second and third optical signals. Using a convolutional neural network backpropagation algorithm or linear inverse method for training improves training efficiency and model accuracy, avoids overfitting or underfitting, and ensures that the trained mapping relationship has good generalization ability, adapting to first optical signals of different intensities and patterns. The trained neural network is directly imported into the signal reconstruction module 6 to achieve fast and accurate signal inversion, further improving the reconstruction quality of the third optical signal and ensuring a high degree of consistency between the reconstructed signal and the original incident light signal. For example, the trained mapping relationship model is imported into the signal reconstruction module 6 to complete model initialization. The signal reconstruction module 6 receives the second optical signal transmitted from the optical detection module 5, analyzes and extracts the encoded features, position information, and light intensity data of the optical signal. Based on the trained mapping model, the extracted second optical signal parameters are inversely inverted to derive the corresponding original imaging optical signal, i.e., the third optical signal. The inverted third optical signal is then precision calibrated to eliminate minor errors generated during encoding and transmission, ensuring that the reconstructed optical signal accurately reflects the characteristics of the original incident optical signal. This achieves precise mapping and complete tomographic reconstruction from the encoded optical signal to the original imaging optical signal.

[0027] To convert incident light information, such as intensity, angle, and pattern distribution, into a signal transmitted to the circumferential side 3, and to further improve the transmission efficiency of the light signal to the circumferential side 3, ensuring that the subsequent convex lens array structure 4 can accurately capture and encode the complete incident light information and avoid loss of incident light information, in a more preferred embodiment of the invention, after receiving the first light signal from the first end face 1 in step S1, the light signal is transmitted to the circumferential side 3 along the parallel direction of the first end face 1 through scattering or photoluminescence. This ensures that the light signal is incident on the incident surface of the convex lens array structure 4, which must be stably fitted or integrally formed on the circumferential side 3 of the imaging plane to ensure array arrangement accuracy. Utilizing the focusing and refraction characteristics of each convex lens in the convex lens array, the incident continuous light signal is split into multiple independent sub-light signals, each sub-light signal corresponding to a convex lens unit. Simultaneously, through the focal length design of the convex lenses, the propagation angle of each sub-light signal is precisely modulated, with different modulation angles corresponding to convex lenses at different positions, thus achieving angle encoding of the light signal. Preferably, the convex lens array structure comprises uniformly arranged convex lens units, each convex lens unit including at least one convex lens, and the spacing between the convex lens units is consistent. The arrangement of the convex lens array follows a pattern, such as a rectangular array or a ring array. The emission position of each sub-light signal corresponds one-to-one with the array coordinates of the convex lens unit, forming a position code. Simultaneously, by adjusting the transmittance of the convex lenses, the light intensity of each sub-light signal is controlled, and combined with the original light intensity characteristics of the incident light signal, light intensity encoding is achieved. The multiple sub-light signals, after being modulated by angle, position, and light intensity, are combined to form at least one complete second light signal, which is directly transmitted to the optical detection module 5.

[0028] To better deploy the transparent tomography method based on an edge-convex lens array in practical applications, and to better transmit the light source signal entering the first end face 1 to the circumferential side face 3 without additional optical path guidance, thereby reducing optical signal transmission loss and improving coding efficiency, this invention also discloses a sensing device for performing the above-described transparent tomography method based on an edge-convex lens array. The sensing device includes a first end face 1, a second end face 2 opposite to the first end face 1, at least one optical detection module 5, and a signal reconstruction module 6. The first end face 1 and the second end face 2 are connected by a circumferential side surface 3; The circumferential side surface 3 is formed with a convex lens array structure 4; The optical detection module 5 is fixedly installed outside the convex lens array structure 4 and can be connected to the signal reconstruction module 6.

[0029] The sensing device directly forms the convex lens array structure 4 on the circumferential side 3, directly capturing the light signal transmitted from the first end face 1 to the circumferential side 3, such as the signal transmitted by scattering or photoluminescence. The optical detection module 5 is fixed outside the convex lens array to accurately receive the encoded second light signal, avoiding signal omission or bias. At the same time, it is directly connected to the signal reconstruction module 6 to realize the rapid transmission of the second light signal, providing a guarantee for subsequent optical signal tomography reconstruction.

[0030] To improve the sensing accuracy of the device and facilitate the capture of all optical signal information on the circumferential side 3, and to encode multiple sets of second optical signals for signal processing operations that enhance the sensing accuracy of the optical signals, in a more preferred embodiment of the invention, the first end face 1 and the second end face 2 are connected via the circumferential side 3 to form a cylindrical structure, which includes a prism structure or a cylindrical structure. The first end face 1, the second end face 2, and the circumferential side 3 form a prism or cylindrical cylindrical structure, resulting in a regular structure and a high degree of integration. The circumferential side 3 of the cylindrical structure is a continuous closed structure, allowing the convex lens array structure 4 to be arranged completely along the circumference, enabling 360° capture of the optical signals transmitted from the first end face 1 to the circumferential side 3 without blind spots. For example, the annular side of a cylindrical structure achieves omnidirectional encoding of the optical signals, avoiding signal loss. In a more preferred embodiment of the invention, such as... Figure 3 As shown, a regular polyhedral prism structure is adopted, and a regular triangular prism structure, a regular tetraangular prism structure, or a regular hexagonal prism structure is particularly preferred. Each convex lens unit in the edge convex lens array structure 4 maps light signals from different propagation directions to different positional responses on the optical detection module 5, thereby forming an angle-related code for the in-plane light field.

[0031] To acquire more comprehensive information about the optical signal, in a more preferred embodiment of the present invention, the prism structure comprises a regular polyhedral prism, and multiple prism structures can be fixedly arranged in parallel. The optical detection module 5 is disposed on at least one circumferential side 3 of the regular polyhedral prism. The parallel and fixed arrangement of multiple prism structures allows for the simultaneous series connection of multiple sensing devices, facilitating actual data acquisition and improving overall sensing accuracy.

[0032] To better acquire and sense the encoded optical signal data and improve the transmission quality of the optical signal during transmission, in a more preferred embodiment of the present invention, the optical detection module 5 includes a fiber optic array or an image sensor. The use of a fiber optic array or image sensor in the optical detection module 5 further enhances detection accuracy and stability. Fiber optic arrays have strong anti-interference capabilities, effectively suppressing the influence of stray light and electromagnetic interference on the detection signal, ensuring high purity of the acquired second optical signal, and providing a high-quality signal for the mapping relationship. Image sensors have high pixel precision, accurately capturing detailed features such as the intensity, angle, and position of the encoded optical signal, reducing signal acquisition errors, and are particularly suitable for high-precision imaging requirements. In a more preferred embodiment, each fiber corresponds to a convex lens encoding unit, eliminating signal crosstalk. During acquisition, the second optical signal of a single encoding unit is directly captured without additional signal separation, enabling rapid capture of weak encoded signals. Furthermore, the low optical fiber guide loss allows for rapid transmission of the second optical signal to the signal reconstruction module 6, significantly improving the detection efficiency of weak signals. If an image sensor is used, the detection surface can be ensured to cover the entire convex lens array by adjusting the installation angle, and the second light signal of all coding units can be captured at once without point-by-point acquisition, which greatly improves the detection speed. It is especially suitable for multi-light source and complex pattern scenes, and can realize the synchronous acquisition and transmission of coding signals.

[0033] In the above embodiments, the descriptions of each embodiment have different focuses. Parts not described in detail in a certain embodiment can be referred to in the relevant descriptions of other embodiments. The above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A transparent tomographic imaging method based on an edge-convex lens array, characterized in that, The transparent tomography method based on an edge convex lens array includes: Step S1: The incident light source is received by the first end face (1) to generate a first light signal. The first light signal is transmitted along the first end face (1) to the circumferential side face (3). Step S2: The first optical signal is encoded by a convex lens array structure (4) arranged on the circumferential side (3), and at least one set of second optical signals is obtained after encoding. The second optical signal is sensed by at least one optical detection module (5). In step S3, the optical detection module (5) transmits the second optical signal to the signal reconstruction module (6) for tomographic reconstruction to obtain the third optical signal.

2. The transparent tomography method based on an edge-convex lens array according to claim 1, characterized in that, The signal reconstruction module (6) establishes a mapping relationship between the second optical signal and the third optical signal.

3. The transparent tomography method based on an edge-convex lens array according to claim 2, characterized in that, The mapping relationship is obtained through supervised learning.

4. The transparent tomography method based on an edge-convex lens array according to claim 3, characterized in that, The mapping relationship is established through supervised learning training, in which the second optical signal is used as the input of the neural network and the third optical signal is used as the output of the neural network; the first optical signal is used as the label for training, the loss function is the error between the third optical signal and the first optical signal, and the training is performed by methods including backpropagation algorithm of convolutional neural network or linear inverse calculation. When the error or the number of training rounds reaches a preset value, the neural network is output as the mapping relationship.

5. The transparent tomography method based on an edge-convex lens array according to claim 1, characterized in that, In step S1, after receiving the incident light source from the first end face (1), a first light signal is generated by scattering or photoluminescence, and the first light signal is transmitted to the circumferential side face (3) along the parallel direction of the first end face (1).

6. The transparent tomography method based on an edge-convex lens array according to any one of claims 1-5, characterized in that, In step S1, after the incident light source passes through the first end face (1), the incident light source of proportion a is converted into a first light signal by the adjustment setting and continues to pass through the second end face (2) along the light path.

7. A sensing device for implementing the transparent tomographic imaging method based on an edge-convex lens array as described in any one of claims 1-6, characterized in that, The sensing device includes a first end face (1), a second end face (2) opposite to the first end face (1), at least one optical detection module (5), and a signal reconstruction module (6). The first end face (1) and the second end face (2) are connected by a circumferential side face (3); The circumferential side surface (3) is formed with a convex lens array structure (4); The optical detection module (5) is fixedly installed outside the convex lens array structure (4) and can be connected to the signal reconstruction module (6).

8. The sensing device according to claim 7, characterized in that, The first end face (1) and the second end face (2) are connected by the circumferential side face (3) to form a column structure, which includes a prism structure or a cylindrical structure.

9. The sensing device according to claim 8, characterized in that, The prism structure includes a regular polyhedral prism, and multiple prism structures can be fixedly arranged in parallel. The optical detection module (5) is provided on at least one of the circumferential side surfaces (3) of the regular polyhedral prism.

10. The sensing device according to any one of claims 7-9, characterized in that, The optical detection module (5) includes a fiber optic array or an image sensor.

Images