Imaging system and imaging method based on full Stokes polarization regulation and medical detection device

Abstract

The invention discloses an imaging system and method based on full Stokes polarization regulation and control and a medical detection device. The imaging system comprises a laser light source; the polarization device and the laser light source are arranged at an interval, and light beams of different polarization channels are coupled to three groups of polarization channels, namely a 0-degree / 90-degree linear polarization channel, a 45-degree / 135-degree linear polarization channel and a right-hand / left-hand circular polarization channel; and the photoelectric detection device is located at the downstream of the polarization device and is provided with three differential output channels corresponding to the three groups of polarization channels respectively. And selectively activating at least one of the channel A, the channel B and the channel C, converting an obtained optical signal into an electric signal by the photoelectric detection device, and converting the electric signal into an image gray scale through depolarization imaging and polarization coding ternary convolutional neural network imaging. According to the invention, polarization response is converted into a logic signal unit, a polarization convolution model for image processing is constructed, and image edge detection and intelligent identification are directly carried out on a detector level.

Description

Technical Field

[0001] This invention relates to the field of micro-nano optoelectronic technology, and in particular to an imaging system, imaging method and medical detection device based on full Stokes polarization modulation. Background Technology

[0002] Polarization, as a key degree of freedom characterized by Stokes vectors (S0 to S3), enables contrast enhancement and multidimensional information encoding in quantum communication, optical logic circuits, and biomedical imaging. The core of full Stokes reconstruction lies in achieving high-fidelity conversion from vector light fields to digital signals. Polarization photoelectric detection systems are advanced photoelectric detection systems integrating polarization sensing and photoelectric detection functions, capable of simultaneously detecting the intensity and polarization state of light signals.

[0003] Existing fully polarized photoelectric detection systems generally rely on optical modulation devices for beam splitting or modulation, followed by measurement via photodetectors. Traditional photodetectors mostly provide scalar intensity responses, lacking the electrical ability to distinguish polarization direction and chirality, and thus cannot perform image processing.

[0004] In view of this, it is necessary to provide an imaging system, imaging method and medical detection device based on full Stokes polarization modulation to solve one of the above-mentioned technical problems. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art, and provides an imaging system, imaging method and medical detection device based on full Stokes polarization modulation, establishing a chip-level pathway from polarization vector → photoelectric detection → image processing at the device level.

[0006] To achieve one of the above-mentioned objectives, the present invention adopts the following technical solution: An imaging system based on full Stokes polarization modulation includes: a laser source; a polarization device spaced apart from the laser source, the polarization device performing beam splitting and phase / amplitude encoding on the incident light, coupling beams from different polarization channels to six polarization channels, the six polarization channels being: a 0° linear polarization channel, a 90° linear polarization channel, a 45° linear polarization channel, a 135° linear polarization channel, a right-hand circular polarization channel, and a left-hand circular polarization channel; and a photodetector downstream of the polarization device, the photodetector having channels A, B, and C, channel A having channels respectively polarized to the 135° linear polarization channel. The photodetector has two detection units corresponding to the focal points of the 45° linear polarization channel; channel B has two detection units corresponding to the focal points of the 90° linear polarization channel and the 0° linear polarization channel, respectively; channel C has two detection units corresponding to the focal points of the left-hand circular polarization channel and the right-hand circular polarization channel, respectively; channels A, B, and C have a differential structure; wherein, at least one of channels A, B, and C is selectively activated, the photodetector converts the obtained optical signal into an electrical signal, and then converts the electrical signal into image grayscale through depolarization imaging and polarization-coded ternary convolutional neural network imaging.

[0007] In some implementations, the amplitude threshold is defined as the 0 state, and the positive and negative polarities of S1, S2, and S3 are mapped to +1 and -1, respectively. The ternary logic state (-1, 0, +1) is derived from the polarization-resolved photocurrent encoded by Stokes vectors, and the image grayscale is output through a polarization-encoded ternary convolutional neural network based on the ternary logic state.

[0008] In some implementations, when channel A is activated, 135° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN1 set to +1; when 0° or 90° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN1 set to 0; when 45° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN1 set to -1. Based on the ternary logic state (-1, 0, +1), the image grayscale is output through a polarization-coded ternary convolutional neural network.

[0009] In some implementations, when channel B is activated, 90° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN2 set to +1; when 45° or 135° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN2 set to 0; when 0° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN2 set to -1. Based on the ternary logic state (-1, 0, +1), the image grayscale is output through a polarization-coded ternary convolutional neural network.

[0010] In some implementations, when channel C is activated, left-hand circularly polarized light generates a positive signal, and the output value IN3 is set to +1; when it is deactivated, a zero signal is generated, and the output value IN3 is set to 0; right-hand circularly polarized light generates a negative signal, and the output value IN3 is set to -1. The image grayscale is output through a polarization-coded ternary convolutional neural network based on the ternary logic state (-1, 0, +1).

[0011] In some implementations, when channel A is activated, 135° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN1 set to +1; when 0° or 90° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN1 set to 0; when 45° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN1 set to -1. The image grayscale is output through a polarization-coded ternary convolutional neural network based on the ternary logic state (-1, 0, +1). When channel B is activated, 90° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN2 set to +1; when 45° or 135° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN2 set to 0; when 0° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN1 set to -1. IN1 is set to -1; channel C is activated, and left-hand circularly polarized light is input to the polarization device. The electrical signal output by the photodetector is positive, and the output value IN3 is set to +1; left / right circularly polarized light is turned off, generating a zero signal, and the output value IN3 is set to 0; right-hand circularly polarized light is input to the polarization device, and the electrical signal output by the photodetector is negative, and the output value IN3 is set to -1; by controlling IN1, IN2, and IN3, a ternary logic state (-1, 0, +1) is output together, and the image grayscale is output through a polarization-encoded ternary convolutional neural network based on the ternary logic state.

[0012] A medical detection device includes the aforementioned imaging system based on full Stokes polarization modulation and an image processing module.

[0013] An imaging method based on a fully Stokes polarization-controlled imaging system includes the following steps: a laser source outputs a beam of light with a predetermined polarization state; the polarized beam passes through the object to be detected or enters a polarization device, where the incident light is split and its phase / amplitude is encoded by the polarization device, coupling beams from different polarization channels to six polarization channels, namely: a 0° linear polarization channel, a 90° linear polarization channel, a 45° linear polarization channel, a 135° linear polarization channel, a right-hand circular polarization channel, and a left-hand circular polarization channel; the beam output by the polarization device enters a photodetector, selectively activating at least one of channels A, B, and C; the photodetector converts the obtained optical signal into an electrical signal, and then converts the electrical signal into image grayscale through depolarization imaging and polarization-encoded ternary convolutional neural network imaging.

[0014] In some implementations, the amplitude threshold is defined as the 0 state, and the positive and negative polarities of S1, S2, and S3 are mapped to +1 and -1, respectively. The ternary logic state (-1, 0, +1) is derived from the polarization-resolved photocurrent encoded by Stokes vectors, and the image grayscale is output through ternary convolution.

[0015] In some implementations, when channel A is activated, 135° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN1 set to +1; when 0° or 90° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN1 set to 0; when 45° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN1 set to -1, and the image grayscale is output through ternary convolution.

[0016] In some implementations, when channel B is activated, 90° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN2 set to +1; when 45° or 135° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN2 set to 0; when 0° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN2 set to -1, and the image grayscale is output through ternary convolution.

[0017] In some implementations, when channel C is activated, left-hand circularly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN3 set to +1; when left-hand / right-hand circularly polarized light is turned off, a zero signal is generated, and the output value IN3 is set to 0; when right-hand circularly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, and the output value IN3 is set to -1, and the image grayscale is output through ternary convolution.

[0018] In some embodiments, when channel A is activated, 135° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN1 set to +1; when 0° or 90° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN1 set to 0; when 45° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN1 set to -1; when channel B is activated, 90° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN2 set to +1; when 45° or 135° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN2 set to 0; when 0° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN2 set to -1; when channel C is activated, left-handed circularly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN1 set to -1. IN3 is set to +1; left / right circularly polarized light is turned off, generating a zero signal, and the output value IN3 is set to 0; right circularly polarized light is input to the polarization device, and the electrical signal output by the photodetector is negative, so the output value IN3 is set to -1; by controlling IN1, IN2, and IN3, a ternary logic state (-1, 0, +1) is output together, and the image grayscale is output through ternary convolution.

[0019] The beneficial effects of this invention are as follows: The imaging system based on full Stokes polarization modulation of this invention uses polarization devices to split the incident light and encode its phase / amplitude, coupling beams from different polarization channels to the aforementioned three sets of orthogonal polarization channels. A photodetector is located downstream of the polarization devices and has three differential output channels corresponding to the three sets of orthogonal polarization channels. By selectively activating at least one of channels A, B, and C, the photodetector converts the acquired optical signal into an electrical signal, which is then converted into image grayscale through depolarization imaging and polarization-encoded ternary convolutional neural network imaging. This invention converts the polarization response into a logic signal unit and uses the detector response matrix in image processing tasks, enabling direct edge enhancement and directional filtering at the detector unit 20. It constructs a polarization convolution model for image processing, achieving direct image edge detection and intelligent recognition at the detector level. This system can perform boundary recognition and semantic segmentation on disordered biological tissues (such as human brain white matter), demonstrating broad application prospects in biomedical imaging, intelligent photon sensing, and machine vision processing. Attached Figure Description

[0020] Figure 1 This diagram illustrates the geometric structure and working principle of a polarization device according to one embodiment of the present invention.

[0021] Figure 2The measurement basis vectors on the Poincaré sphere are illustrated: six analytical states form a maximally inscribed regular octahedron, minimizing the reconstruction error.

[0022] Figure 3 Simulated diagram of focal plane light intensity distribution of a superlens with an aperture of 50 μm under incident conditions of 0° linear polarization, 90° linear polarization, 45° linear polarization, 135° linear polarization, right-hand circular polarization (RCP), and right-hand circular polarization (LCP).

[0023] Figure 4 The target value (left), optimization result (middle), and full-wave simulation result (right) of the average polarization dichroism of the six polarization channels are compared.

[0024] Figure 5 For the design of micro / nano structures of metasurfaces and metasurface structures, (a) micro / nano structures and library mapping strategies for superlenses; (b) optical photographs of metasurfaces.

[0025] Figure 6 This is a schematic diagram showing the arrangement of the six polarization channels of the polarization device in one embodiment of the present invention.

[0026] Figure 7 This is a schematic diagram showing the arrangement of the six polarization channels of the polarization device in another embodiment of the present invention.

[0027] Figure 8 This is a schematic diagram showing the arrangement of the six polarization channels of the polarization device in another embodiment of the present invention.

[0028] Figure 9 This is a schematic diagram showing the arrangement of the six polarization channels of the polarization device in another embodiment of the present invention.

[0029] Figure 10 This is a schematic diagram showing the arrangement of the six polarization channels of the polarization device in another embodiment of the present invention.

[0030] Figure 11 This is a schematic diagram of the channel structure of a photodetector according to one embodiment of the present invention.

[0031] Figure 12 for Figure 11 An optical photograph of the photodetector shown.

[0032] Figure 13 This is a schematic diagram of the channel structure of the photodetector in another embodiment of the present invention.

[0033] Figure 14 for Figure 11 The diagram shows a reconfigurable wiring layout for a photodetector. (a) shows the reconfigurable wiring layout for the photodetector in differential mode, and (b) shows the reconfigurable wiring layout for the photodetector in summation mode.

[0034] Figure 15 Indication Figure 14 Photocurrent distribution diagrams of the six detection units under two working modes.

[0035] Figure 16 Indication Figure 14 Box plot of photocurrent intensity in the activated state of the middle channel A.

[0036] Figure 17 Indication Figure 14 The normalized response matrices of the four detection channels demonstrate low crosstalk characteristics and consistent multi-channel readout performance.

[0037] Figure 18 Indication Figure 14 The four detection channels in the middle have uniform and stable photocurrent output.

[0038] Figure 19 This is a schematic diagram of a metasurface cascaded with a photodetector in one embodiment of the present invention.

[0039] Figure 20 The diagram illustrates the Poincaré sphere full Stokes parameter distribution analyzed using a lightweight fully connected neural network based on the measured photocurrent response.

[0040] Figure 21 The diagram illustrates the results of the photoelectric pump-probe test, (i) the pump-probe timing diagram used to evaluate the time response, (ii) the measured results of the 6.1 ps time-resolved pulse response, and (iii) the stable switching of the normalized photocurrent.

[0041] Figure 22 This is a schematic diagram showing the test results of the stability, orthogonal response, and broadband performance of the multiple detection channels 21 of the present invention.

[0042] Figure 23 This paper presents a binary-ternary logic system with polarization encoding, reconfigurable logic gates, and optical encryption applications. (a) illustrates a reconfigurable optoelectronic logic platform; (b) illustrates the polarization-selective response of three logic inputs; (c) illustrates the physical diagram of the extended logic circuit and devices; (d) illustrates a reversible two-stage gate cascade system formally represented as an affine mapping on GF(2); (e) illustrates the experimentally implemented XOR-XNOR cascade; (f) illustrates the all-optical encryption-decryption process; and (g) illustrates the normalized photocurrent trajectory measured during character sequence transmission, verifying the correct encoding / decoding function and robust multi-channel operation performance.

[0043] Figure 24 It implements all seven basic binary logic gates.

[0044] Figure 25The structure and implementation of ternary logic gates are shown; (a) illustrates the ternary logic state defined based on the Stokes vector response; (b) illustrates the truth table of a typical ternary logic operation; and (c) illustrates the experimentally measured quantization step response of the "ternary adder" gate when the N-Z-P input changes.

[0045] Figure 26 This section describes the processing of ternary polarization convolution and in-sensor convolution. The left image shows the input image; the middle image shows the polarization resolution intensity set being mapped to ternary weights and the convolution operation being performed within the sensor pixel array; the green miniature matrix illustrates the physical mapping relationship between each polarization channel and the weights; the right image shows an example of Prewitt-X convolution.

[0046] Figure 27 This diagram illustrates the edge extraction process based on polarization convolution.

[0047] Figure 28 This is a schematic diagram of image processing and interventional polarization disordered tissue detection; (a) is a schematic diagram of an intravascular imaging scene: fiber optic integrated probes acquire depolarization contrast images of disordered tissue; (b) shows the size relationship between blood vessels, fiber optic probes and polarization-sensitive detector arrays; (c) shows a depolarization rate mapping map highlighting tissue scattering heterogeneity; (d) shows a normalized grayscale image used for processing; and (e) shows edge features extracted by polarization ternary convolution. Detailed Implementation

[0048] The present invention will now be described in detail with reference to the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, methodological, or functional modifications made by those skilled in the art based on these embodiments are included within the scope of protection of the present invention.

[0049] In the various figures of this invention, for ease of illustration, some dimensions of structures or parts may be exaggerated relative to other structures or parts; therefore, only the basic structure of the subject matter of this invention is used to illustrate the invention.

[0050] Please refer to Figures 1 to 28 As shown, the present invention provides a polarization device based on full Stokes polarization modulation, a photodetector device 2 for detecting full Stokes polarization, a photodetector system and reconstruction method for reconstructing full Stokes vectors, a logic control device based on full Stokes polarization modulation, and an imaging system based on full Stokes polarization modulation, so as to establish a chip-level path from polarization vector → photoelectric logic → image convolution at the device level.

[0051] Please refer to Figures 1 to 10As shown, this invention provides a polarization device based on full Stokes polarization modulation for beam splitting and phase / amplitude encoding of incident light, achieving multi-dimensional control of the optical field. Specifically, the polarization device first splits the incident light, physically separating it into different spatial positions to obtain beams with different polarization states (polarization channels). The polarization device then performs phase / amplitude encoding on the beams of different polarization channels, focusing them onto predetermined, precise positions to ensure they accurately fall onto the downstream photodetector device 2. The light spots of the multiple beams after spatial separation and phase modulation are spatially separated on the focal plane to achieve decoupling of the physical information channels. The intensity of each beam's spot represents the intensity component of the incident light on the corresponding polarization basis.

[0052] The polarization device directionally couples (focuses) the incident light beams from different polarization channels into six polarization channels (three sets of orthogonal polarization channels). The six polarization channels are: 0° (horizontal, also labeled X or H) linear polarization channel, 90° (vertical, also labeled Y or V) linear polarization channel, 45° (also labeled A) linear polarization channel, 135° (-45°, also labeled D) linear polarization channel, right-hand (R) circular polarization channel, and left-hand (L) circular polarization channel.

[0053] The six polarization channels form the smallest and most complete basis set for calculating the full Stokes parameters (S0, S1, S2, S3), representing the optimal and redundancy-free solution for achieving full polarization measurement. The focal points of the six polarized beams output by the polarization device correspond to the six polarization channels, allowing the positions of each polarization state on the Poincaré sphere to form the largest inscribed regular octahedron, laying the foundation for reconstructing the complete Stokes parameters (S0 to S3).

[0054] In order to split the incident light beam and encode it into six polarization channels, the polarization device is selected from at least one of a superlens 1, a grating, and a micro / nano photonic device.

[0055] In some embodiments, the polarization device is a superlens 1 based on full Stokes polarization control, which has a metasurface formed by several micro / nano structures 12. The phase and amplitude of the outgoing light can be precisely controlled through the metasurface, and the incident light can be split and phase / amplitude encoded to achieve precise control of the polarization channels and enhanced absorption. The superlens 1 maps different polarization basis vectors (such as H / V, ±45°, R / L) in the incident light to the above six polarization channels, realizing spatial separation of polarization channels (spatial decoupling).

[0056] By integrating polarization beam splitting and focusing functions into a single metasurface, six polarization channels are split and focused through six polarization detection points, enabling simultaneous, real-time, and parallel capture of the dynamic changes in six polarization states. This invention integrates the complex optical system of a full Stokes parameter measurement system onto a tiny metasurface, resulting in a simple structure, small size, and convenient integration design with photodetectors.

[0057] To achieve efficient polarization analysis, this invention first employs the Jones matrix phase retrieval algorithm to design the Jones matrix distribution of the metasurface. Simultaneously, to suppress coma caused by the inherent beam deflection of the superlens 1 profile, a ray tracing algorithm is further used, and software is employed to optimize the profile of the superlens 1.

[0058] Furthermore, taking a superlens 1 with an aperture diameter of 50 μm as an example, this invention calculates and optimizes the focal plane light intensity distribution of the vector metasurface when it is incident with linearly polarized light of 0° (H), 90° (V), 45° (D), 135° (A), right-handed (R) circularly polarized light, and left-handed (L) circularly polarized light, respectively. The results are as follows: Figure 3 As shown.

[0059] For incident light in the analytical state, the central intensity at the corresponding focal point reaches its maximum, while the intensity of its orthogonal states is eliminated, indicating high polarization extinction performance. Zero-order focusing efficiency is effectively suppressed, which is beneficial for achieving efficient, low-crosstalk polarization detection. For example, when the central intensity of 45° linearly polarized light is strongest, the central intensity of 135° linearly polarized light is 0; conversely, when the central intensity of 135° linearly polarized light is strongest, the central intensity of 45° linearly polarized light is 0.

[0060] To verify the polarization analysis performance of the superlens 1, this invention defines the polarization extinction ratio at each focal point as follows: ,in, and They represent the analytical state respectively. and its orthogonal state The intensity of light at the center of the m-th focal point under incident light.

[0061] This invention takes a superlens 1 with an aperture diameter of 50μm as an example and tests its polarization dichroism (PD) through full-spectrum simulation. Figure 4The average PD values ​​for all six polarization channels are shown, corresponding to the target value, optimized results, and simulation results, respectively. The optimized PD performance of the superlens 1 is above 0.95 in all six polarization channels, while the simulation results show a significant decrease in the circular polarization channel. This difference can be attributed to neglecting the coupling effect between adjacent micro / nano structures 12 during the design process. Nevertheless, the overall focusing efficiency of the superlens 1 of this invention still reaches 74%, maintaining a high average PD performance of 0.94, which is sufficient to meet the requirements of high-precision polarization detection.

[0062] Please refer to Figure 5 As shown, this invention utilizes the library mapping strategy of the micro / nano structure 12 to determine the design parameters of the metasurface and achieve an optimized spatial Jones matrix distribution.

[0063] Specifically, the superlens 1 includes a transparent substrate 11, micro / nano structures 12 located on the transparent substrate 11, and fillers 13 filling the spaces between the micro / nano structures 12. For ease of description, a three-dimensional spatial coordinate system xyz is established, where the x, y, and z directions are mutually perpendicular. The transparent substrate 11 lies in the xy plane, with its thickness direction along the z direction. The micro / nano structures 12 extend from the transparent substrate 11 along the z direction to at least one side.

[0064] The transparent substrate 11 can be, but is not limited to, a glass substrate. Preferably, the transparent substrate 11 is an optical-grade glass wafer, such as BF33 glass or BK7 glass.

[0065] The micro / nano structure 12 is one or more combinations of micro / nanowires, micro / nanopilles, and micro / nanopores. Utilizing the electromagnetic resonance effect of the micro / nano structure 12, the phase, amplitude, and polarization of light waves can be precisely controlled at the subwavelength scale, which is the basis for realizing complex polarization beam splitting and focusing functions.

[0066] The metasurface formed by the subwavelength micro / nano structure 12 has higher transmission efficiency and lower absorption loss, and can be manufactured using semiconductor micro / nano fabrication technology, which is conducive to achieving low-cost, mass production.

[0067] The micro / nanostructure 12 extends from the transparent substrate 11 along the z-direction to one side, with its height H in the z-direction. The cross-sectional shape of the micro / nanostructure 12 is rectangular or elliptical, with extension in either the x or y direction; that is, the major axis of the ellipse or the long side of the rectangle extends along the x or y direction. In this art, the cross-section of the micro / nanostructure 12 refers to the cross-section formed by cutting in a direction perpendicular to its height.

[0068] The micro / nano structure 12 has a dimension Dx in the x-direction, which can produce a phase delay in the x-direction; and a dimension Dy in the y-direction, which can produce a phase delay in the y-direction. The rotation angle θ of the micro / nano structure 12 about its center from the x-direction to the y-direction is θ. By controlling the geometric parameters Dx, Dy, and θ, this invention can precisely "encode" each superlens unit, thereby realizing complex and efficient polarization optics functions.

[0069] Specifically, by changing its dimensions Dx and Dy, its phase response to x- and y-polarized light can be independently tuned; by changing the rotation angle θ, the orientation of its optical axis can be controlled. This is key to realizing polarization-dependent phase profiles (i.e., Jones matrix design), enabling the metasurface to apply different phase modulations to light of different polarization states, thereby guiding them to different focal points.

[0070] In one embodiment, Dx ≥ Dy, -90° ≤ θ ≤ +90°.

[0071] In some embodiments, the micro / nano structure 12 has a wire diameter ratio of 0.2 to 1, which provides high mechanical stability, high optical efficiency, and ease of fabrication, ensuring the feasibility and high performance of the device in actual manufacturing.

[0072] In some embodiments, the micro-nano structures 12 are arranged at equal intervals, which simplifies the design of the metasurface and helps to suppress optical crosstalk between different superlens units, ensuring that each superlens unit can independently and accurately perform its preset electromagnetic control function.

[0073] The filler 13 is used to protect the micro / nano structure 12 and can be photoresist or silicon dioxide. The filler 13 is higher than the micro / nano structure 12, and its height is represented by Hsup. In one embodiment, such as Figure 5 As shown in (a), the transparent substrate 11 is BF33 glass, the micro / nano structure 12 is a polycrystalline silicon nanopillar, and the filler 13 is photoresist. The photoresist completely covers the nanopillar, and the height H of the photoresist is... sup It is greater than the height H of the nanopillar.

[0074] The cross-sections of the micro / nanopillars are rectangular. Some micro / nanopillars have larger cross-sectional areas than others, and the deflection angles differ in different regions. The superlens 1 operates at a wavelength of 940 nm and has a lattice period of 500 nm. The nanopillars have the same height of 658 nm; however, they possess different dimensional parameters (Dx, Dy) and rotation angles θ relative to the reference coordinate system in different directions within the xy plane.

[0075] The superlens 1 fabricated based on the above design has the following structure: Figure 5 As shown in (b) of the image, the scanning electron microscope image illustrates the structure and arrangement of the micro / nano structure array.

[0076] In polarization devices, the positions of the six polarization channels and their focal points are determined by the polarization detection points. However, this invention optimizes the arrangement of the six polarization detection points by comprehensively considering factors such as the ease of fabrication of the photodetector, crosstalk between the detection signals at each point, and detection efficiency.

[0077] In this invention, the focal points of the six polarization channels are located in the same plane, and the lines connecting the two focal points of the 0° / 90° linear polarization channel, the two focal points of the ±45° linear polarization channel, and the two focal points of the left / right circular polarization channel do not intersect within the same plane. Based on this, the three differential output channels used to measure the focal spots of the 0° / 90° linear polarization channel, the ±45° linear polarization channel, and the left / right circular polarization channel, respectively, do not intersect, and the signal paths are physically independent, avoiding electrical crosstalk. The measurement of any one differential output channel is unaffected by interference from other polarization channels, ensuring measurement accuracy.

[0078] In some embodiments, taking the superlens 1 as an example, the arrangement of the polarization detection points is determined by the grating diffraction order distribution of the metasurface. Theoretically, any six polarization detection points can be selected from all grating diffraction orders. However, this invention comprehensively considers factors such as the ease of fabrication of the photodetector, crosstalk between the detection signals at each point, and detection efficiency, and arranges the six polarization detection points in a straight line along a first direction, so that the six polarization channels and / or their focal points are arranged in a straight line along the first direction. The first direction is the x-direction, but it can also be the y-direction or other directions.

[0079] One-dimensional linear arrangement is one of the most space-efficient multi-channel arrangements, facilitating the miniaturization of the entire photodetector / detection system. Furthermore, the linear arrangement of focal points allows the differential signal pairs used to calculate Stokes parameters (such as 0° and 90°, 45° and 135°, left-handed and right-handed) to naturally become adjacent or near-adjacent pixels in physical space. This greatly simplifies the layout and routing of subsequent differential amplifier circuits, reduces differences and crosstalk in signal transmission paths, and improves measurement speed and signal-to-noise ratio. It also significantly simplifies the design and manufacturing of downstream photodetector devices, greatly reducing alignment difficulties and accuracy requirements during optical system packaging, and improving production yield.

[0080] The structure with six focal points arranged in a straight line allows for more precise design of the spacing between each detection spot, and makes it easier to achieve isolation of photoelectric signals with different polarization states in the physical structure and readout circuit, thereby significantly reducing signal crosstalk and achieving high-precision polarization detection.

[0081] Considering that the beam splitting deflection angle of left / right circular polarization is limited by diffraction efficiency, in order to ensure high detection efficiency of the detector, this invention places the left / right circular polarization channel in the middle, that is, in a position close to the center; the 0° / 90° linear polarization channel is located on both sides of the left / right circular polarization channel, and / or the ±45° linear polarization channel is located on both sides of the left / right circular polarization channel.

[0082] With the left / right circularly polarized channels centered and at least one set of linearly polarized channels positioned on either side, this arrangement reduces the impact of manufacturing errors or near-field coupling between the left / right circularly polarized channels and the edge linearly polarized channels, thus helping to reduce optical crosstalk between channels. This invention places the left / right circularly polarized channels in the region of optimal aberration and illumination uniformity of the optical system, thereby obtaining two high-quality, best-matched focal spots. This significantly improves the signal-to-noise ratio and accuracy of differential measurements. Furthermore, this arrangement provides a basis for the symmetrical design of the entire device, facilitating a more symmetrical and balanced phase distribution, thereby reducing overall device aberrations, improving the focusing efficiency and spot quality of each channel, and facilitating subsequent differential measurements.

[0083] In some implementations, the measurement of the degree of circular polarization (DoCP) is typically used to calculate S3, where S3 = I R -I L Based on a linear arrangement, this invention symmetrically sets the focal points of the left / right circular polarization channels relative to the center of the first direction. This helps to compensate for systematic errors caused by uneven illumination or lens edge effects, thereby improving the measurement signal-to-noise ratio and accuracy of the differential signal. Furthermore, the symmetrical layout also helps to optimize the optical design, ensuring that the optical performance (such as focusing efficiency and spot quality) of the two circular polarization channels remains consistent. Simultaneously, the 0° / 90° linear polarization channels and the ±45° linear polarization channels are located on either side of the left / right circular polarization channels, respectively, which avoids interference from edge beams on the central left / right circular polarization channels.

[0084] In some embodiments, the six polarization channels along the first direction are arranged sequentially as follows: one of the ±45° linear polarization channels, one of the 0° / 90° linear polarization channels, one of the left / right circular polarization channels, another of the left / right circular polarization channels, another of the 0° / 90° linear polarization channels, and another of the ±45° linear polarization channels. This arrangement facilitates the wiring design of the internal electrodes of the photodetector 2, clearly groups the pixel pairs used for differential output, and reduces signal crosstalk. Furthermore, this arrangement forms a polarization angle gradient distribution along the first direction, allowing the phase modulation of linearly polarized light to transition smoothly along the gradient direction, avoiding abrupt interference between adjacent linear polarization channels, resulting in a low crosstalk rate.

[0085] In one specific embodiment, the six polarization channels are sequentially arranged along the first direction as follows: 135° linear polarization channel, 90° linear polarization channel, left-handed circular polarization channel, right-handed circular polarization channel, 0° linear polarization channel, and 45° linear polarization channel.

[0086] In some embodiments, the six polarization channels along the first direction are arranged sequentially as follows: one of the 0° / 90° linear polarization channels, one of the ±45° linear polarization channels, one of the left / right circular polarization channels, another of the left / right circular polarization channels, another of the ±45° linear polarization channels, and another of the 0° / 90° linear polarization channels. This arrangement also facilitates the wiring design of the internal electrodes of the photodetector 2. This symmetrical arrangement helps to better balance the aberrations of each channel during the design of the superlens 1, resulting in more uniform quality at the six focal points.

[0087] In one specific embodiment, the six polarization channels are sequentially arranged along the first direction as follows: 90° linear polarization channel, 135° linear polarization channel, left-handed circular polarization channel, right-handed circular polarization channel, 45° linear polarization channel, and 0° linear polarization channel.

[0088] In addition, with the two arrangement methods described above, the two sets of linear polarization channels are located symmetrically on both sides of the left / right circular polarization channel, which can ensure the signal integrity of the circular polarization channel and ensure the high accuracy of circular polarization measurement.

[0089] To ensure accurate detection of the S0 component in the Stokes parameter, the diffraction efficiency of the two outermost channels for each polarization channel distribution is set as one of the objective functions of the optimization algorithm during the design of metasurface devices using optimization algorithms.

[0090] In some embodiments, the focal point of the left / right circular polarization channel is located on the first side of the center in the first direction, that is, the focal point of the left / right circular polarization channel is adjacent to the center and offset to one side in the first direction. One set of 0° / 90° linear polarization channels and ±45° linear polarization channels are located on both sides of the left / right circular polarization channel, respectively, while the two channels of the other set are arranged adjacent to each other and located on the second side of the center in the first direction. This is to meet the requirements of other application scenarios, such as asymmetric optical path systems.

[0091] This invention utilizes a corresponding objective function to achieve a metasurface device that balances high diffraction efficiency and high polarization extinction ratio through iterative optimization.

[0092] The six polarization channels, arranged sequentially along the first direction, are: one of the ±45° linear polarization channels, one of the left / right circular polarization channels, another of the left / right circular polarization channels, another of the ±45° linear polarization channels, one of the 0° / 90° linear polarization channels, and another of the 0° / 90° linear polarization channels. Grouping the circular polarization channel group and the ±45° linear polarization channel group together, while placing the 0° / 90° linear polarization channels sideways and adjacent, is more conducive to the design of the back-end readout circuit.

[0093] In one specific embodiment, the six polarization channels are sequentially arranged along the first direction as follows: a 135° linear polarization channel, a left-handed circular polarization channel, a right-handed circular polarization channel, a 45° linear polarization channel, a 0° linear polarization channel, and a 90° linear polarization channel.

[0094] In other embodiments, the six polarization channels along the first direction are, in sequence: one of the 0° / 90° linear polarization channels, one of the left / right circular polarization channels, another of the left / right circular polarization channels, another of the 0° / 90° linear polarization channels, one of the ±45° linear polarization channels, and another of the ±45° linear polarization channels. Grouping the circular polarization channel group and the 0° / 90° linear polarization channel group together, and placing the ±45° linear polarization channels sideways and adjacent to each other, also facilitates the design of the back-end readout circuit.

[0095] In one specific embodiment, the six polarization channels are sequentially arranged along the first direction as follows: 0° linear polarization channel, left-hand circular polarization channel, right-hand circular polarization channel, 90° linear polarization channel, 45° linear polarization channel, and 135° linear polarization channel.

[0096] In both of the above embodiments, the total light intensity can be measured using either the 0° / 90° linear polarization channel or the ±45° linear polarization channel. Preferably, a pair of adjacent linear polarization channels are used to measure the total light intensity. For example, when the six polarization channels are sequentially a 135° linear polarization channel, a left-handed circular polarization channel, a right-handed circular polarization channel, a 45° linear polarization channel, a 0° linear polarization channel, and a 90° linear polarization channel, the total light intensity is measured using the 0° / 90° linear polarization channel.

[0097] In some implementations, the left / right circular polarization channel is placed in the middle, the ±45° linear polarization channel is located on one side of the left / right circular polarization channel, and the 0° / 90° linear polarization channel is located on the other side of the left / right circular polarization channel. In this case, the three sets of orthogonal polarization channels are arranged sequentially along the first direction, and each set can be used to measure the total light intensity, which is beneficial for the design of the back-end readout circuit.

[0098] The six polarization channels are, in sequence along the first direction: one of the ±45° linear polarization channels, another of the ±45° linear polarization channels, one of the left / right circular polarization channels, another of the left / right circular polarization channels, one of the 0° / 90° linear polarization channels, and another of the 0° / 90° linear polarization channels.

[0099] In one specific embodiment, the six polarization channels are sequentially arranged along the first direction as follows: 135° linear polarization channel, 90° linear polarization channel, left-hand circular polarization channel, right-hand circular polarization channel, 0° linear polarization channel, and 45° linear polarization channel.

[0100] In another specific embodiment, the six polarization channels are sequentially arranged along the first direction as follows: 45° linear polarization channel, 135° linear polarization channel, left-hand circular polarization channel, right-hand circular polarization channel, 90° linear polarization channel, and 0° linear polarization channel.

[0101] Of course, such as Figure 10 The six polarization channels and / or their focal points shown can be arranged in an array, concentrating the channels within a very small area. In some implementations, such as... Figure 10 As shown in (a), the six polarization channels and / or their focal points are arranged in two rows and three columns. The 0° / 90° linear polarization channels are in one column, the ±45° linear polarization channels are in another column, and the left / right circular polarization channels are in another column. The lines connecting the six focal points are as follows: Figure 13 As shown in channel 21 of (c), it is also possible to achieve three output channels without crossing.

[0102] In one specific embodiment, such as Figure 10 As shown in (a), in the first direction, the left / right circular polarization channel is located between the ±45° linear polarization channel and the 0° / 90° linear polarization channel. The left / right circular polarization channel being located in the middle prevents interference from surrounding light.

[0103] It should be noted that in any of the above arrangements, the positions of any set of orthogonal polarization channels can be interchanged. The positions of the 135° linear polarization channel and the 45° linear polarization channel can be interchanged, the positions of the 90° linear polarization channel and the 0° linear polarization channel can be interchanged, and the positions of the left-hand circular polarization channel and the right-hand circular polarization channel can be interchanged.

[0104] Furthermore, in polarization devices such as the superlens 1, the six polarization channels share the same focal length f, and they are focused onto the same plane along the optical axis. When the photodetector is precisely placed on this focal plane, the images of all six polarization channels are simultaneously clear without any defocusing differences. This is the basis for performing accurate polarization calculations (such as S0, S1, S2, and S3).

[0105] In one embodiment, the focal length f is from 270 μm to 2 mm. The polarization device can be coupled to a miniature photoelectric sensor.

[0106] The spacing d between the focal points (spot centers) of the six polarization channels is 20μm±10μm, ensuring complete differentiation between the multiple spots to prevent signal interference and minimize optical and electrical crosstalk between channels. A larger spacing results in less signal interference, while a smaller spacing leads to a smaller overall size. Considering both device size and detection accuracy, the preferred spacing d is 20μm±8μm, 20μm±6μm, or 20μm±3μm.

[0107] The six polarization channels decomposed from the polarization device form six measurement basis vectors, which can make the positions of each polarization state on the Poincaré sphere form the largest inscribed regular octahedron, thereby ensuring that the polarization measurement error is minimized.

[0108] In summary, the polarization device of this invention directionally couples incident light beams from different polarization channels to six polarization channels, allowing the beams from the six polarization channels to be output in parallel. The photodetector can simultaneously acquire the light intensity information of all channels in a single exposure. Combined with calibration and reconstruction algorithms, the complete Stokes vectors (S0, S1, S2, S3) can be retrieved in real time. Based on the fully Stokes polarization-controlled superlens 1, a multifocal vector lens is formed through a metasurface, resulting in a small size, ease of mass production, and convenient integration with the photodetector device 2.

[0109] Please refer to Figures 11 to 13 As shown. The present invention also provides a photodetector 2 for detecting fully Stokes polarized light beams, which detects Stokes polarized light beams and outputs polarized electrical signals.

[0110] To achieve full Stokes vectoring in conjunction with polarization devices The photodetector 2 has at least six detection units 20, each matched to the focal point of one of the six polarization channels, and outputs an electrical signal in response to the light intensity of the light spot. Two detection units 20 corresponding to a set of orthogonal polarization channels are paired to form a detection channel 21. The six detection units 20 are connected to form at least three differential output channels with differential response mechanism and one summation output channel with summation response mechanism.

[0111] The photodetector 2 includes multiple detection channels 21, each including at least three differential output channels and at least one summing output channel. Each differential output channel has two detection units 20 corresponding to two focal points of a set of orthogonal polarization channels, and the three differential output channels correspond to three sets of orthogonal polarization channels. The summing output channel is connected in parallel to any one of the differential output channels to measure the total light intensity.

[0112] Some detection channels 21 employ a differential structure to form differential output channels, mapping optically orthogonal polarization states into electrically symmetrical differential signals. The "signed" electrical quantities are read differentially and presented in the electrical signal as a combination of polarity and amplitude. This differential structure not only improves polarization contrast but also suppresses thermal drift and common-mode noise, effectively enhancing system stability and sensitivity. Other detection channels 21 are connected in parallel to any one of the differential output channels to obtain the total intensity information of the emitted light.

[0113] The differential response naturally generates bipolar (positive / negative) electrical signals, which perfectly match the sign characteristics of the Stokes parameters S1, S2, and S3, realizing a direct mapping from optical polarization information to electrical logic symbols, laying the physical foundation for subsequent analog calculations or logic operations.

[0114] This invention employs a collaborative design of three differential output channels and one summing output channel to simultaneously capture and directly convert the intensity information of six polarization basis vectors into Stokes parameters. This parallel processing mechanism fundamentally eliminates errors introduced by time-division measurement, enabling unambiguous, real-time reconstruction of the polarization state of dynamic light fields, and providing a hardware foundation for high-speed polarization imaging and sensing.

[0115] Furthermore, the signed electrical signal provided by the differential output channel can be directly used as the input for logic operations, making the detector itself an analog preprocessing unit. It can directly perform basic logic judgments or realize convolution operations in the polarization domain at the sensing front end, providing key hardware support for the "sensing and computing integrated" intelligent optical system for edge computing.

[0116] The six detection units 20 are pin 1, pin 2, pin 3, pin 4, pin 5, and pin 6, which can be combined in any two pairs to form different detection channels 21.

[0117] Please refer to Figure 11 and Figure 12 As shown, one of pin 1 and pin 2 is a detection unit 20 for the 135° linear polarization channel, and the other is a detection unit 20 for the 45° linear polarization channel. Pin 1 and pin 2 are connected to form channel A. That is, channel A has two detection units 20 corresponding to the focal points of the 135° linear polarization channel and the 45° linear polarization channel, respectively, for decoding the S2 vector polarization component.

[0118] One of pins 3 and 4 is a detection unit 20 for the 0° linear polarization channel, and the other is a detection unit 20 for the 90° linear polarization channel. Pins 3 and 4 are connected to form channel B. That is, channel B has two detection units 20 corresponding to the focal points of the 90° linear polarization channel and the 0° linear polarization channel, respectively, to decode the S1 vector polarization component.

[0119] One of pins 5 and 6 is a detection unit 20 for the right-hand circular polarization channel, and the other is a detection unit 20 for the left-hand circular polarization channel. Pins 5 and 6 are connected to form channel C. That is, channel C has two detection units 20 corresponding to the focal points of the left-hand circular polarization channel and the right-hand circular polarization channel, respectively, to decode the S3 vector polarization component.

[0120] Channels A, B, and C employ a differential structure, serving as differential output channels. They output positive and negative polarity currents in response to the light spot intensity of the corresponding orthogonal polarization channels. Since S1, S2, and S3 are inherently differential signals, the ratio of these differential signals remains relatively stable even when the total incident light intensity fluctuates. This allows the photoelectric sensor to obtain more reliable polarization information, unaffected by changes in brightness. Furthermore, the differential calculations are performed at the device level, eliminating the need for subsequent processing and significantly reducing the data volume and computational burden on subsequent digital processing units.

[0121] For ease of description, the summation output channel is denoted as channel D. Channel D is connected in parallel with any one of channels A, B, and C via electrodes to achieve a summation response. The total light intensity is measured through the existing detection channel 21 to decode the S0 vector polarization component.

[0122] Resource allocation can be optimized according to specific application scenarios, and channel D can be selectively connected in parallel with one of channels A, B, and C to obtain the total light intensity.

[0123] In some implementations, channel D is connected in parallel to the channel with the strongest light among channels A, B, and C, such as channel A. From a signal quality perspective, channel A corresponds to S2 and is typically the strongest and most stable channel among the linearly polarized channels. Connecting channel D, which is responsible for measuring the total light intensity S0, in parallel with channel A allows for a more reliable total light intensity measurement by leveraging the robustness of channel A.

[0124] In other implementations, channel D is connected in parallel to the outermost of channels A, B, and C. From a layout space perspective, selecting the outermost channel to lead out the parallel electrode avoids crossing other functional channels, minimizing the impact on the layout and wiring of internal channels, simplifying layout design, and reducing process complexity. Furthermore, external channels facilitate heat dissipation, preventing heat accumulation at the device center and improving the long-term operational stability of the device.

[0125] In other embodiments, channel D is connected in parallel to the channel with the largest physical size among channels A, B, and C. The larger physical size of the detector unit 20 typically results in higher photocurrent output and better driving capability, contributing to a more stable and stronger SO signal. Selecting the channel with the largest physical size to lead out the parallel electrodes can reduce manufacturing complexity and improve device reliability. Furthermore, when there are process deviations, the larger detector unit 20 exhibits relatively smaller fluctuations in performance parameters (such as responsivity), which is beneficial for maintaining consistent device performance during mass production.

[0126] Specifically, channel D includes an electrode pin 7, which is connected to any one of channels A, B, or C. Pin 7 can be connected to trench channel D on multiple differential output channels. During testing, one or more pins can be selected to acquire the total light intensity to verify measurement accuracy. Multiple pins 7 can also be used as backup test units.

[0127] It should be noted that the arrangement of the detection units 20 of the photodetector 2 is consistent with the arrangement of the focal points of the polarization channel of the polarization device. The description of the arrangement of the focal points of the polarization channel of the polarization device in this invention essentially includes the distribution of the detection units 20 of the corresponding photodetector 2 in the same manner, and will not be repeated here. Similarly, the description of the arrangement of the detection units 20 of the photodetector 2 in this invention essentially includes the distribution of the focal points of the polarization channel of the corresponding polarization device in the same manner, and will not be repeated here.

[0128] In this invention, such as Figure 11 and Figure 13 As shown in (a), (b), and (c), the three differential output channels do not overlap, and each probe channel 21 processes only one differential pair. The signal paths are physically independent, avoiding electrical crosstalk. The measurement of any differential output channel is not affected by other polarization channels, ensuring measurement accuracy.

[0129] like Figure 13 As shown in (d), when some of the differential output channels intersect, a dielectric layer 22 is provided at the intersection to electrically isolate the two intersecting differential output channels and prevent signal crosstalk.

[0130] In some implementations, channels A, B, and C each have two interconnected probe arms, symmetrically arranged to form a differential structure. This not only optimizes the shape but also improves the electrical performance and signal output. The probe unit 20 is located on the probe arm, preferably at its end; that is, the end of the probe arm is the probe unit 20 of the channel.

[0131] It should be noted that the symmetry settings of the present invention include both mirror symmetry and chiral symmetry.

[0132] The symmetrical probe arms ensure a high degree of consistency in the propagation path, parasitic parameters, and environmental interferences (such as thermal gradients and electromagnetic crosstalk) experienced by photogenerated carriers. This allows common-mode noise to be canceled to the greatest extent during differential amplification, thereby maximizing the inherent advantages of the differential architecture and achieving ultra-high signal-to-noise ratio and measurement accuracy.

[0133] The symmetrical structure also ensures a high degree of matching between the two detector units 20 in each differential pair in terms of photoelectric response characteristics (such as responsivity, response speed, and dark current). This allows the differential signal to truly reflect the differences in optical polarization information, greatly reducing calibration complexity and error.

[0134] The detector arms of channel AC are arranged in a mirror-symmetric manner. In this mode, according to the Shockley-Ramothe theorem, symmetrical detector pairs will generate photocurrents of opposite polarities, thereby enabling differential current operation within each channel.

[0135] In some implementations, channels A, B, and C are all U-shaped, with the two probe arms of each differential output channel symmetrically arranged along a first direction, corresponding to the polarization channels arranged in a straight line along the first direction. The U-shaped channels can be arranged in various ways.

[0136] In some implementations, such as Figure 11 and 12 As shown, the three channels are nested and compactly arranged, integrating six detector units 20 and their interconnections onto a minimal device, which is key to achieving device miniaturization. Furthermore, the detector units 20 in the three channels face the same direction and are located on the same straight line, facilitating interface with polarization devices. This structure is suitable for... Figure 6 or Figure 7 The polarization channel / focus arrangement is shown.

[0137] In one embodiment, channel A surrounds channel B, and channel B surrounds channel C; that is, from the outside to the inside, the channels are channel A, channel B, and channel C, respectively, to suit various applications. Figure 6 The polarization channel / focus arrangement is shown. Channel A (corresponding to S2) is usually the channel with the strongest and most stable signal among the linearly polarized channels. Channel D is connected to the outermost channel A through an electrode, making the detection circuit simple and easy to fabricate. Furthermore, channel D, which is responsible for measuring the total light intensity S0, is connected in parallel with it, relying on the robust performance of channel A to obtain a more reliable total light intensity.

[0138] In another embodiment, channel B is interchanged with channel A. Channel B wraps around channel A, and channel A wraps around channel C, to suit [the specific needs of the channel]. Figure 7 The polarization channel / focus arrangement is shown.

[0139] In some implementations, such as Figure 13 As shown in (a), channels A and B are arranged along a first direction, and channel C is nested within channel A or channel B. The arrangement of some channels surrounding or nesting them avoids interference from external signals to internal channels, ensuring signal stability and fidelity. The placement of some channels to the side is more beneficial for the design of the back-end readout circuit.

[0140] In one embodiment, channel B surrounds channel C, and channel A is located on one side of channel B in the first direction. Channel B filters out external signal interference to channel C, ensuring signal stability and fidelity. Placing channel A adjacent to channel C in a side-by-side configuration is more advantageous for the design of the back-end readout circuit. Furthermore, channel D can be connected in parallel to channel A or channel B via a single electrode.

[0141] In another embodiment, channel B is interchanged with channel A. Channel A wraps around channel C, and channel B is located on one side of channel A in the first direction, to suit... Figure 8 The polarization channel / focus arrangement is shown.

[0142] In some implementations, such as Figure 13 As shown in (b), channels A, B, and C are arranged sequentially along a first direction. Positioning channels A and B on opposite sides of channel C prevents interference from external signals and simplifies circuit design. Furthermore, channel D can be connected in parallel to channels A, B, or C via a single electrode to suit various circuit configurations. Figure 9 The polarization channel / focus arrangement is shown.

[0143] The above arrangement, by physically isolating channels that respond to different Stokes parameters and combining this with a reasonable gap and ground shielding design, effectively cuts off electrical crosstalk paths between channels. In particular, placing channel C, which responds to the weakest signal (such as S3), in the innermost layer minimizes interference from external high-speed switching noise, ensuring the reconstruction fidelity of all parameters, especially the weak polarization channel.

[0144] In some implementations, the six polarization channels defined by the polarization device are arranged in a straight line along a first direction. The two detection units 20 of channel A, the two detection units 20 of channel B, and the two detection units 20 of channel C are also arranged in a straight line along the first direction. All the detection units 20 are linearly arranged along the first direction, forming a very good topological match with the linear focal array generated by the metasurface isopolarization device. They are indiscriminately aligned with the six polarization channels on a single baseline to eliminate interference and improve detection accuracy and precision. This greatly simplifies the packaging process for opto-electric coupling and reduces alignment difficulty and tolerance requirements. Simultaneously, this arrangement effectively shortens the length of the critical differential signal path.

[0145] In one embodiment, pins 1, 3, 5, 6, 4, and 2 are arranged in a mirror-symmetric manner. Channel A is located on the outermost side, and pin 7 is connected to channel A on the outermost side. From both a physical design and light intensity perspective, this is the optimal choice.

[0146] Furthermore, the mirror-symmetric arrangement ensures that each differential pair (e.g., PIN 1 and PIN 2, PIN 3 and PIN 4, PIN 5 and PIN 6) is completely symmetrical in space. This strict geometric symmetry guarantees that the two units of the differential pair are highly consistent in photoelectric response characteristics, parasitic parameters, and environmental disturbances (such as thermal gradients and electromagnetic noise), thereby achieving maximum common-mode noise cancellation during differential amplification and pushing the device's signal-to-noise ratio and long-term stability to the theoretical limit.

[0147] In one specific embodiment, pin 7 is connected to the channel on channel A that connects two detection units 20 (pin 1 and pin 2), and is spaced apart from pin 1 and pin 2, which can effectively reduce signal interference; it also avoids directly drawing the S0 signal from a certain detection unit 20, which would damage the perfect symmetry of the original differential pair, and ensures that the performance of channel A when performing differential detection is not affected.

[0148] In another embodiment, the six detection units 20 are arranged as pin 1, pin 2, pin 3, pin 5, pin 6, and pin 4, with pin 3, pin 5, pin 6, and pin 4 arranged in a mirror-symmetrical manner. Alternatively, the six detection units 20 are arranged as pin 3, pin 4, pin 1, pin 5, pin 6, and pin 2, with pin 1, pin 5, pin 6, and pin 2 arranged in a mirror-symmetrical manner. Furthermore, pin 7 is preferably connected to the channel connecting two detection units 20 on channel A or channel B.

[0149] In another embodiment, the six detection units 20 are arranged as pin 1, pin 2, pin 5, pin 6, pin 3, and pin 4. Preferably, pin 7 is connected to the channel connecting two detection units 20 on channel A, channel B, or channel C.

[0150] Regardless of the arrangement of the detector units 20, the dimensions of pins 1, 2, 3, 4, 5, and 6 in the first direction are 15μm ± 2μm, and the spacing between adjacent detector units 20 is 5μm ± 0.5μm. This size design ensures sufficient tolerance when aligned with polarization devices. Furthermore, the unit size of approximately 15μm ensures sufficient photosensitive area for good responsivity while achieving an ultra-high integration density of approximately 120μm in the core region of the detector array. The ±2μm tolerance range provides the necessary window for process implementation, ensuring mass production feasibility.

[0151] In one embodiment, pins 1, 2, 3, 4, 5, and 6 are arranged at equal intervals along a first direction. This equal-interval arrangement of the probe units 20 greatly simplifies the layout design and process control of key fabrication processes such as photolithography and etching, significantly improving manufacturing yield and batch consistency. The 5μm spacing provides ample space for metal interconnects while ensuring optical and electrical isolation between units, effectively suppressing crosstalk between adjacent channels.

[0152] In addition, the two detection units 20 of channel A, the two detection units 20 of channel B, and the two detection units 20 of channel C can also be arranged in multiple rows and columns, making the test area more concentrated.

[0153] In one embodiment, such as Figure 13 As shown in (c), pins 1 and 2 form one column, pins 3 and 4 form another, and pins 5 and 6 form a third, arranged along the first direction, with the column containing pins 3 and 4 in the middle. Pin 7 can be connected to the channel connecting two detection units 20 on channel A, channel B, or channel C. In this case, the three differential output channels do not intersect.

[0154] In another embodiment, such as Figure 13 As shown in (d), the six detection units are arranged in two rows and three columns. Pins 3 and 4 are located in the middle column, pins 1 and 2 are arranged along one diagonal, and pins 5 and 6 are arranged along the other diagonal. Preferably, pin 7 is connected to the channel connecting two detection units 20 on channel A, channel B, or channel C. In this case, the three differential output channels intersect, and a dielectric layer 22 is disposed at the intersection.

[0155] Please refer to Figures 12 to 14 As shown, based on the above four detection channels 21A-D, the photodetector of the present invention has at least two operating modes: Mode I (differential operating mode) and Mode II (summation operating mode) to achieve complete extraction of all Stokes parameters.

[0156] Figure 14 Figure (a) illustrates the differential operating mode. The detector arms and detector units 20 of channel AC are arranged in a mirror-symmetric manner. According to the Shockley-Ramotheorem, the symmetrical detector units 20 (pin 1 and pin 2, pin 3 and pin 4, pin 5 and pin 6) will generate photocurrents with opposite polarities, thereby realizing differential current operation in each channel. Utilizing the characteristic of symmetrical changes in each pair of light spots, Stokes parameters S1 to S3 are extracted using photoelectric signals of the same intensity but opposite direction.

[0157] like Figure 15 As shown in (i), in mode I, channels A–C are activated individually. At an incident power of 0.5 mW, the three sets of symmetrical detector units 20 pairs generate photocurrents with similar amplitudes but opposite polarities, mainly localized at the gold-graphene interface. This indicates that channels A–C exhibit symmetrical bipolar signals. Specifically, activating channel A individually decodes the S2 vector polarization component. Activating channel B individually decodes the S1 vector polarization component. Activating channel C individually decodes the S3 vector polarization component.

[0158] Figure 14 Figure (b) illustrates the summation operation mode. The channel connected in parallel with channel D is activated, and channel D is reconstructed via external circuitry; other differential output channels are deactivated. The activated differential output channels are connected in parallel to one end of the external circuit (e.g., positive +), and pin 7 is connected to the other end of the external circuit (e.g., negative -). The intensities of the two light spots acquired by the activated differential output channels are summed, and the absolute light intensity parameter S0 is obtained based on the summation response mechanism. This reconfigurable hybrid scheme can selectively extract vector and scalar information from the incident light field.

[0159] In one embodiment, taking channel A surrounding channel B, channel B surrounding channel C, and channel D connected in parallel with channel A as an example, channel B and channel C are turned off, channel A is activated, and channel D is reconstructed through an external circuit. The two detection units 20 PIN 1 and PIN 2 of channel A are connected in parallel to align (have the same) photocurrent polarity. PIN 1 and PIN 2 are connected in parallel to one end of an external circuit (e.g., positive +), and pin 7 is connected to the other end of the external circuit (e.g., negative -). The intensities of the two light spots (the light spots of the 0° linearly polarized channel and the 90° linearly polarized channel) acquired by channel A are summed, and the absolute light intensity parameter S0 is obtained based on the summation response mechanism.

[0160] In mode II, the two detector units 20 in channel D exhibit the same polarity, clearly demonstrating the intrinsic reconfigurability between modes I and II. Figure 15Figures (ii)–(iii) further demonstrate the cross-channel photocurrent distribution of all six detector units 20 when only channel A is selectively activated, confirming that the photocurrent is strictly limited to the activated detector units 20 PIN 1 and PIN 2.

[0161] This invention embeds the differential response mechanism and the additive response mechanism into the device level through the symmetry of the spatial electrode layout and the junction (e.g., a PN junction made of PN material), forming an intrinsic analog computing architecture. Therefore, signal encoding based on polarity and intensity is directly implemented physically on the photodetector device 2.

[0162] Tests show that the photodetector exhibits robust reconfigurability and multi-channel electrical readout capability; details can be found in [reference needed]. Figure 15 The photocurrent distribution diagrams of the six detection units 20 are shown under the two working modes.

[0163] Please refer to Figure 16 This invention employs statistical analysis using normalized box plots. When channel A is activated individually, detector elements PIN 1 and pin 2 exhibit highly matched photocurrent intensities, with a median value of approximately 20 nA. In contrast, the signals from the inactive detector channels (PIN 3, pin 5) and the background region (BG) are significantly lower, remaining at approximately 2 nA–3 nA, with a signal-to-noise ratio of approximately 10 dB. Based on this, this invention systematically evaluates the dynamic response of all four detector channels 21.

[0164] Please refer to Figure 17 This invention uses a normalized response matrix to quantitatively demonstrate extremely low inter-channel crosstalk and excellent multi-channel electrical readout performance.

[0165] Please refer to Figure 18 This invention summarizes the photocurrent distribution of all four output channels, highlighting the uniformity and stability of the entire array signal.

[0166] The photodetector 2 is a device that outputs an electrical signal in response to an optical signal. In this invention, the photodetector 2 can be selected from any one of a photothermal electroelectric detector (PTE detector), a photodetector 2 (PV detector), or a PC detector.

[0167] PTE (Photothermoelectric) detectors convert optical signals into electrical signals based on the photothermoelectric effect. After absorbing light energy, the material does not directly excite electrons; instead, it converts the light energy into localized heat through nonradiative relaxation. Then, due to the asymmetry present in the material, a temperature gradient is formed under illumination. This temperature gradient drives charge carriers (electrons or holes) to diffuse from the hot end to the cold end, thereby generating a voltage across the device, known as the photothermoelectric voltage. PTE detectors are temperature gradient driven, self-powered, and have extremely fast response times. Their differential and polarity-sensitive characteristics make them suitable for polarization detection.

[0168] In one embodiment, the PTE detection device includes multiple PTE detection units 20 composed of graphene and gold electrodes, i.e., a graphene-Au photodetector array. Based on the Schottky junction formed by the graphene-Au material, the response time is less than 10 picoseconds, and it operates under zero bias conditions, possessing advantages such as self-powering, wide bandwidth, and polarization sensitivity.

[0169] Specifically, Au forms pins, and graphene is connected in pairs to the corresponding pins to form multiple detection channels 21.

[0170] Each PTE detector unit 20 operates via the photothermoelectric effect, exhibiting an intrinsic broadband self-powered response (covering the visible-infrared band), intrinsic polarization independence, and ultrafast response characteristics. Under illumination, localized heating at the gold-graphene interface establishes an asymmetric thermal gradient. This gradient, combined with the non-uniform doping induced by the gold contact, generates an internal Seebeck gradient, thereby producing a photocurrent under external zero bias.

[0171] PV detectors convert light signals into electrical signals based on the photovoltaic effect. A PV detector has a PN junction formed by P-type doped material (P-region) and N-type doped material (N-region). When the photon energy is greater than the semiconductor bandgap, electron-hole pairs are generated in or near the PN junction. The built-in electric field in the junction pushes the electrons and holes in opposite directions (N-region and P-region), thus generating a photovoltage or photocurrent across the PV detector. Photodetector 2 is driven by its built-in electric field, is self-powered, and has a fast response speed.

[0172] In one embodiment, the PV detection device includes a photodetector array composed of transition metal chalcogenides (TMDs) and black phosphorus. This material combination can cover conventional test wavelengths.

[0173] PC detectors convert optical signals into electrical signals based on the photoconductivity effect. When photons strike a semiconductor material, if their energy exceeds its band gap, they excite electron-hole pairs (photogenerated carriers). These additional carriers significantly increase the material's conductivity. Under an applied bias voltage, this increase in conductivity translates into an increase in current within the circuit. Unlike self-powered PTE and PV detectors, PC detectors require an external power supply to provide a bias voltage and create a current path.

[0174] In one embodiment, the material of the PC detector is a narrow bandgap semiconductor, which is selected from one or more combinations of graphene, telluride, and oxide.

[0175] In summary, by setting at least three differential output channels and at least one summation output channel on the photodetector device 2, this invention can detect the light intensity and polarization information of six polarization channels as well as the total light intensity, and output bipolar (positive / negative) electrical signals, realizing a direct mapping from optical polarization information to electrical logic symbols, and providing a foundation for reconstructing the full Stokes parameters.

[0176] The present invention provides a photoelectric detection system for reconstructing the full Stokes vector, which includes a polarization device and a photoelectric detection device 2 located downstream of the polarization device.

[0177] The polarization device splits the incident light and encodes its phase / amplitude, coupling beams from different polarization channels to the six polarization channels mentioned above: 0° / 90° linear polarization channel, ±45° linear polarization channel, and left / right circular polarization channel.

[0178] The photodetector 2 has six detection units 20 corresponding one-to-one with the six polarization channels. The six detection units 20 are connected to form three differential output channels with a differential response mechanism and one summation output channel with a summation response mechanism. The three differential output channels map optically orthogonal polarization states (such as 0° and 90°, 45° and 135°, left-handed and right-handed polarization) into electrically symmetrical differential signals.

[0179] The photodetector 2 being located downstream of the polarization device means that the photodetector 2 is located downstream of the polarization device in the direction of light transmission, and the photodetector 2 can receive the outgoing light from the polarization device.

[0180] This invention achieves synchronous detection of all Stokes parameters by cascading polarization devices and photodetectors, directly converting optical signals into polarized electrical signals, thus providing the possibility of realizing polarization logic control at the electrical signal level.

[0181] Specifically, the incident light is decomposed into six independent polarization channels in one step by a polarization device, and combined with the parallel differential and summation readout capabilities of photodetector 2, a signed bipolar photocurrent is output. On the one hand, during differential amplification, common-mode signals caused by interference such as laser power fluctuations and ambient temperature drift can be effectively canceled, thereby significantly improving the signal-to-noise ratio and measurement stability. On the other hand, the differential response naturally generates bipolar (positive / negative) analog signals, which perfectly match the mathematical characteristics of Stokes parameters S1, S2, and S3. This allows for direct encoding of the positive and negative characteristics of the Stokes components in the analog domain, achieving a direct physical mapping from optical polarization state to electrical logic state. This enables synchronous capture and real-time calculation of the four Stokes parameters, providing a hardware foundation for subsequent analog calculations or logic processing. Furthermore, the signed electrical signals provided by the differential output channels can be directly used as inputs for logic control, making the photodetector itself an analog preprocessing unit. It can directly perform feature extraction (such as edge detection) or implement polarization-encoded neural network convolution at the sensing end before the data is digitized.

[0182] This architecture of "parallel optical decomposition at the front end and coordinated electrical processing at the back end" fundamentally eliminates measurement errors in dynamic scenarios, laying the hardware foundation for high-speed polarization imaging and sensing. The polarization device and photodetector work together to form a low condition number near-diagonal system response matrix, ensuring high-fidelity full Stokes reconstruction; at the same time, it has sign-fidelity differential readout and high bandwidth response (up to 160 GHz), providing the hardware foundation for high-speed binary / ternary logic operations.

[0183] This invention enables on-chip electrical decoding of complete Stokes parameters. A positive Stokes parameter is denoted as +1; a negative Stokes parameter is denoted as 0; and a negative Stokes parameter is denoted as -1. Based on this, a reconfigurable differential readout structure can be established, allowing different polarization states to be directly mapped to polarized electrical signals (+1 / 0 / -1).

[0184] This invention breaks away from the reliance on optical components and external modulation devices in traditional polarization measurements, achieving direct electrical output of all Stokes parameters (S0, S1, S2, S3) at the device level. By switching between differential and additive response mechanisms, it not only possesses on-chip real-time polarization resolution capabilities but also provides signed polarization electrical signals that can be used for logical judgments. Compared with traditional polarization measurement schemes, it significantly improves integration, response speed, and polarization resolution, achieving a compact, high-speed, stable, and low-power electrical polarization sensing solution.

[0185] As described above, polarization devices and photodetectors can be cascaded together using any set of polarization devices and photodetectors corresponding to the arrangement of the focal point and the detection unit 20, which will not be elaborated further here.

[0186] The photodetector 2 of the present invention operates in the infrared band, including but not limited to 850nm, 940nm and 980nm.

[0187] This invention also provides a method for reconstructing the full Stokes vector, comprising the following steps: splitting the incident light and encoding its phase / amplitude using a polarization device to focus beams from different polarization channels onto the aforementioned six polarization channels; activating three differential output channels to obtain the light intensity of the light spots on the 0° / 90° linear polarization channel, the 45° / 135° linear polarization channel, and the right / left circular polarization channel; and activating the summation output channel to obtain the total light intensity.

[0188] The three differential output channels can be activated individually to prevent measurement interference from other channels. They can also be activated simultaneously to improve measurement speed and data consistency.

[0189] The full Stokes vector is reconstructed from the acquired light intensity signal in any of the following ways: , , .

[0190] Among them, I x I y I 45 I 135 I R I L The light intensities of the light spots on the 0° linear polarization channel, 90° linear polarization channel, 45° linear polarization channel, 135° linear polarization channel, right-hand circular polarization channel, and left-hand circular polarization channel are respectively.

[0191] S1 describes the degree to which light is linearly polarized relative to the vertical direction (90°, y) towards the horizontal direction (0°, x), describing the intensity difference between light linearly polarized in the horizontal direction (0°) and the vertical direction (90°). S1 = I0 - I 90 A positive S1 value indicates that the light is more linearly polarized in the horizontal direction; a negative S1 value indicates that the light is more linearly polarized in the vertical direction; and S1 = 0 indicates that the light has no preference in either the horizontal or vertical direction.

[0192] S² describes the intensity difference of light between linearly polarized at 45° and 135° (-45°), and describes the degree to which light is linearly polarized towards 45° relative to 135°. S² = I 45 -I 135 A positive S2 value indicates that the light is more polarized at 45°; a negative S2 value indicates that the light is more polarized at 135°. An S2 value of 0 indicates that the light has no polarization in the ±45° direction.

[0193] S3 describes the intensity difference of light between right-handed and left-handed circularly polarized light, and describes the degree to which light is biased towards right-handed circular polarization relative to left-handed circular polarization. S3 = I R -I L A positive S3 value indicates that the light is more right-handed circularly polarized; a negative S3 value indicates that the light is more left-handed circularly polarized. An S3 value of 0 indicates that the light has no circular polarization characteristics.

[0194] S0 describes the total intensity of the incident light, that is, the sum of the intensities of all light in all polarization directions, S0=I 总 .

[0195] In some implementations, the sum of light intensities on the 45° and 135° linear polarization channels is selected and reconstructed as follows: .

[0196] In other embodiments, the sum of light intensities on the 0° and 90° linear polarization channels can also be selected and reconstructed as follows: .

[0197] In other embodiments, the sum of the light intensities on the right-handed and 9-handed circularly polarized channels can also be selected and reconstructed as follows: .

[0198] In one specific embodiment, channels A, B, and C employ a differential structure, and channel D is connected in parallel to channel A via electrodes. Channel A is activated to decode the S2 vector polarization component. Channel B is activated to decode the S1 vector polarization component. Channel C is activated to decode the S3 vector polarization component. Channels B and C are closed, and channel A is activated. Channel D is reconstructed via external circuitry, aligning the photocurrent polarities of the two detection units 20 in channel A to achieve an additive response and decode the S0 vector polarization component.

[0199] The following explanation uses the cascaded connection of superlens 1 and a PTE detector as an example. During readout, the light spot focused by the polarizing device (e.g., superlens 1) can be defined as an intensity vector:

[0200] When the six focused light spots corresponding to the polarization vector components illuminate the photodetector, each light spot triggers a local photocurrent in its corresponding unit. The electrical output Y in different modes can be controlled by mode-specific operators. The unified representation is:

[0201] It should be noted that the structure of this operator is not arbitrarily defined, but rather originates from the intrinsic physical mapping of the PTE detection device's operating mechanism. Therefore, as... Figure 19 As shown, this operator exhibits different mathematical forms in different operating modes. In Mode I, the difference operator... :

[0202] Three contrast signals are generated These correspond to S1, S2, and S3 respectively, and are directly presented as the differential signals observed by channels A–C:

[0203] The obtained components are proportional to the Stokes parameters S1, S2, and S3, and because they satisfy... It inherently suppresses common-mode noise (such as uniform background interference like laser power fluctuations) due to its differential weighting characteristics.

[0204] In Mode II (addition mode, channel D), only the outermost 20 pairs of detector units are used, and the operator is...

[0205] This operator-based framework provides a compact, real-time model for polarization readout, enabling the direct interpretation of experimentally measured photocurrent signals via Stokes basis.

[0206]

[0207] Please refer to Figures 19 to 22 The superlens 1 and PTE photodetector 2 are cascaded to achieve full Stokes reconstruction and ultrafast performance. The six polarization channels of the superlens 1 are focused along a straight line. Channel A of the photodetector 2 wraps around channel B, channel B wraps around channel C, and channel D is connected in parallel to channel A via an electrode.

[0208] Please refer to Figure 19 The focal points of the six polarization channels constitute the intensity vector. Through electrical output Implement the mapping. Difference operator in Mode I. Generate three sets of contrast signals proportional to S2, S1, and S3. Summation operator in Mode II. Given S0=I 135 +I 45 .

[0209] Please refer to Figure 20 The present invention system measures the photocurrent output of all output channels (A, B, C and D) under different polarization states.

[0210] Please refer to Figure 20In section (i), the differential output channel exhibits bipolar output characteristics, with its signal varying sinusoidally with the incident polarization angle θ (corresponding polarization states are indicated at the top), thus enabling the analysis of vector information. The electrical signal of channel D explicitly depends on the incident light power; for the effect of the additive mode, please refer to [reference needed]. Figure 22 (a) In this paper, the complete Stokes vector is reconstructed by associating each Stokes parameter with the corresponding output channel. Channel B has a selective response to orthogonal linear polarization (0° linear polarization, 90° linear polarization), and channel A has a selective response to diagonal linear polarization (45° linear polarization, 135° linear polarization).

[0211] Specifically, channels A and B vary sinusoidally with the linear polarization angle θ, with channel A corresponding to the S2 component and channel B corresponding to the S1 component; channel C varies with the quarter-wave plate angle and corresponds to the S3 component; and channel D is proportional to the incident power and corresponds to the S0 component.

[0212] The present invention further tested the device's response under left / right circular polarization modulation. The incident beam passes sequentially through a fixed half-wave plate and a rotatable quarter-wave plate, and is then focused by a superlens 1 onto the photodetector array. Figure 20 The lower left subplot (i) shows the variation of the normalized photocurrent PC with the quarter-wave plate angle φ.

[0213] The incident polarization angle is encoded in a dual-period feature vector formed by channels A and B. To resolve this angle, this invention employs a lightweight fully connected neural network for angle regression calculation. Figure 20 As shown in (ii), the incident angle is decoded using the two-dimensional feature vectors of channels A and B. The network receives two-dimensional input vectors within the range of 0°–180°, and its structure includes three hidden layers and a linear output unit. After 300 rounds of training, the mean absolute error on the test set converges to approximately 0.22°. This accurate polarization reconstruction directly benefits from the ability of the polarity-sensitive differential design to suppress common-mode noise. Figure 22 (a)-(b)), this characteristic is usually difficult to achieve in isotropic semiconductor devices.

[0214] Please refer to Figure 20 As shown in (iii), the measured and predicted Poincaré sphere full Stokes parameter distribution at a wavelength of 940 nm demonstrates the broadband operating characteristics. Figure 20 The upper subplot of (iii) compares the distribution of predicted and measured values ​​on the Poincaré sphere at typical angles (0°–180°); Figure 20 The lower subplot of (iii) shows a comparison between the measured and predicted values ​​of the circular polarization parameter S3. Similar results were obtained at wavelengths of 850 nm and 980 nm (see reference). Figure 22 (d) highlights the system's broadband capabilities.

[0215] Please refer to Figure 21 As shown, pump-probe characteristic tests reveal that photodetector 2 exhibits a picosecond response time of 6.1 ps, corresponding to an operating bandwidth exceeding 160 GHz. Time-resolved measurements further confirm its pulse resolution and signal reproducibility, making the device suitable for ultrafast sensing and high-speed optical logic applications. Stable switching of the normalized photocurrent validates its pulse resolution, low noise / crosstalk characteristics, and robust multi-channel readout performance.

[0216] Figure 22 The multi-channel stability, orthogonal response, and broadband performance of the photodetector are illustrated. (a) shows the photocurrent timing signal of a representative summing channel (channel D). The photocurrent variation with the linear polarization angle θ exhibits a sinusoidal response with an approximately 90° phase difference, forming a "double-period" eigenvector used for angle regression and vector information analysis, demonstrating a stable baseline and long-term operational reliability. (b) Under non-differential conditions (before differential operation), a significant repetitive frequency noise signal is visible. This signal originates from factors such as vibration in the test environment. (c) Under differential conditions (after differential operation), the photocurrent timing signal of a typical differential output channel with fixed incident polarization shows only minor changes despite slight input power fluctuations, demonstrating excellent common-mode noise suppression capabilities. (d) The Stokes parameter distribution reconstructed on a Poincaré sphere at wavelengths of 850 nm and 980 nm is illustrated, verifying the broadband operating characteristics and robust vector selectivity from the near-infrared (NIR) to mid-infrared (MIR) bands.

[0217] This invention performs polarization beam splitting at the device level and directly carries vector information in a signed differential manner, avoiding the bandwidth and power consumption overhead of the "intensity-numerical solution" link from the source, while significantly improving robustness under weak signal and complex lighting conditions.

[0218] This invention preferably operates the readout link in zero / low bias mode, which, in conjunction with differential sampling and reference, consciously suppresses common-mode disturbances and low-frequency drift. Guard rings and shielding are used on the array traces to reduce channel crosstalk. Pixels can be arrayed in one-dimensional or two-dimensional configurations and are compatible with conventional readout circuits, enabling full Stokes area array imaging and high-speed scanning.

[0219] This invention also provides a logic controller and image processing system based on full Stokes polarization modulation, constituting an all-optical control logic system. The logic controller includes a polarization device and a photodetector 2.

[0220] The polarization device splits the incident light and encodes its phase / amplitude, coupling beams from different polarization channels to the aforementioned six polarization channels.

[0221] The photodetector 2 is located downstream of the polarization device, and the photodetector 2 has channels A, B, and C. Channel A has two detection units 20 corresponding to the focal points of the 45° linear polarization channel and the 135° linear polarization channel, respectively; channel B has two detection units 20 corresponding to the focal points of the 0° linear polarization channel and the 90° linear polarization channel, respectively; and channel C has two detection units 20 corresponding to the focal points of the right-hand circular polarization channel and the left-hand circular polarization channel, respectively. Channels A, B, and C are differential output channels.

[0222] By selectively activating at least two of channels A, B, and C, the photodetector 2 directly converts the acquired optical signal into a logic output. This invention converts the polarization response into logic signal units and, based on fully Stokes-state reconstructed optical logic gates (binary or ternary logic gates), realizes a polarization-driven all-optical control logic system.

[0223] This invention constructs a differential response mechanism through symmetrical photodetector units 20, enabling the electrical signals generated under excitation in different polarization directions to have controllable polarity (positive / negative) and amplitude, and the response requires no external bias. This polarity can naturally be used as a carrier of logic levels, supporting polymorphic digital encoding including binary (0 / 1) and ternary (+1 / 0 / -1), forming the basic elements of polarization logic. Simultaneously, the differential structure can suppress background noise and temperature drift, improving intrinsic anti-interference capability. This mechanism is not only applicable to logic calculations but can also be used to construct "activation function"-like nodes, analog adders, or difference filter structures in neural networks, possessing photoelectric analog computing capabilities.

[0224] Based on the aforementioned photoelectric detection system and method for reconstructing the full Stokes vector, this invention introduces an amplitude threshold to define the 0 state, while the positive and negative polarities of S1, S2, and S3 are mapped to +1 and -1, respectively. This achieves a direct and high-speed conversion from optical polarization to electrical logic, forming a binary / ternary hybrid encoding. Therefore, the photoelectric detection device 2 based on full Stokes polarization control supports all seven basic Boolean logic gates and composite logic operations, laying the physical foundation for the polarization-logic-learning integrated chip architecture.

[0225] As described above, polarization devices and photodetectors can be cascaded together using any set of polarization devices and photodetectors corresponding to the arrangement of the focal point and the detection unit 20, which will not be elaborated further here.

[0226] The following section uses the cascade of superlens 1 and PET detector as an example to provide a detailed explanation of the polarization-driven all-optical control logic system.

[0227] This invention integrates the polarization beam-splitting capability of a metasurface with the bipolar differential response of a photodetector array, thereby realizing a shift from passive polarization sensing to active logic modulation driven by programmable optical input. By selectively modulating specific Stokes components, the corresponding detection channels 21 can be independently addressed, each generating a signed bipolar electrical output. This direction-selective, polarity-preserving response characteristic lays a solid physical foundation for polarization-coded binary and ternary logic operations.

[0228] To realize optical logic inputs based on Stokes state modulation, this invention experimentally verifies the orthogonality of the output polarization states after metasurface projection, as well as the channel isolation characteristics of the detector array for each Stokes parameter. Three independent optical logic inputs are encoded using precisely defined polarization states and spatially routed to their corresponding graphene photodetectors.

[0229] Please refer to Figure 23 The diagram shows polarization-encoded binary-ternary logic, reconfigurable logic gates, and optical encryption applications. Figure 23 (a) illustrates a reconfigurable optoelectronic logic platform, in which Stokes-coded optical inputs are fed to a photodetector device 2 (e.g., a graphene-Au detector array) via a superlens 1; external wiring is configured to configure the logic output mode.

[0230] Specifically, such as Figure 23 As shown in Figure (b), the polarization-selective response of the three logic inputs is defined. Channel A is activated by linearly polarized light at θ=135°, defined as IN 1. Channel B is activated by linearly polarized light at θ=90°, defined as IN 2. Channel C is activated by left-handed circularly polarized light, defined as IN 3, and channel C is driven by right-handed circularly polarized light, defined as IN 3R. The bottom mapping of the figure shows the ternary weights assigned to the detector units 20 (pin 1, pin 3, pin 5, pin 6, pin 4, pin 2).

[0231] Linearly polarized light with θ=135° selectively activates channel A. At this time, the differential signals generated by pin 3 and pin 4 of the detector unit have equal amplitudes and opposite polarities, thus canceling each other out and keeping channel B in a non-responsive state. Conversely, linearly polarized light with θ=0° / 180° triggers channel B while suppressing channel A, confirming the mutual exclusion between the two channels and ensuring crosstalk-free characteristics during logic operations.

[0232] The third logic input is defined by circular polarization states. Channel C exhibits a polarity-selective response: right-handed circular polarization produces a positive signal, and left-handed circular polarization produces a negative signal. Thanks to the symmetrical projection of the metasurface and the differential output channel architecture, channels A and B remain silent under circularly polarized illumination, confirming complete vector selectivity. These mappings together establish a fully optical Stokes-modulated three-input logic configuration, where each input is uniquely determined by its polarization basis vector and spatial routing. Notably, the bipolar electrical outputs at the three logic inputs support direct encoding of a ternary logic character set.

[0233] By utilizing three polarization-coded optical inputs and simultaneously controlling the signal polarity and amplitude, this invention achieves integrated implementation of all seven basic dual-input logic gates on a chip. Figure 23 (c) shows a schematic diagram of the extended logic circuit and devices. The operating modes include self-powered, externally biased, and externally extended logic circuits. Wiring examples are provided to implement AND gates / OR gates, XOR gate → NOT gate, OR gate, NAND gate, etc.

[0234] Please refer to Figure 23 and Figure 24 As shown, in different operating modes, the logic controller can output different logic gates.

[0235] Channel A is activated by 45° or 135° linearly polarized light. When 45° or 135° linearly polarized light is in the input state, the input value IN1 is defined as 1; when 45° or 135° linearly polarized light is in the off state, the input value IN1 is defined as 0. Channel B is activated by 0° or 90° linearly polarized light. When 0° or 90° linearly polarized light is in the input state, the input value IN2 is defined as 1; when 0° or 90° linearly polarized light is in the off state, the input value IN2 is defined as 0. Channel C is activated by left- or right-handed circularly polarized light. When left- or right-handed circularly polarized light is in the input state, the input value IN3 is defined as 1; when left- or right-handed circularly polarized light is in the off state, the input value IN3 is defined as 0.

[0236] By controlling any two of the input values ​​IN1, IN2, and IN3 as the input value IN, the photodetector 2 directly converts the obtained optical signal into a binary logic output.

[0237] For example, by controlling the input values ​​IN1 and IN2, photodetector 2 directly converts the acquired optical signal into a binary logic output. Alternatively, by controlling the input values ​​IN1 and IN3, photodetector 2 directly converts the acquired optical signal into a binary logic output. By controlling the input values ​​IN2 and IN3, photodetector 2 directly converts the acquired optical signal into a binary logic output.

[0238] In self-powered mode, by controlling the input values ​​IN1 and IN2, the photodetector 2 directly converts the obtained optical signal into one of the following logic gates: OR, NOT, or XOR.

[0239] In self-powered mode, an OR logic gate can be constructed by superimposing the unidirectional photocurrent responses of any two of channels A, B, and C. Under 0V bias, when the input sequence changes according to IN-00, IN-01, IN-10, and IN-11, the logic controller outputs (0, 1, 1, 1) sequentially, exhibiting standard OR truth table characteristics, with its normalized switching current ratio exceeding one order of magnitude.

[0240] In self-powered mode, NOT logic gates and XOR logic gates can be implemented by extracting the reverse symmetrical current modes of any two of channels A, B, and C.

[0241] At 0V bias, when the input sequence changes according to IN-10 and IN-11, the logic controller outputs (1, 0) in sequence, exhibiting standard NOT truth table characteristics, and its normalized switching current ratio exceeds one order of magnitude.

[0242] Under 0V bias, when the input sequence changes according to IN-00, IN-01, IN-10, IN-11, the logic controller outputs (0, 1, 1, 0) in sequence, exhibiting standard XOR truth table characteristics, and its normalized switching current ratio is about one order of magnitude.

[0243] In external bias mode, a bias voltage is applied, and by controlling the input value IN, the photodetector 2 directly converts the obtained optical signal into one of XNOR logic gates or ANAD logic gates.

[0244] In external bias mode, XNOR logic gates can be constructed. In V bias =-V PC Under bias, when the input sequence changes according to IN-00, IN-01, IN-10, IN-11, the logic controller outputs (1, 0, 0, 1) in sequence, exhibiting the standard XNOR truth table characteristics.

[0245] In external bias mode, XAND logic gates can be constructed. In V bias =-2V PC Under bias, when the input sequence changes according to IN-00, IN-01, IN-10, IN-11, the logic controller outputs (1, 1, 1, 0) in sequence, exhibiting the standard XAND truth table characteristics.

[0246] The logic controller has an external extended logic circuit mode, introducing an NMOSFET to control any two of the input values ​​IN1, IN2, and IN3 as the input value IN. The photodetector 2 directly converts the obtained optical signal into an AND logic gate and a NOR logic gate.

[0247] In the external extended logic circuit mode, a depletion-type NMOSFET is introduced. The photocurrent of one of channels A, B, and C is connected to the source of the NMOSFET, and the photocurrent of the other of channels A, B, and C is connected to the gate of the NMOSFET. When the input sequence changes according to IN-00, IN-01, IN-10, and IN-11, the drain of the NMOSFET outputs (0, 0, 0, 1) in sequence, exhibiting standard AND truth table characteristics.

[0248] In one embodiment, the photocurrent of channel A is connected to the source of the NMOSFET, and the photocurrent of channel B is connected to the gate of the NMOSFET. When the input sequence changes according to IN-00, IN-01, IN-10, IN-11, the drain of the NMOSFET outputs (0, 0, 0, 1) in sequence, exhibiting standard AND truth table characteristics.

[0249] In the external extended logic circuit mode, an enhancement-mode NMOSFET is introduced, and a positive voltage V is applied between the drain and source. D The unidirectional photocurrent response of any two of channels A, B, and C is superimposed and connected to the gate of the NMOSFET. When the input sequence changes according to IN-00, IN-01, IN-10, IN-11, the drain of the NMOSFET outputs (1, 0, 0, 0) in sequence, exhibiting the standard NOR truth table characteristics.

[0250] In one embodiment, under external extended logic circuit mode, an enhancement-mode NMOSFET is introduced, and a positive voltage V is applied between its drain and source. D The unidirectional photocurrent response of channels A and B is superimposed and connected to the gate of the NMOSFET. When the input sequence changes according to IN-00, IN-01, IN-10, IN-11, the drain of the NMOSFET outputs (1, 0, 0, 0) in sequence, exhibiting the standard NOR truth table characteristics.

[0251] As can be seen, the present invention is based on a logic controller with full Stokes polarization modulation, which can cover all seven dual-input Boolean logic functions without any physical structure modification, and can achieve switching between different logic gates simply by reconfiguring the external circuit connection.

[0252] To extend the platform's functionality to multi-input and multi-level control systems, any two of channels A, B, and C can be used to construct a single logic gate. This single logic gate can be any one of the following: OR, NOT, XOR, XNOR, ANAD, AND, or NOR. This single logic gate can be superimposed with the unidirectional photocurrent of another channel (Channel A, B, or C) and input to any other single logic gate to form a composite logic gate.

[0253] In one embodiment, a single logic gate is constructed based on channel A and channel B. The single logic gate is superimposed with the unidirectional photocurrent of the channel and then input into another arbitrary single logic gate to form a composite logic gate.

[0254] like Figure 23 As shown in (d), this is a reversible two-stage gate cascade system with affine mapping on GF(2). When the parameters meet specific conditions, the cascade constitutes a permutation operation. In this invention, an all-optical XOR gate is superimposed with the unidirectional photocurrent of channel C and then input into an XNOR gate for cascading to construct a composite logic gate.

[0255] Figure 23 (e) shows the circuit diagram, truth table, and normalized photocurrent time-domain output (logic output characteristics) of the cascaded XOR-NOR compound logic gates implemented in the experiment. The XOR gate is biased at 0V, and the XNOR gate at V... bias =-V PC Under bias, when the input sequence changes according to IN-000, IN-001, IN-010, IN-011, IN-100, IN-101, IN-110, IN-111, the compound logic gate outputs (1, 0, 1, 0, 1, 1, 0) in sequence.

[0256] Figure 23 The diagram in (f) illustrates the all-optical encryption-decryption process: characters are encoded by polarization and then encrypted and decrypted through cascaded logic gates. Figure 23 Figure (g) illustrates the normalized photocurrent trajectory measured during the transmission of the character sequence, verifying the correct encoding and decoding functions and robust multi-channel operation performance.

[0257] To formally describe this reversible two-stage system, this invention models each stage of operation as an affine mapping on GF(2).

[0258]

[0259] The mapping is invertible if and only if the coefficient a = 1. Its two-level combination can be expressed as:

[0260] When the coefficient At that time, the overall transformation constitutes A permutation on the bit; the decryption process is performed in reverse order and is equivalent to a single masking operation at the bit level. Using this cascaded system, this invention implements an optical encryption-decryption process: input characters (such as "N", "K", "U") are encoded using binary polarization states and sequentially input to ports IN1–IN3; the cascaded XOR-XNOR logic chain generates ciphertext, which is then decrypted through reverse mapping. Figure 23 (f–g). This process validates secure, deterministic all-optical logic operations based entirely on polarization modulation and passive photodetection.

[0261] Based on the above scheme, this invention can construct a binary control scheme. On the basis of differential readout, two or more differential output channels are selected, and a binary signal is obtained through threshold determination. By adding / subtracting and connecting the channels through external circuitry, basic logic gates (such as AND, OR, NOT, XOR) and their cascades can be implemented for polarization-based triggering and discrimination.

[0262] The switching of logic functions is accomplished through external connections and polarity configuration, and in principle, it does not involve any changes to the layout of the superlens 1 or the graphene active region. Since each differential output channel corresponds to a different polarization semantic, the angle or ellipticity of the incident polarization can directly correspond to the valid / invalid state of the logic input, thereby realizing logic operation on the sensor side directly driven by the polarization state.

[0263] Multi-valued logic (especially ternary logic) offers a powerful alternative to binary encoding. Ternary computation increases the information density of each symbol to approximately log₂₃ ≈ 1.585 bits, reducing logic circuit depth and enabling the implementation of complex arithmetic functions in a more compact manner. However, the difficulties in designing tri-state hardware and the lack of mature software and manufacturing ecosystems continue to limit its practical application.

[0264] This invention utilizes the inherent bipolar electrical response of the detector to directly derive the ternary logic state (-1, 0, +1) from the polarization-resolved photocurrent encoded by Stokes vectors. Therefore, the polarization device not only serves as a signal carrier for the logic layer but also becomes a multidimensional encoding mode, particularly suitable for data processing scenarios with rich spatial gradients and structural complexity.

[0265] Activate at least two of channels A, B, and C, define the amplitude threshold as the 0 state, map the positive and negative polarities of S1, S2, and S3 to +1 and -1 respectively, and derive the ternary logic state (-1, 0, +1) from the polarization-resolved photocurrent encoded by Stokes vectors.

[0266] Based on this, ternary logic control is constructed without the need to manufacture complex three-state transistors. Instead, it utilizes the physical characteristics of optical polarization and the differential response of the detector to directly realize the generation and operation of ternary states in the analog domain, significantly reducing the hardware barriers of ternary logic. This provides a unique control platform for realizing ultra-high-speed, low-power "physical native ternary computation" for specific applications (such as real-time polarization image processing).

[0267] When channel A is activated, 135° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN1 set to +1; when 0° or 90° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN1 set to 0; when 45° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN1 set to -1.

[0268] When channel B is activated, 90° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN2 set to +1; when 45° or 135° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN2 set to 0; when 0° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN2 set to -1.

[0269] When channel C is activated, left-hand circularly polarized light generates a positive signal, and the output value IN3 is set to +1; when it is deactivated, a zero signal is generated, and the output value IN3 is set to 0; right-hand circularly polarized light generates a negative signal, and the output value IN3 is set to -1.

[0270] By controlling any two of the input values ​​IN1, IN2, and IN3, the photodetector 2 directly converts the obtained optical signal into a ternary logic output.

[0271] Ternary logic outputs include ternary loop gates, ternary AND gates (MIN gates), ternary OR gates (MAX gates), ternary inverters, etc.

[0272] Based on the above scheme, this invention can construct a ternary control scheme. While preserving the differential polarity, the output is divided into three intervals, corresponding to three discrete values: negative, zero, and positive. A threshold is used to determine the discrete state, and basic operations such as ternary addition, comparison, and selection are implemented through external connections. Switching between binary and ternary modes is achieved by adjusting the judgment threshold and the connection method, without changing the optical and electrical structure of the front end.

[0273] Unlike existing polarization measurement devices that only operate at the optical detection level, this invention directly converts polarization information into digital logic signals, constructing an all-optical input, electrically modulated, and reconfigurable polarization logic gate system. This system implements basic logic operations such as AND, OR, NOT, and XOR, as well as ternary addition. Furthermore, the polarization excitation signals used in this system are mutually orthogonal, enabling complete mutual exclusion and crosstalk-free control between input channels, significantly enhancing logic reliability and coding capability.

[0274] In addition, the PTE detector (graphene-metal photothermal detector unit 20) used in the device has self-powered capability, requires no external bias, has a picosecond-level response time (6.1ps), and corresponds to a theoretical bandwidth of over 160GHz, which can meet the needs of next-generation high-speed optical computing and communication systems.

[0275] This invention also provides an imaging system based on full Stokes polarization modulation, which converts the polarization response into logic signal units, constructs a polarization convolution model for image processing, and realizes image edge detection and intelligent recognition directly at the detector level.

[0276] An imaging system based on full Stokes polarization control includes a laser source, polarization devices, and photodetectors.

[0277] A laser source is used to output a beam of light with a predetermined polarization state. For example, it can output a beam with a 0° / 90° linear polarization state, a beam with a 45° / 135° linear polarization state, or a beam with a right-hand / left-hand circular polarization state.

[0278] The polarizing device and the laser source are spaced apart to facilitate placing the object to be imaged between them. The polarizing device splits the incident light and encodes its phase / amplitude, coupling beams from different polarization channels to the six polarization channels mentioned above.

[0279] The photodetector 2 is located downstream of the polarization device. The photodetector 2 has channels A, B, and C. Channel A has two detection units 20 corresponding to the focal points of the 135° linear polarization channel and the 45° linear polarization channel, respectively; channel B has two detection units 20 corresponding to the focal points of the 90° linear polarization channel and the 0° linear polarization channel, respectively; and channel C has two detection units 20 corresponding to the focal points of the left-handed circular polarization channel and the right-handed circular polarization channel, respectively.

[0280] Channels A, B, and C are differential output channels. At least one of channels A, B, and C is selectively activated. The photodetector 2 converts the acquired optical signal into an electrical signal, which is then converted into image grayscale through depolarization imaging and polarization-coded ternary convolutional neural network imaging.

[0281] The polarization device and the photodetector 2 can be combined using any of the above methods, which will not be elaborated further here.

[0282] With its ultra-small size (core size approximately 200-500μm), this platform can be integrated with optical fibers or interventional catheters. Combining depolarization imaging with polarization-coded ternary convolutional neural networks, it provides a proximal, low-redundancy intelligent sensing and computing solution for interventional disordered tissue detection and boundary delineation.

[0283] This invention introduces the concept of "polarization convolution" and applies the detector response matrix to image processing tasks, enabling edge enhancement and directional filtering directly within the detection unit 20. This system can perform boundary recognition and semantic segmentation on disordered biological tissues (such as human brain white matter), demonstrating broad application prospects in biomedical imaging, intelligent photon sensing, and machine vision processing. Overall, this invention realizes an integrated optoelectronic platform from "polarization detection" to "logic operation" to "intelligent image processing," laying the physical foundation for a truly on-chip polarization intelligent system.

[0284] Disordered human tissues (such as brain white matter) often exhibit a stronger depolarization effect on incident polarized light due to their disordered microstructure, altered collagen fiber orientation, and enhanced multiple scattering: linear polarization contrast decreases, polarization angles become more unstable in local regions, and polarization characteristics show significant abrupt changes at tissue boundaries. Simultaneously, surface specular reflection typically maintains high polarization, masking the true tissue contrast. Therefore, imaging and discrimination procedures based on "depolarization rate" and related polarization characterization quantities have become an effective approach to distinguish disordered human tissues (such as disordered brain tissue) from surrounding healthy tissues and to highlight lesion boundaries.

[0285] This invention provides a single acquisition of all Stokes data at the device level, simultaneously obtaining intensity, linear polarization, and circular polarization information at the pixel level, providing direct input for constructing characterization parameters such as depolarization rate. Based on this, the system first performs polarization-oriented preprocessing on-chip: utilizing the difference between linear and circular polarization channels to suppress specular highlights and uniform backgrounds, and leveraging the spatial variations in polarization angle and polarization contrast to highlight texture and boundary abrupt changes, forming a multi-channel polarization map as input for subsequent algorithms. This preprocessing primarily uses simple, deterministic operators to ensure stable latency and power consumption performance at the edge.

[0286] To facilitate integration with clinical or engineering applications, the system simultaneously performs polarization channel consistency correction and intensity normalization during the calibration phase; during the inference phase, it operates with a fixed model and thresholds to ensure the reproducibility and interpretability of the results.

[0287] Based on the robustness of the full Stokes information and differential readout provided by the device at the front end, the overall process can maintain high resolvability under complex lighting and surface reflection conditions, and is suitable for applications such as end-side imaging and medical auxiliary judgment.

[0288] Based on the vector encoding capability of the full Stokes vector, this invention constructs a ternary logic system by discretizing the signed photocurrent into three logic levels: +1 (positive polarity), 0 (zero level), and -1 (negative polarity). Figure 25 (As shown in a). These levels are directly mapped from the bipolar optical responses of channels A–C under linearly and circularly polarized illumination, thereby defining mutually independent and orthogonal logical foundations.

[0289] The amplitude threshold is defined as the 0 state, and the positive and negative polarities of S1, S2, and S3 are mapped to +1 and -1, respectively. The ternary logic state (-1, 0, +1) is derived from the polarization-resolved photocurrent encoded by Stokes vector. Based on the ternary logic state, the image grayscale is output through a polarization-encoded ternary convolutional neural network.

[0290] For channel A (primarily responding to S0 and S2 components), positive polarity corresponds to 135° linear polarization, zero level corresponds to 0° or 180° polarization, and negative polarity corresponds to the dark state. This three-state encoding mechanism not only supports multi-valued calculations but also compactly represents signal amplitude and polarization direction information.

[0291] Figure 25 The ternary "adder" gate shown in b–c can quantize and sum any two Stokes-defined input signals and exhibit stable state transition characteristics between discrete logic levels.

[0292] Based on the ternary optoelectronic logic framework, this invention further extends the device to the field of polarization convolution image processing. For example... Figure 26 As shown, performing a ternary convolution operation on the depolarization rate map can transform the microstructural disorder caused by multiple scattering into stable and extractable edge / texture features; by adjusting the convolution weights, it can be adapted to tasks such as edge detection and image sharpening.

[0293] This invention supports a single-channel operating mode, avoiding channel crosstalk at the hardware level.

[0294] Taking channel A as an example, when 135° linearly polarized light is input to the polarization device, the photodetector outputs a positive electrical signal, and the output value IN1 is set to +1; when 0° or 90° linearly polarized light is input to the polarization device, the photodetector outputs a 0 electrical signal, and the output value IN1 is set to 0; when 45° linearly polarized light is input to the polarization device, the photodetector outputs a negative electrical signal, and the output value IN1 is set to -1, thus achieving a direct mapping from polarization orientation to logic state. Based on the ternary logic state (-1, 0, +1), the image grayscale is output through a polarization-coded ternary convolutional neural network.

[0295] Taking channel B as an example, when 90° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is positive, and the output value IN2 is set to +1; when 45° or 135° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, and the output value IN2 is set to 0; when 0° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, and the output value IN2 is set to -1. The image grayscale is output through a polarization-coded ternary convolutional neural network based on the ternary logic states (-1, 0, +1).

[0296] Taking channel C as an example, when left-hand circularly polarized light is input to the polarization device, the electrical signal output by the photodetector is positive, and the output value IN3 is set to +1; when left / right circularly polarized light is turned off, a zero signal is generated, and the output value IN3 is set to 0; when right-hand circularly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, and the output value IN3 is set to -1. The image grayscale is output through a polarization-coded ternary convolutional neural network based on the ternary logic states (-1, 0, +1).

[0297] It should be noted that in a symmetrical set of polarized beams (±45°, 0° / 90°, left / right rotation), one beam is positive and the other is negative. For example: Taking channel A as an example, 135° linearly polarized light can be input into the polarization device, and the electrical signal output by the photodetector is negative, with the output value IN 1 set to -1; 45° linearly polarized light can be input into the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN 1 set to +1.

[0298] Taking channel B as an example, 90° linearly polarized light can be input into the polarization device, and the electrical signal output by the photodetector is negative, with the output value IN 1 set to -1; 0° linearly polarized light can be input into the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN 1 set to +1.

[0299] Taking channel C as an example, left-handed circularly polarized light can be input into the polarization device, and the electrical signal output by the photodetector is negative, with the output value IN1 set to -1; right-handed circularly polarized light can be input into the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN1 set to +1.

[0300] By utilizing the bipolar response of channel AC under incident angles of 45° / 90° / 135°, this invention directly generates ternary weights in the sensing layer and reassembles them into Prewitt-like differential convolutions and edge-sharpening convolutions.

[0301] This invention also supports a multi-channel working mode, simultaneously activating channels A, B, and C. By controlling IN1, IN2, and IN3, a ternary logic state (-1, 0, +1) is jointly output. Based on the ternary logic state, the image grayscale is output through a polarization-coded ternary convolutional neural network.

[0302] This invention also provides an imaging method based on a fully Stokes polarization-controlled imaging system, comprising the following steps: a laser source outputs a beam of light with a predetermined polarization state; the polarized beam passes through the object to be detected or enters a polarization device, where the incident light is split and its phase / amplitude encoded by the polarization device, coupling beams from different polarization channels to the aforementioned six polarization channels. The beam output by the polarization device enters a photodetector 2, selectively activating at least one of channels A, B, and C; the photodetector 2 converts the obtained optical signal into an electrical signal, and then converts the electrical signal into image grayscale through depolarization imaging and polarization-encoded ternary convolutional neural network imaging.

[0303] The laser source relies on the aforementioned single-channel and multi-channel modes to output a polarized beam. The imaging method is as described above and will not be repeated here.

[0304] This invention also provides a medical detection device, including an imaging system based on full Stokes polarization modulation and an image processing module. The image processing module outputs medical image information based on the grayscale of the output image.

[0305] like Figure 27 As shown, the depolarized image processed by polarization convolution effectively highlights the tissue edge structure. The upper layer image illustrates the original image and the edge enhancement processing result, while the lower layer image shows the response mapping of the ternary convolution synthesized from multiple polarization bases.

[0306] In glioma boundary delineation, traditional intensity imaging struggles to achieve accurate edge identification due to the invasive growth characteristics and low contrast of disordered tissue. In contrast, polarization-resolved imaging offers greater sensitivity to the microstructural heterogeneity of disordered tissue.

[0307] Taking disordered tissue as an example of tumors, the invasive growth of tumors and the inherently low contrast of traditional intensity imaging methods have long hindered the accurate identification of tumor edges in glial tumor boundary delineation. In contrast, Mueller polarization imaging—especially depolarization contrast imaging—is more sensitive to the microstructural heterogeneity of tumors, significantly improving intraoperative visualization capabilities. Depolarization can distinguish between ordered and disordered tissues and achieve robust and rapid stratification of gray and white matter in the brain. More importantly, this prior information greatly improves sensitivity to the linear phase delay decrease and optical axis orientation coherence loss caused by infiltration, while reducing cross-sample threshold drift and false positive risks. This invention further constructs an application scenario for intravascular polarization imaging: fiber-optic integrated probes acquire polarization-resolved images, and the depolarization rate parameter effectively highlights the spatial variation characteristics of scattering, clearly presenting soft tissue heterogeneity information in the generated structural map. The core size of this polarization-sensitive detector array is only about 150 μm, much smaller than the diameter of the simulated blood vessel and the size of the fiber-optic probe, and can be seamlessly integrated into interventional catheter systems, supporting minimally invasive in vivo real-time high-resolution boundary identification.

[0308] In the data processing flow, the original depolarization rate map is first normalized to a grayscale image, and then edge enhancement is performed using the aforementioned polarization ternary convolution. Thanks to the directional selectivity of the hardware-based ternary convolution and the spatial continuity of the polarization cues, accurate segmentation of the tumor region is achieved, fully demonstrating the technical advantages of the synergistic optimization of polarization detection and deep learning.

[0309] The technology of the present invention will be described in detail below with reference to an embodiment.

[0310] A six-channel graphene photoelectric array (PIN 1–PIN 6) is registered and constructed with a superlens 1; differential mode (Mode I) and additive mode (Mode II) configurations are used; incident light is controlled by polarization state, allowing selective activation of channels A–C. The logic input consists of three polarization states (linear polarization 135°, 0° / 180°, circular polarization); the output current polarity serves as the logic state encoding (P / Z / N). Applications include Prewitt convolution for edge detection images, character encryption / decryption, and disordered organization boundary recognition; it also enables collaborative image semantic segmentation with the U-Net model. Please refer to [link / reference needed] for details. Figure 28 The diagram clearly illustrates the process of image processing and disordered organization detection.

[0311] In summary, this invention designs a polarization device based on full Stokes polarization modulation and a photodetector 2 for detecting full Stokes polarization. The polarization device and photodetector 2 are cascaded to construct a photodetector system and reconstruction method for reconstructing the full Stokes vector, a logic control device based on full Stokes polarization modulation, and an imaging system based on full Stokes polarization modulation. This establishes a chip-level pathway from polarization vector → photoelectric logic → image convolution at the device level.

[0312] It should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This way of describing the specification is only for clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0313] The detailed descriptions listed above are merely specific descriptions of feasible embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. All equivalent embodiments or modifications made without departing from the spirit of the present invention should be included within the scope of protection of the present invention.

Claims

1. An imaging system based on full Stokes polarization modulation, characterized in that, include: Laser source; A polarizing device is disposed at a distance from the laser source. The polarizing device splits the incident light and encodes its phase / amplitude, coupling beams from different polarization channels to six polarization channels. The six polarization channels are: 0° linear polarization channel, 90° linear polarization channel, 45° linear polarization channel, 135° linear polarization channel, right-hand circular polarization channel, and left-hand circular polarization channel. A photodetector device is located downstream of the polarization device. The photodetector device has channels A, B, and C. Channel A has two detection units corresponding to the focal points of the 135° linear polarization channel and the 45° linear polarization channel, respectively. Channel B has two detection units corresponding to the focal points of the 90° linear polarization channel and the 0° linear polarization channel, respectively. Channel C has two detection units corresponding to the focal points of the left-hand circular polarization channel and the right-hand circular polarization channel, respectively. Channel A, channel B, and channel C have a differential structure; Among them, at least one of channel A, channel B, and channel C is selectively activated, and the photodetector converts the obtained optical signal into an electrical signal, and then converts the electrical signal into image grayscale through depolarization imaging and polarization-coded ternary convolutional neural network imaging.

2. The imaging system based on full Stokes polarization modulation according to claim 1, characterized in that, The amplitude threshold is defined as the 0 state, and the positive and negative polarities of S1, S2, and S3 are mapped to +1 and -1, respectively. The ternary logic state (-1, 0, +1) is derived from the polarization-resolved photocurrent encoded by Stokes vector. Based on the ternary logic state, the image grayscale is output through a polarization-encoded ternary convolutional neural network.

3. The imaging system based on full Stokes polarization modulation according to claim 1, characterized in that, When channel A is activated, 135° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN1 set to +1; when 0° or 90° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN1 set to 0; when 45° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN1 set to -1. Based on the ternary logic state (-1, 0, +1), the image grayscale is output through a polarization-coded ternary convolutional neural network.

4. The imaging system based on full Stokes polarization modulation according to claim 1, characterized in that, When channel B is activated, 90° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN2 set to +1; when 45° or 135° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN2 set to 0; when 0° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN2 set to -1. Based on the ternary logic state (-1, 0, +1), the image grayscale is output through a polarization-coded ternary convolutional neural network.

5. The imaging system based on full Stokes polarization modulation according to claim 1, characterized in that, When channel C is activated, left-hand circularly polarized light generates a positive signal, and the output value IN3 is set to +1; when it is deactivated, a zero signal is generated, and the output value IN3 is set to 0; right-hand circularly polarized light generates a negative signal, and the output value IN3 is set to -1. Based on the ternary logic state (-1, 0, +1), the grayscale of the image is output through a polarization-encoded ternary convolutional neural network.

6. The imaging system based on full Stokes polarization modulation according to claim 1, characterized in that, When channel A is activated, 135° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN1 set to +1; when 0° or 90° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN1 set to 0; when 45° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN1 set to -1. Based on the ternary logic state (-1, 0, +1), the image grayscale is output through a polarization-coded ternary convolutional neural network. When channel B is activated, 90° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN2 set to +1; when 45° or 135° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN2 set to 0; when 0° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN2 set to -1. When channel C is activated, left-hand circularly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN3 set to +1; when left / right circularly polarized light is turned off, a zero signal is generated, and the output value IN3 is set to 0; when right-hand circularly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, and the output value IN3 is set to -1. By controlling IN1, IN2, and IN3, a ternary logic state (-1, 0, +1) is jointly output. Based on the ternary logic state, the image grayscale is output through a polarization-coded ternary convolutional neural network.

7. A medical detection device, characterized in that, The imaging system and image processing module based on full Stokes polarization modulation as described in any one of claims 1 to 6 are included.

8. An imaging method based on the imaging system based on full Stokes polarization modulation as described in any one of claims 1 to 6, characterized in that, Includes the following steps: The laser source outputs a beam with a predetermined polarization state; The polarized beam passes through the object to be detected or enters the polarization device. The polarization device splits the incident light and encodes its phase / amplitude, coupling beams from different polarization channels into six polarization channels. The six polarization channels are: 0° linear polarization channel, 90° linear polarization channel, 45° linear polarization channel, 135° linear polarization channel, right-hand circular polarization channel, and left-hand circular polarization channel. The light beam output from the polarization device enters the photodetector and selectively activates at least one of channels A, B, and C. The photodetector converts the obtained optical signal into an electrical signal, and then converts the electrical signal into image grayscale through depolarization imaging and polarization-coded ternary convolutional neural network imaging.

9. The imaging method according to claim 8, characterized in that, The amplitude threshold is defined as the 0 state, and the positive and negative polarities of S1, S2, and S3 are mapped to +1 and -1, respectively. The ternary logic state (-1, 0, +1) is derived from the polarization-resolved photocurrent encoded by Stokes vectors, and the image grayscale is output through ternary convolution.

10. The imaging system based on full Stokes polarization modulation according to claim 9, characterized in that, When channel A is activated, 135° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN1 set to +1; when 0° or 90° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN1 set to 0; when 45° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN1 set to -1. The image grayscale is then output through ternary convolution. Alternatively, channel B can be activated. When 90° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is positive, and the output value IN2 is set to +1. When 45° or 135° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, and the output value IN2 is set to 0. When 0° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, and the output value IN2 is set to -1. The image grayscale is then output through ternary convolution. Alternatively, channel C can be activated, and left-hand circularly polarized light can be input to the polarization device. The electrical signal output by the photodetector is positive, and the output value IN3 is set to +1. When left-hand / right-hand circularly polarized light is turned off, a zero signal is generated, and the output value IN3 is set to 0. When right-hand circularly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, and the output value IN3 is set to -1. The image grayscale is then output through ternary convolution.

11. The imaging method according to claim 9, characterized in that, When channel A is activated, 135° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN1 set to +1; when 0° or 90° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN1 set to 0; when 45° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN1 set to -1. When channel B is activated, 90° linearly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN2 set to +1; when 45° or 135° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is 0, with the output value IN2 set to 0; when 0° linearly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, with the output value IN2 set to -1. When channel C is activated, left-hand circularly polarized light is input to the polarization device, and the electrical signal output by the photodetector is positive, with the output value IN3 set to +1; when left / right circularly polarized light is turned off, a zero signal is generated, and the output value IN3 is set to 0; when right-hand circularly polarized light is input to the polarization device, the electrical signal output by the photodetector is negative, and the output value IN3 is set to -1. By controlling IN1, IN2, and IN3, a ternary logic state (-1, 0, +1) is output, and the image grayscale is output through ternary convolution.

Images