Self-reference three-phase polarization imaging system and method based on integrated photon neural network

Abstract

The invention discloses a self-reference three-phase polarization imaging system and method based on an integrated photon neural network, and the method comprises the steps: setting a fixed phase difference in a silicon-based photon chip to achieve three-phase projection and single-port reading, outputting a three-phase interference signal at the same time, obtaining all phase and polarization information in one-time imaging to achieve full-polarization measurement, and carrying out the single-port reading of the silicon-based photon chip; according to the invention, the effective resolution ratio is consistent with that of an original detection array, power and temperature drift is eliminated, single-frame acquisition has no time sequence error and motion artifact, a stable and compact single-port self-reference reading mode is realized through a three-phase interference measurement architecture constructed on a chip in combination with channel normalization and rotation correction, and complete Stokes parameters and polarization information can be measured. According to the system provided by the invention, the MZI is adopted, so that the phase bias can be flexibly adjusted, and different bandwidth requirements can be met by adjusting an interference path. The silicon-based photon platform has a broadband response characteristic and can cover a visible wave band, so that the device can keep stable interference performance in a communication wave band.

Description

Technical Field

[0001] This invention belongs to the field of integrated photonics and holographic optical imaging technology, specifically relating to a self-referenced three-phase polarization imaging system and method based on an integrated photonic neural network. Background Technology

[0002] In recent years, three-dimensional coherent imaging has matured, providing spatial distribution and tomographic information of samples. However, this type of imaging technology typically focuses only on light intensity or phase distribution, and its characterization of polarization information of the light field remains insufficient, failing to directly reveal the sample's response in polarization degrees of freedom. Polarization imaging has important applications in remote sensing, biophotonics, and materials science, extracting rich information such as material composition, microstructure, stress, and chiral characteristics.

[0003] Existing technologies include multi-frame rotating polarizers, division of focal plane (DoFP) polarization cameras, and metasurface single-shot imaging systems.

[0004] The article “Dai Hu, Yan Changxiang. Study on the azimuth error of a rotating waveplate polarizer [J]. Spectroscopy and Spectral Analysis, 2017, 37 (07): 2288-2293” mentions a multi-frame rotating polarizer: the polarization detection function is realized by a polarization component consisting of a rotating waveplate, a fixed linear polarizer and a detector. By rotating the polarization element and switching the polarizer angles of 0°, 45°, 90° and 135°, the light intensity at different positions is obtained to construct a system of linear equations, and then the Stokes parameters are solved by matrix.

[0005] The article “Wang Wei, Zhang Haiyu, Ke Lianbao, et al. Polarization error analysis and polarization calibration of a focal plane polarization camera [J]. Acta Photonica Sinica, 2024, 53 (09): 35-46” mentions a DoFP polarization camera: a micro-polarization array is integrated on the pixel array of the focal plane detector. Each 2×2 superpixel unit contains four sub-pixels, which are covered by micro-polarizers with polarizer orientations of 0°, 45°, 90°, and 135° respectively. After the light is converged by the microlens array, it passes through the corresponding polarizers and is finally received by the detector pixels. Polarization information is obtained by the transmission intensity difference of four adjacent directions. The polarization directions of the four pixels in the volume pixel of the micro-polarizer array are 90°, 45°, 0°, and 135° clockwise. The polarization state of the incident light field can be represented by the Stokes parameters (S0, S1, S2, S3). The transmission intensity of the polarizer follows Malus's law: ; in, θ Let be the transmission axis angle of the polarizer, when θAt angles of 0°, 45°, 90°, and 135°, four different strength measurements were obtained: I 0 = 1 / 2 ( S 0 +S 1), I 90 = 1 / 2 ( S 0 -S 1), I 45 = 1 / 2 ( S 0 +S 2), I 135 = 1 / 2 ( S 0 -S2 By using the linear combination of the intensities in these four directions, we can calculate: S 0 =I 0 +I 90 =I 45 +I 135 ,S 1 =I 0 -I 90 ，S 2 =I 45 -I 135 Furthermore, the degree of polarization (DoLP) and the angle of polarization (AoP) can be calculated.

[0006] The article "Zhou Junzhuo, Hao Jia, Yu Xiaochang, et al. Research progress of metasurfaces for polarization imaging [J]. Chinese Optics (Chinese and English), 2023, 16 (05): 973-995" mentions a metasurface polarization camera: the metasurface is composed of micro-nano structural units arranged at subwavelength intervals, which can apply different phase modulations to light with different polarization states. Through reasonable design, beam splitting of orthogonal polarization states such as horizontal / vertical (x,y), diagonal / anti-diagonal (a,b), and left / right circular polarization (l,r) can be achieved. After the incident light shines on the metasurface, the light with different polarization states is spatially separated and focused to different positions on the focal plane of the detector. By measuring the light intensity distribution at each focal point, the full Stokes parameters of the incident light can be obtained in a single frame.

[0007] However, the existing solutions proposed above have the following problems: Multi-frame rotating polarizers require multi-frame acquisition, necessitating the sequential change of polarization states via mechanical rotation of the polarizer or electro-optic modulator. Each frame corresponds to a measurement state, which cannot meet real-time requirements in dynamic scenes and introduces motion artifacts. Power drift of the light source and fluctuations in ambient temperature during multi-frame acquisition directly introduce measurement errors, affecting the repeatability and accuracy of long-term measurements. Mechanical rotation also increases system instability, resulting in poor instantaneous measurement stability. A single superpixel in a DoFP polarization camera must contain four micro-polarizers in different directions, which are essentially four detection units. This results in an equivalent imaging resolution that is only one-quarter of the detector's original resolution, leading to a reduction in spatial resolution. Furthermore, because it can only acquire linear polarization component information and cannot acquire circular or general complex polarization components, it cannot directly obtain complex coherence. The polarization response of metasurface polarization cameras is strongly dependent on wavelength, and polarization control is limited to a single band. They typically operate only in a narrow band near the designed wavelength, exhibiting poor broadband performance. Once fabricated, their polarization response characteristics are fixed and cannot be flexibly adjusted; the fixed structure is also sensitive to temperature. They are highly sensitive to the incident angle and polarization state, and have poor broadband performance. Summary of the Invention

[0008] To address the aforementioned problems in the prior art, this invention provides a self-referenced three-phase polarization imaging system and method based on an integrated photonic neural network. The technical problem to be solved by this invention is achieved through the following technical solution: In a first aspect, embodiments of the present invention provide a self-referenced three-phase polarization imaging system based on an integrated photonic neural network, the self-referenced three-phase polarization imaging system comprising: A polarization beam splitter rotator is used to split and rotate the input light field to obtain horizontal and vertical polarization components, and then map the horizontal and vertical polarization components into TE modes respectively. A silicon-based photonic chip includes: a first 1×4 MMI beam splitter for equally splitting the mapped horizontal polarization component into four output light intensity signals, the first output light intensity signal being used for monitoring, and the second to fourth output light intensity signals being input to the MZI at a projection angle of 0°, a projection angle of +90°, and a projection angle of -90°, respectively; a second 1×4 MMI beam splitter for equally splitting the mapped vertical polarization component into four output light intensity signals, the first output light intensity signal being used for monitoring; and a first 2×2 MMI interferometer for... The second output light intensity signal of the 1×4 MMI beam splitter is coupled with the output of the MZI at a 0° projection angle to obtain the first interference output light intensity signal; the second 2×2 MMI interferometer is used to couple the third output light intensity signal of the second 1×4 MMI beam splitter with the output of the MZI at a +90° projection angle to obtain the second interference output light intensity signal; the third 2×2 MMI interferometer is used to couple the fourth output light intensity signal of the second 1×4 MMI beam splitter with the output of the MZI at a -90° projection angle to obtain the third interference output light intensity signal. The photodetector is used to convert the first interference output light intensity signal, the second interference output light intensity signal, the third interference output light intensity signal, the first output light intensity signal of the first 1×4MMI beam splitter, and the first output light intensity signal of the second 1×4MMI beam splitter into analog electrical signals, and then convert them into digital electrical signals by the built-in analog-to-digital converter. The data processing module is used to perform paired channel normalization processing on the first, second, and third interference output light intensity signals after digitization of the first output light intensity signal of the first 1×4MMI beam splitter and the first output light intensity signal of the second 1×4MMI beam splitter, and to establish a rotation correction matrix. Based on the rotation correction matrix and the normalized first, second, and third interference output light intensity signals, the module calculates the rotationally corrected Stokes parameters and performs polarization imaging based on the rotationally corrected Stokes parameters.

[0009] In a second aspect, embodiments of the present invention provide a self-referenced three-phase polarization imaging method based on an integrated photonic neural network, applied to any of the self-referenced three-phase polarization imaging systems based on an integrated photonic neural network described in the first aspect, the corresponding method comprising: The input light field is split and polarized to obtain horizontal and vertical polarization components, and the horizontal and vertical polarization components are mapped to TE modes respectively. The mapped horizontal polarization component is equally divided into four output light intensity signals. The first output light intensity signal is used for monitoring, and the second to fourth output light intensity signals are respectively input to the MZI of the 0° projection angle, the MZI of the +90° projection angle, and the MZI of the -90° projection angle. The mapped vertical polarization components are equally divided into four output light intensity signals, with the first output light intensity signal used for monitoring. The second output light intensity signal of the second 1×4MMI beam splitter is coupled with the output of the MZI at a projection angle of 0° to obtain the first interference output light intensity signal. The third output light intensity signal of the second 1×4MMI beam splitter is coupled with the output of the MZI at a projection angle of +90° to obtain the second interference output light intensity signal. The fourth output light intensity signal of the second 1×4MMI beam splitter is coupled with the output of the MZI at a projection angle of -90° to obtain the third interference output light intensity signal. The first interference output light intensity signal, the second interference output light intensity signal, the third interference output light intensity signal, as well as the first output light intensity signal of the first 1×4MMI beam splitter and the first output light intensity signal of the second 1×4MMI beam splitter are all converted into analog electrical signals, and then converted into digital electrical signals by the built-in analog-to-digital converter. Based on the first output light intensity signal of the first 1×4MMI beam splitter and the first output light intensity signal of the second 1×4MMI beam splitter after digitization, the first, second, and third interference output light intensity signals after digitization are subjected to paired channel normalization processing, and a rotation correction matrix is ​​established. Based on the rotation correction matrix and the normalized first, second, and third interference output light intensity signals, the rotated Stokes parameters are calculated, and polarization imaging is performed based on the rotated Stokes parameters.

[0010] The beneficial effects of this invention are: This invention proposes a self-referenced three-phase polarization imaging system based on an integrated photonic neural network. It innovatively achieves three-phase projection by setting a fixed phase difference within a silicon-based photonic chip, enabling single-port readout and simultaneous output of three-phase interference signals. This allows for the acquisition of all phase and polarization information in a single imaging operation, achieving an effective resolution consistent with the original detector array. It eliminates power and temperature drift, and single-frame acquisition is free of timing errors and motion artifacts. Furthermore, through an on-chip three-phase interferometry measurement architecture, combined with paired channel normalization and rotation correction, a stable and compact single-port self-referenced readout method is achieved, enabling the measurement of complete Stokes parameters and polarization information. The system can simultaneously recover the polarization state and complex amplitude information of each pixel in a single acquisition, thereby obtaining complete characteristics of the light field in the three dimensions of intensity, phase, and polarization. The system employs a programmable MZI network as the measurement layer, and its phase bias can be flexibly adjusted to match different bandwidth requirements. Because silicon-based photonic platforms possess broadband response characteristics, covering the visible band, and the devices maintain stable interferometric performance within the communication band, this invention's system offers greater adaptability in terms of wavelength compatibility and functional reusability compared to metasurface solutions that rely on fixed resonant structures. Compared to traditional multi-frame phase shift measurement methods, this invention eliminates the dependence on local oscillator light and multi-port balancing, requiring only a single calibration. The system proposed in this invention is suitable for scenarios such as polarization imaging, optical coherence tomography, and material property analysis. The on-chip interferometric structure technology is mature, with high repeatability, enabling more stable polarization and complex amplitude measurements, and providing an on-chip solution for high-stability, high-throughput optical vector measurements.

[0011] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the structure of a self-referenced three-phase polarization imaging system based on an integrated photonic neural network provided in an embodiment of the present invention; Figure 2 The embodiments of the present invention provide Figure 1 A schematic diagram of the structure of a silicon-based photonic chip; Figure 3 This is a flowchart illustrating a self-referenced three-phase polarization imaging method based on an integrated photonic neural network provided in an embodiment of the present invention. Figure 4 This is a flowchart illustrating another self-referenced three-phase polarization imaging method based on an integrated photonic neural network provided in an embodiment of the present invention. Detailed Implementation

[0013] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0014] The rapid development of integrated photonics technology allows silicon photonics processes to integrate optical devices such as polarization splitter-rotators (PSRs), multimode interferometers (MMIs), and programmable Mach-Zehnder interferometer (MZI) arrays onto chips, enabling low-drift, highly reproducible optical interferometry and linear computation. On-chip optical neural networks have been used for inference and classification, but are less commonly used directly for physical quantity measurement and polarization state recovery. This invention extends traditional 3D imaging by introducing a polarization dimension. By designing a three-phase parallel interferometry architecture on an on-chip photonic integrated system, the complete Stokes parameter orientation of each pixel in the imaging region can be simultaneously acquired in a single exposure. S 0, S 1, S 2, S 3). Therefore, the imaging dimension is expanded from the traditional "three-dimensional (space + depth)" to "four-dimensional (space + depth + polarization)," achieving a higher-dimensional light field characterization. This high-dimensional imaging method can significantly improve the system's capabilities in applications such as high-speed imaging, polarization-sensitive material analysis, and biological tissue structure detection, providing a novel technical approach for related fields. Specifically: Firstly, please see Figure 1 and Figure 2 This invention provides a self-referenced three-phase polarization imaging system based on an integrated photonic neural network. The self-referenced three-phase polarization imaging system includes: A polarization beam splitter rotator is used to split and rotate the input light field to obtain horizontal and vertical polarization components, and then map the horizontal and vertical polarization components into TE modes respectively. A silicon-based photonic chip includes: a first 1×4 MMI beam splitter for equally splitting the mapped horizontal polarization component into four output light intensity signals, the first output light intensity signal being used for monitoring, and the second to fourth output light intensity signals being input to the MZI at a projection angle of 0°, a projection angle of +90°, and a projection angle of -90°, respectively; a second 1×4 MMI beam splitter for equally splitting the mapped vertical polarization component into four output light intensity signals, the first output light intensity signal being used for monitoring; and a first 2×2 MMI interferometer for... The second output light intensity signal of the 1×4 MMI beam splitter is coupled with the output of the MZI at a 0° projection angle to obtain the first interference output light intensity signal; the second 2×2 MMI interferometer is used to couple the third output light intensity signal of the second 1×4 MMI beam splitter with the output of the MZI at a +90° projection angle to obtain the second interference output light intensity signal; the third 2×2 MMI interferometer is used to couple the fourth output light intensity signal of the second 1×4 MMI beam splitter with the output of the MZI at a -90° projection angle to obtain the third interference output light intensity signal. The photodetector is used to convert the first interference output light intensity signal, the second interference output light intensity signal, the third interference output light intensity signal, the first output light intensity signal of the first 1×4MMI beam splitter, and the first output light intensity signal of the second 1×4MMI beam splitter into analog electrical signals, and then convert them into digital electrical signals by the built-in analog-to-digital converter. The data processing module is used to perform paired channel normalization processing on the first, second, and third interference output light intensity signals after digitization of the first output light intensity signal of the first 1×4MMI beam splitter and the first output light intensity signal of the second 1×4MMI beam splitter, and to establish a rotation correction matrix. Based on the rotation correction matrix and the normalized first, second, and third interference output light intensity signals, the module calculates the rotationally corrected Stokes parameters and performs polarization imaging based on the rotationally corrected Stokes parameters.

[0015] The data processing module is also used to calculate the polarization azimuth angle, linear polarization degree, and complex cross-polarization coherence degree based on the rotated Stokes parameters.

[0016] The system proposed in this invention includes: a fiber-coupled laser, a polarization beam rotator (PSR), a 1×4 MMI beam splitter, a 2×2 MMI interferometer, and a photodetector. These modules are connected via a silicon-based waveguide to form an integrated optical path, enabling the readout of the incident light field polarization vector information on a silicon-based photonic chip. It also includes a data processing module, which can be executed by a computer. Therefore, the system proposed in this invention mainly comprises four parts: input light source and polarization separation, power beam splitting and detection, phase modulation and interference, and output detection. Specifically: 1. Input Light Source and Polarization Separation: The laser source provided by the fiber-coupled laser is coupled into the silicon-based photonic chip via a fiber array. A polarization beam splitter and rotator is set at the input end of the silicon-based photonic chip to perform beam splitting and polarization rotation, decomposing the input light field into horizontal polarization (H) components and vertical polarization (V) components. Simultaneously, the horizontal and vertical polarization components are mapped to TE modes and introduced into the A / B waveguide branches respectively. The input Jones vector can be characterized as: E= [ E H ,E V ] T , E H Represents the horizontal polarization component. E V Indicates the vertical polarization component, [] T This indicates the transpose operation.

[0017] The polarization state of the incident light field is determined by the Stokes parameter ( S 0 ,S 1 ,S 2 ,S 3) Characterization: ; ; ; .

[0018] 2. Power Beam Splitting and Detection: Path A enters the first 1×4 MMI beamsplitter, and path B enters the second 1×4 MMI beamsplitter. Each path is equally split into four output light intensity signals. The first output light intensity signal A1 from the first 1×4 MMI beamsplitter and the first output light intensity signal B1 from the second 1×4 MMI beamsplitter serve as monitoring ports to monitor changes in optical power in real time. They are also used for pairwise channel normalization (PCN) correction to eliminate system drift and ensure the stability of subsequent detection data. The remaining three pairs of branches (A2–A4 and B2–B4) enter subsequent 2×2 MMI interferometers.

[0019] 3. Phase Modulation and Interference: On-chip phase modulation is achieved using a programmable Mach-Zehnder interferometer (MZI), whose phase bias is programmable to perform different projection operations. The second to fourth output light intensity signals A2 and A4 of the first 1×4 MMI beamsplitter are respectively fed into three groups of MZIs. Each MZI has a built-in thermo-optical phase modulator, which is set to three projection phases: 0°, +90°, and -90°. After modulation, the signals are coupled in pairs with the B-channel signals in a 2×2 MMI interferometer. That is, the second output light intensity signal B2 of the second 1×4 MMI beamsplitter is coupled with the output of the MZI at a 0° projection angle, the third output light intensity signal B3 of the second 1×4 MMI beamsplitter is coupled with the output of the MZI at a +90° projection angle, and the fourth output light intensity signal B4 of the second 1×4 MMI beamsplitter is coupled with the output of the MZI at a -90° projection angle. The 2×2 MMI interferometer retains only one output port as a readout, achieving "single-port three-phase interference," thus avoiding the signal imbalance and phase drift problems caused by multiple output ports. The three output interference light intensity signals, namely the first, second, and third interference output light intensity signals, are denoted as I0, I... +90 I -90 .

[0020] Compared to traditional fixed-structure interferometer arrays, the system proposed in this invention employs a single-port three-phase self-referenced interferometry architecture, achieving three-phase phase projection on-chip. The phase reference is provided by the internal optical path, eliminating the need for an external local oscillator to acquire complex amplitude information, thereby simplifying the system structure and improving stability. Furthermore, the internal programmable MZI network enables phase tunability and path reconfigurability, allowing for flexible configuration of the interferometric path in different wavelength bands or measurement modes. This design overcomes the limitations of narrow wavelength response and fixed functionality in metasurface polarization imaging devices, significantly improving the system's adaptability and scalability.

[0021] 4. Output Detection: Three interference output ports (first 2×2 MMI interferometer, second 2×2 MMI interferometer, and third 2×2 MMI interferometer) and two detection ports (A1 monitoring port of the first 1×4 MMI beam splitter and B1 monitoring port of the second 1×4 MMI beam splitter) are connected to a photodetector. The photodetector converts all received light intensity signals into analog electrical signals and uses a built-in analog-to-digital converter (ADC) to convert all analog electrical signals into digital electrical signals, ensuring that the three interference results are recorded at the same time as the monitoring reference, thus avoiding time drift.

[0022] To address the signal inconsistency issues caused by light source power fluctuations and device drift in traditional systems, this invention employs a paired channel normalization method in the data processing module. This method is a system calibration and adjustment method proposed for the system architecture of this invention, ensuring high stability and consistency of the three-phase interference output during long-term operation. The data processing module performs paired channel normalization processing on the digitized first, second, and third interference output light intensity signals, including: using the first output light intensity signals of the digitized first 1×4MMI beamsplitter and the first output light intensity signals of the second 1×4MMI beamsplitter as light intensity reference signals, establishing a normalization relationship expression between the interference output and the light intensity reference signals; and based on this normalization relationship expression, performing paired channel normalization processing on the digitized first, second, and third interference output light intensity signals respectively. More specifically: This invention embodiment monitors the output light intensity signals of ports A1 and B1 ( I A1 、I B1 Using this as a reference signal, the interference output light intensity signals of the first, second, and third 2×2 MMI interferometers are normalized. Before data acquisition, the system response is measured through a monitoring channel to establish a normalization relationship expression between the interference output light intensity signal and the reference signal. A differential method is used to normalize the three interference output signals, thereby obtaining a stable normalization result independent of changes in light source intensity and fluctuations in photodetector gain. Specifically: In actual measurements, light source power drift and gain channel mismatch will cause the light intensity signal I of each measurement channel to be subject to a slow-varying multiplicative factor. and additional offset Impact: ; in, To achieve true interference, output light intensity signals, while simultaneously monitoring two channels that do not participate in interference. and Also affected and Influence: ; Define the normalized relationship between the interference output and the light intensity reference signal, expressed as: ; in, This represents the normalized light intensity signal. Represents the normalization coefficient. This represents the first output optical intensity signal of the first 1×4 MMI beam splitter after digitization. This represents the first output optical intensity signal of the digitized second 1×4 MMI beam splitter. This represents the first, second, or third interference output light intensity signal after digitization. It can be seen that when... Very small and It is the slow-varying multiplicative factor common to all channels, obtained from the above formula. It can precisely offset This ensures that subsequent signals are not affected by gain mismatch.

[0023] Normalization coefficient For the system calibration phase, this involves determining, for example, selecting one or more known signal sources as references, observing the impact of different α values ​​on system stability, and selecting the value that makes the signal most stable. Normalization coefficient Once determined, this standard remains unchanged throughout subsequent measurements, providing a stable and consistent benchmark for the entire system operation. It is applied to the signals of each channel to compensate for gain and phase differences, thereby normalizing the interferometric output light intensity signal. This results in a stable normalized result independent of changes in light source intensity and detector gain fluctuations. The relative amplitude of the signal is independent of the light source intensity, ensuring stable and consistent subsequent calculations. This difference-by-sum approach overcomes the power and gain drift problems in single-port readout mode. The PCN method can correct drift in real time during each measurement in signal acquisition, significantly reducing errors in experimental repetition.

[0024] Because 2×2 MMI interferometers have an inherent phase due to their manufacturing process. Slight amplitude imbalances can cause the Stokes parameters to rotate and scale, potentially introducing static biases and resulting in outputs deviating from ideal values. To eliminate this effect, during system calibration, the correction angle is measured by inputting a known polarized light field. Establish rotation correction matrix The established rotation correction matrix ， The formula is expressed as: ; in, Indicates the correction angle. This represents the rotation correction matrix.

[0025] During actual measurement, all normalized signals in the data processing module undergo a one-time matrix rotation operation, using a pre-measured rotation correction matrix. To eliminate the phase error caused by the 2×2 MMI interferometer and correct the internal imbalance of the MMI, a rotation correction is performed on the complex vector signal output from the three-phase interference. The Stokes parameters after rotation correction are expressed by the formula: ; in, , All are Stokes parameters. Indicates before rotation correction , , This indicates the normalized light intensity signal corresponding to the output light intensity signal of the third interference. This indicates the normalized light intensity signal corresponding to the second interference output light intensity signal. Indicates before rotation correction , , This represents the normalized light intensity signal corresponding to the first interference output light intensity signal, where β represents the scaling factor. Indicates static deviation. For rotation correction matrix middle Values The expression at that time. At this time, No rotation correction is required; the original calculated values ​​are preserved. Correction angle. Scaling factor β, static deviation All of these are pre-calculated during the system calibration phase. For example, during the system calibration phase, an eight-step phase scan is performed on a uniform region, such as 0° to 135°, with one phase point acquired every 45°. The original data combinations are recorded, and then β can be accurately calculated by fitting these data with a sine curve. and These are the three key error parameters. β, and Once determined, it remains unchanged in subsequent actual measurement and calculation processes, providing a stable and consistent benchmark for the entire operation of the system.

[0026] The three signals after normalization ( , , The signals satisfy the coherent superposition relationship, and the monitoring branch can measure the total light intensity signal and the branch light intensity signal. The three-channel interference light intensity signals can be expressed by the following equation: , ; The Stokes parameters can be solved using closed-form analytical formulas, thereby obtaining the complex amplitude and phase information of the input light field: Re ( ) ∝ ( - ); Im ( ) ∝ ( - ); S 0 ∝ ( I A1 + I B1 ); S 1 ∝ ( I A1 - I B1 ); Finally, based on the rotated Stokes parameters, AoP (polarization azimuth angle), DoLP (linear polarization degree), and complex cross-polarization coherence (W) were further calculated. HV These parameters characterize the polarization state distribution of the light field and can be used for imaging reconstruction and material classification.

[0027] ; ; .

[0028] In summary, the self-referenced three-phase polarization imaging system based on an integrated photonic neural network proposed in this invention innovatively achieves three-phase projection by setting a fixed phase difference within a silicon-based photonic chip, enabling single-port readout and simultaneous output of three-phase interference signals. It acquires all phase and polarization information in a single imaging operation, achieving the same effective resolution as the original detector array. It eliminates power and temperature drift, and single-frame acquisition is free of timing errors and motion artifacts. Furthermore, through an on-chip three-phase interferometric measurement architecture, combined with channel normalization and rotation correction, a stable and compact single-port self-referenced readout method is achieved, enabling the measurement of complete Stokes parameters and polarization information. The system proposed in this invention can simultaneously recover the polarization state and complex amplitude information of each pixel in a single acquisition, thereby obtaining complete characteristics of the light field in the three dimensions of intensity, phase, and polarization. The system uses a programmable MZI network as the measurement layer, and its phase bias can be flexibly adjusted to match different bandwidth requirements by adjusting the interference path. Because silicon-based photonic platforms possess broadband response characteristics, covering the visible band, and the devices maintain stable interferometric performance within the communication band, this invention's system offers greater adaptability in terms of wavelength compatibility and functional reusability compared to metasurface solutions that rely on fixed resonant structures. Compared to traditional multi-frame phase shift measurement methods, this invention eliminates the dependence on local oscillator light and multi-port balancing, requiring only a single calibration. The system proposed in this invention is suitable for scenarios such as polarization imaging, optical coherence tomography, and material property analysis. The on-chip interferometric structure technology is mature, with high repeatability, enabling more stable polarization and complex amplitude measurements, and providing an on-chip solution for high-stability, high-throughput optical vector measurements.

[0029] Secondly, please see Figure 3 This invention proposes a self-referenced three-phase polarization imaging method based on an integrated photonic neural network, applied to the self-referenced three-phase polarization imaging system based on an integrated photonic neural network described in the first aspect. The corresponding method includes: S10. The input light field is split and polarized to obtain horizontal polarization components and vertical polarization components, and the horizontal polarization components and vertical polarization components are mapped to TE modes respectively. S20. The mapped horizontal polarization component is equally divided into four output light intensity signals. The first output light intensity signal is used for monitoring. The second to fourth output light intensity signals are respectively input to the MZI of the 0° projection angle, the MZI of the +90° projection angle, and the MZI of the -90° projection angle. S30. The mapped vertical polarization components are equally divided into four output light intensity signals. The first output light intensity signal is used for monitoring. S40. Couple the second output light intensity signal of the second 1×4MMI beam splitter with the output of the MZI at a 0° projection angle to obtain the first interference output light intensity signal. S50. Couple the third output light intensity signal of the second 1×4MMI beam splitter with the output of the MZI at a projection angle of +90° to obtain the second interference output light intensity signal. S60. Couple the fourth output light intensity signal of the second 1×4MMI beam splitter with the output of the MZI at a projection angle of -90° to obtain the third interference output light intensity signal. S70. The first interference output light intensity signal, the second interference output light intensity signal, the third interference output light intensity signal, the first output light intensity signal of the first 1×4MMI beam splitter, and the first output light intensity signal of the second 1×4MMI beam splitter are all converted into analog electrical signals, and then converted into digital electrical signals by the built-in analog-to-digital converter. S80. Based on the first output light intensity signal of the first 1×4MMI beam splitter and the first output light intensity signal of the second 1×4MMI beam splitter after digitization, the first interference output light intensity signal, the second interference output light intensity signal, and the third interference output light intensity signal after digitization are subjected to paired channel normalization processing, and a rotation correction matrix is ​​established. Based on the rotation correction matrix and the normalized first interference output light intensity signal, the second interference output light intensity signal, and the third interference output light intensity signal, the rotated Stokes parameters are calculated, and polarization imaging is performed based on the rotated Stokes parameters.

[0030] Further, please see Figure 4 The self-referenced three-phase polarization imaging method based on integrated photonic neural networks proposed in this invention also includes: S90. Calculate the polarization azimuth angle, linear polarization degree, and complex cross-polarization coherence degree based on the rotated Stokes parameters.

[0031] As for the method embodiment of the second aspect, since it is basically similar to the system embodiment of the first aspect, the description is relatively simple, and relevant details can be found in the description of the system embodiment of the first aspect.

[0032] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0033] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the specification and accompanying drawings, will understand and implement other variations of the disclosed embodiments in carrying out the claimed invention. In the specification, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality. While certain measures are described in different embodiments, this does not mean that these measures cannot be combined to produce good results.

[0034] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. An integrated photonic neural network based self-referenced three-phase polarization imaging system, characterized in that, The self-referencing three-phase polarization imaging system comprises: A polarization beam splitter rotator is configured to split and rotate the input light field to obtain horizontal polarization components and vertical polarization components, and map the horizontal polarization components and the vertical polarization components into TE modes respectively; The silicon-based photonic chip comprises: a first 1*4 MMI beam splitter configured to equally split the mapped horizontal polarization components into four output light intensity signals, a first output light intensity signal is used for monitoring, and second to fourth output light intensity signals are input into a 0° projection angle MZI, a +90° projection angle MZI, and a -90° projection angle MZI respectively; a second 1*4 MMI beam splitter configured to equally split the mapped vertical polarization components into four output light intensity signals, a first output light intensity signal is used for monitoring; a first 2*2 MMI interferometer configured to couple the second output light intensity signal of the second 1*4 MMI beam splitter with the output of the 0° projection angle MZI to obtain a first interference output light intensity signal; a second 2*2 MMI interferometer configured to couple the third output light intensity signal of the second 1*4 MMI beam splitter with the output of the +90° projection angle MZI to obtain a second interference output light intensity signal; and a third 2*2 MMI interferometer configured to couple the fourth output light intensity signal of the second 1*4 MMI beam splitter with the output of the -90° projection angle MZI to obtain a third interference output light intensity signal; A photodetector is configured to convert the first interference output light intensity signal, the second interference output light intensity signal, the third interference output light intensity signal, the first output light intensity signal of the first 1*4 MMI beam splitter, and the first output light intensity signal of the second 1*4 MMI beam splitter into analog electrical signals, and further convert the analog electrical signals into digital electrical signals by a built-in analog-to-digital converter; A data processing module is configured to perform paired channel normalization processing on the digitized first interference output light intensity signal, the second interference output light intensity signal, and the third interference output light intensity signal according to the digitized first output light intensity signal of the first 1*4 MMI beam splitter and the digitized first output light intensity signal of the second 1*4 MMI beam splitter, establish a rotation correction matrix, calculate the Stokes parameters after rotation correction according to the rotation correction matrix and the normalized first interference output light intensity signal, the second interference output light intensity signal, and the third interference output light intensity signal, and perform polarization imaging according to the Stokes parameters after rotation correction.

2. The integrated photonic neural network-based self-referenced three-phase polarization imaging system of claim 1, wherein, The data processing module is further configured to calculate a polarization azimuth angle, a linear polarization degree, and a complex cross-polarization coherence degree according to the Stokes parameters after rotation correction.

3. The integrated photonic neural network-based self-referenced three-phase polarization imaging system of claim 1, wherein, The paired channel normalization processing on the digitized first interference output light intensity signal, the second interference output light intensity signal, and the third interference output light intensity signal in the data processing module comprises: The digitized first output light intensity signal of the first 1*4 MMI beam splitter and the digitized first output light intensity signal of the second 1*4 MMI beam splitter are taken as light intensity reference signals to establish a normalization relationship expression between the interference output and the light intensity reference signals. The normalized relationship expression between the interference output and the light intensity reference signal is expressed by a formula as follows:

4. The integrated photonic neural network-based self-referenced three-phase polarization imaging system of claim 3, wherein, The normalized relationship expression between the interference output and the light intensity reference signal is expressed by a formula as follows: ； wherein, represents the normalized optical intensity signal, represents the normalization coefficient, represents the digitized first 1x4 MMI beam splitter first path output optical intensity signal, represents the digitized second 1x4 MMI beam splitter first path output optical intensity signal, represents the digitized first interference output optical intensity signal or the second interference output optical intensity signal or the third interference output optical intensity signal.

5. The integrated photonic neural network-based self-referenced three-phase polarization imaging system of claim 4, wherein, Normalization coefficient Is predetermined at the system calibration stage.

6. The integrated photonic neural network-based self-referenced three-phase polarization imaging system of claim 1, wherein, The established rotation correction matrix ， The formula is expressed as: ； wherein, denotes a correction angle, denotes a rotation correction matrix.

7. The integrated photonic neural network based self-referenced three-phase polarization imaging system of claim 6, wherein, The Stokes parameter after rotation correction is expressed by a formula as follows: ； wherein , are both Stokes parameters, denotes the pre-rotation corrected , , denotes the normalized intensity signal corresponding to the third interference output intensity signal, denotes the normalized intensity signal corresponding to the second interference output intensity signal, denotes the pre-rotation corrected , , denotes the normalized intensity signal corresponding to the first interference output intensity signal, and β denotes a scaling factor, denotes a static bias.

8. The integrated photonic neural network based self-referenced three-phase polarization imaging system of claim 7, wherein, corrected angle , scaling factor β, static bias are all calculated in advance in the system calibration stage.

9. An integrated photonic neural network based self-referenced three-phase polarization imaging method, characterized in that, The self-referenced three-phase polarization imaging system based on the integrated photonic neural network is applied to the method according to any one of claims 1-8, and the method comprises: The input light field is split and polarized to obtain horizontal polarization components and vertical polarization components, and the horizontal polarization components and the vertical polarization components are respectively mapped into TE modes; The mapped horizontal polarization components are equally divided into four output light intensity signals, the first output light intensity signal is used for monitoring, and the second output light intensity signal to the fourth output light intensity signal are respectively input into the MZI with a projection angle of 0°, the MZI with a projection angle of +90° and the MZI with a projection angle of -90°; The mapped vertical polarization components are equally divided into four output light intensity signals, and the first output light intensity signal is used for monitoring; The second output light intensity signal of the second 1x4 MMI beam splitter is coupled with the output of the MZI with a projection angle of 0° to obtain a first interference output light intensity signal; The third output light intensity signal of the second 1x4 MMI beam splitter is coupled with the output of the MZI with a projection angle of +90° to obtain a second interference output light intensity signal; The fourth output light intensity signal of the second 1x4 MMI beam splitter is coupled with the output of the MZI with a projection angle of -90° to obtain a third interference output light intensity signal; The first interference output light intensity signal, the second interference output light intensity signal, the third interference output light intensity signal, the first output light intensity signal of the first 1x4 MMI beam splitter and the first output light intensity signal of the second 1x4 MMI beam splitter are all converted into analog electrical signals, and then converted into digital electrical signals by a built-in analog-to-digital converter; The first interference output light intensity signal, the second interference output light intensity signal and the third interference output light intensity signal are respectively processed by paired channel normalization according to the first output light intensity signal of the first 1x4 MMI beam splitter and the first output light intensity signal of the second 1x4 MMI beam splitter after digitization, a rotation correction matrix is established, the Stokes parameter after rotation correction is calculated according to the rotation correction matrix and the normalized first interference output light intensity signal, the second interference output light intensity signal and the third interference output light intensity signal, and polarization imaging is performed according to the Stokes parameter after rotation correction.

10. The integrated photonic neural network based self-referenced three-phase polarization imaging method of claim 9, wherein, The method further comprises: The polarization azimuth angle, the linear polarization degree and the complex cross-polarization coherence degree are calculated according to the Stokes parameter after rotation correction.

Images