Incoherent Full-Color Polarization Edge Imaging System and Method Based on Reverse Design of Photonic Crystals

Abstract

This invention discloses an incoherent full-color polarization edge imaging system and method based on reverse-designed photonic crystals, belonging to the fields of optical imaging and computational optics. The system includes an incoherent light source module, a polarization-dependent photonic crystal (PDPC) device, an imaging detection module, and an image processing module. The PDPC device is constructed from stacked different all-dielectric materials, with the thickness of each layer as a trainable parameter. Iterative optimization using a gradient descent algorithm, combined with a ReLU function constraint on the thickness lower limit, yields the optimal structure. The method acquires intermediate images of orthogonally linearly polarized states, performs spatial difference calculations, and utilizes a polarization difference optical transfer function that satisfies the first-order spatial differential condition to output an edge-enhanced image containing full-color and polarization information. This invention requires no coherent light source, operates stably in the 400-700nm visible light band, and exhibits high imaging efficiency and excellent contrast, providing a feasible solution for large-aperture broadband optical simulation processing.

Description

Technical Field

[0001] This invention belongs to the field of optical imaging and computational optics technology, and in particular refers to an incoherent full-color polarization edge imaging system and method based on reverse-designed photonic crystals. Background Technology

[0002] With the rapid iteration of artificial intelligence technology and the continuous expansion of high-throughput data processing scenarios, the demand for real-time analysis and efficient processing of image data has exploded, creating an urgent technical need for ultra-high-speed, low-power image processing methods and hardware platforms. Traditional digital image processing relies on electronic chips to complete signal acquisition, computation, and output. While mature in general computing scenarios, it suffers from inherent limitations due to the physical transmission characteristics and clock frequency of electronic devices, including processing speed bottlenecks, high power consumption, and poor scalability for large-scale parallel computing. This makes it difficult to meet the application requirements of real-time feature extraction in high-speed dynamic scenes and massive image streams. Optical analog computing integrates optical front-end sensing with optical field physical computation. Leveraging the natural parallelism and low-loss characteristics of photon propagation, it possesses core advantages such as ultra-high-speed computation, high-throughput processing, and low-energy operation. It can bypass the speed limitations of electronic processing and has become an important technical route for next-generation high-speed image processing.

[0003] Current mainstream optical simulation computing architectures are mostly based on classic 4f optical systems. These systems utilize optical masks arranged in the Fourier transform plane formed by two lenses to modulate the amplitude and phase distribution of the light field, thereby performing mathematical operations such as convolution, integration, and filtering. This enables large-scale parallel processing of two-dimensional images with low power consumption. However, these traditional 4f systems have significant structural and technological limitations: Firstly, the system consists of lens groups, optical path alignment structures, mask loading mechanisms, and other components, resulting in a large overall size and long optical path, making it unsuitable for miniaturized, integrated, and portable optical device applications. Secondly, the computational function is highly dependent on the spatial positioning accuracy of the mask within the Fourier plane. Alignment errors between the mask and the optical axis, and between the lens focus, directly alter the light field modulation effect, causing computational distortion and reduced image processing accuracy. Furthermore, high-precision alignment processes are complex and costly, significantly hindering the system's integration and industrial application.

[0004] With the development of micro-nano fabrication and planar optics technology, compact optical analog computing devices and chip-based solutions have become a research focus, and the verification of functional devices such as optical convolution, optical integration, and optical spatial differentiation has been gradually realized. Among them, optical spatial differentiation can directly extract key features such as image edges and contours, and is a core computational prerequisite for computer vision tasks such as target recognition, feature classification, and image segmentation. Related technology research and device development have received widespread attention from academia and industry. Existing optical spatial differentiation schemes are still based on the 4f system framework, achieving edge imaging by loading optical devices with specific wave vector spatial responses or phase distributions onto the Fourier plane. For example, an aluminum nanofin array metasurface is used to construct a radial secondary transmission distribution to complete the Laplacian second-order edge operation under incoherent illumination; or a polarization filter structure is integrated into the 4f system, and orthogonal polarizers are used to construct the radial gradient of the polarization state of the light field to achieve spatial differentiation. These schemes have not deviated from the inherent architecture of the 4f system and still suffer from problems such as large size, stringent alignment requirements, and high integration difficulty.

[0005] To circumvent the size and alignment defects of 4f optical systems, researchers have successively proposed edge imaging schemes without Fourier transform systems. One type of scheme is based on a geometric phase polarization grating, which modulates linearly polarized incident light to generate circularly polarized light with opposite chirality and a lateral displacement difference. Combined with shearing interference and linear polarization filtering, the edge image is output. This scheme can achieve broadband operation by relying on the broadband characteristics of geometric phase, but the optical path combination and polarization control structure are complex, the effective numerical aperture of the device is limited, and it is difficult to simultaneously improve the imaging aperture and field of view. Another type of scheme, under coherent illumination conditions, integrates vortex phase and focusing phase into a metal lens, or superimposes Bessel function amplitude modulation, to integrate high-pass filtering and imaging functions in a single device, achieving single-order and multi-order edge detection. However, this type of scheme is strictly dependent on coherent light sources and cannot work stably in incoherent practical scenarios such as natural environments and ordinary lighting, resulting in extremely poor compatibility with applicable scenarios and environments.

[0006] Optical spatial differentiation schemes based on the Green's function method, by engineering the angle response of planar optical devices in real space, avoid explicit Fourier transform steps, further simplifying the system structure and reducing device size. These schemes have been validated on device platforms such as metallic metasurfaces, photonic crystals, and all-dielectric metasurfaces. For example, by controlling the guided-mode resonance characteristics of photonic crystal thin films, a secondary angle transmission response at a specific wavelength can be achieved, completing Laplace operator operations; or, quasi-continuum bound states (q-BIC) can be used to achieve Laplace differentiation functions in specific wavelength bands. However, these devices and schemes generally suffer from two major drawbacks: first, their operating conditions are highly dependent on coherent illumination; the differentiation effect degrades sharply after the absence of a coherent light source, making them unsuitable for practical scenarios such as everyday natural light and broadband incoherent light sources; second, their effective operating bandwidth is narrow, with a relatively narrow operating bandwidth (…). Limited by its limitations, it can only maintain stable differential performance within a specific narrow band, and its multispectral and wide-spectral adaptability is insufficient.

[0007] To overcome the dual limitations of coherence and operating bandwidth, existing technologies further explore optical spatial differentiation methods under incoherent illumination. These methods include designing the difference between visible light dual-wavelength optical transfer functions to satisfy the wave vector quadratic relationship to achieve incoherent differentiation; utilizing the incident angle-dependent transfer function difference to achieve second-order differentiation under dual narrowband incoherent light sources; relying on the angular gradient and geometric phase effect of the Fresnel coefficient of a uniaxial crystal to achieve tunable first-order and second-order differentiation; and constructing polarization-dependent point diffusion functions based on the anisotropy of metasurface nanopillars and achieving incoherent edge imaging through polarization difference. The above-mentioned solutions have achieved breakthroughs in incoherent operating conditions, but there are still irreconcilable technical contradictions: the solutions based on crystal tilt angle and polarization control are extremely sensitive to the alignment accuracy of the incident angle and device tilt angle. Assembly and adjustment errors and small offsets during use will significantly reduce the differential accuracy, resulting in insufficient stability and robustness; various wide-spectrum, incoherent solutions generally cannot achieve the best overall balance between incoherent operation, wide spectral response, large imaging aperture and high computational efficiency. They either sacrifice spectral width for stability or reduce the device aperture to ensure differential effect. High efficiency and large aperture are difficult to achieve simultaneously, and the overall performance cannot meet the comprehensive requirements of multi-scenario adaptability, high integration and high stability in practical engineering applications.

[0008] In summary, existing optical spatial differentiation and related image processing technologies generally suffer from single or multiple technical defects, such as limitations in system size and integration, coherence of illumination sources, operating bandwidth, dependence on alignment accuracy, and balance between aperture and efficiency. These limitations make it difficult to simultaneously meet the comprehensive application requirements of incoherent broadband adaptation, miniaturized integration, large aperture, high efficiency, and high robustness, thus restricting the implementation and large-scale application of such technologies in fields such as real-time machine vision, vehicle imaging, and portable inspection equipment. Summary of the Invention

[0009] Therefore, this invention aims to solve the technical problems of existing optical edge imaging technology, which struggles to achieve an optimal balance between incoherent illumination, wide spectral response, large aperture applications, high efficiency, and functional dimensions (full color and polarization information), and suffers from defects such as dependence on coherent illumination, narrow spectral bandwidth, complex system structure, or sensitivity to angle alignment.

[0010] To address the aforementioned technical problems, this invention provides an incoherent full-color polarization edge imaging system and method based on reverse-designed photonic crystals. The incoherent full-color polarization edge imaging system includes: an incoherent light source module, a polarization-correlated one-dimensional photonic crystal device, an imaging detection module, and an image processing module. The incoherent light source module is used to provide incoherent illumination light in the visible light band of the target; The polarization-dependent one-dimensional photonic crystal device is composed of different all-dielectric materials stacked along the direction perpendicular to the light propagation direction, and is used to generate differentiated optical transfer functions for different linearly polarized light. The imaging detection module is used to collect light intensity image data after passing through the polarization-correlated one-dimensional photonic crystal device when arbitrary linearly polarized light and its orthogonally linearly polarized light are incident. The image processing module is used to perform differential operations on a set of light intensity image data corresponding to orthogonal polarization states, and output a first-order spatial differential edge enhancement image. In this process, the thickness of each dielectric layer in the polarization-dependent one-dimensional photonic crystal device is used as a trainable parameter. An iterative optimization algorithm based on error backpropagation is used to perform optimization. The optimization process introduces a nonlinear activation function as a boundary constraint operator to limit the thickness of a single dielectric layer to no less than a preset minimum value, thereby obtaining the optimal thickness of each dielectric layer.

[0011] In one embodiment of the present invention, obtaining the optimal thickness of each dielectric layer includes: Within the target visible light band, multiple discrete sampling wavelengths are selected according to a preset wavelength step size, and a target differential optical transfer function is constructed for each discrete sampling wavelength, so that the target differential optical transfer function satisfies a preset linear proportional relationship with the spatial angular frequency in the spatial frequency domain. Based on the gradient descent optimization algorithm, the optimization objective is to minimize the deviation between the actual differential optical transfer function and the corresponding target differential optical transfer function of the polarization-dependent one-dimensional photonic crystal device at each discrete sampling wavelength. The trainable parameters are iteratively updated, and the optimal thickness of each dielectric layer in the polarization-dependent one-dimensional photonic crystal device is determined through multiple rounds of convergence calculation.

[0012] In one embodiment of the present invention, the loss function for iteratively optimizing the trainable parameters is as follows: , in, The sampling wavelength within the target visible light band. The angle of incidence of the incident light; and To represent the optimization process, at the th... Each sampling wavelength and angle of incidence Under the given conditions, the optical transfer functions of the first linearly polarized light and the second linearly polarized light, which are orthogonal to it, are... To match the sampling wavelength The corresponding adaptive coefficients; Indicates the sampling wavelength sequence number; This indicates the absolute value operation; N is the total number of discrete sampling wavelengths within the target visible light band.

[0013] In one embodiment of the present invention, the first linearly polarized light is TE-polarized light and the second linearly polarized light is TM-polarized light.

[0014] In one embodiment of the present invention, the target differential optical transfer function In the spatial frequency domain, it is related to the spatial angular frequency. The preset linear proportional relationship is satisfied as follows: , in, It represents the sign of direct proportion.

[0015] In one embodiment of the present invention, the target differential optical transfer function The expression is as follows: , in, and Let represent the optical transfer functions corresponding to arbitrary linearly polarized light and its orthogonally polarized light, respectively. These two represent a set of optical transfer functions corresponding to orthogonal polarization states. This indicates the operation of taking the absolute value.

[0016] In one embodiment of the present invention, the nonlinear activation function is the ReLU function.

[0017] In one embodiment of the present invention, the system further includes a polarization control module, which is disposed in the imaging optical path between the incoherent light source module and the object to be imaged, for generating linearly polarized light in a specified direction and synchronously outputting its orthogonally linearly polarized light.

[0018] Based on the same inventive concept, this invention also provides an incoherent full-color polarization edge imaging method based on reverse-designed photonic crystals. Utilizing the aforementioned incoherent full-color polarization edge imaging system, the method includes the following steps: S1: Perform incoherent illumination on the surface of the target object to be imaged, and acquire at least one set of intermediate light intensity images corresponding to each of the orthogonal linear polarization states; S2: Perform spatial domain difference calculation on the intermediate light intensity image under the orthogonal polarization state. Through the differential modulation of polarization information, make the absolute difference of the optical transfer function corresponding to the orthogonal linearly polarized light satisfy the first-order spatial differential condition, and output the first-order spatial differential edge enhancement image of the surface under test.

[0019] In one embodiment of the present invention, the mathematical representation of the first-order spatial differential edge enhancement image in the frequency domain is as follows: , in, It is the spatial angular frequency. This represents the frequency domain light intensity distribution of the edge-enhanced image output after the first-order optical spatial differentiation operation. This represents the frequency domain light intensity distribution of the target surface to be imaged. This is the differential optical transfer function.

[0020] Compared with the prior art, the above-described technical solution of the present invention has the following advantages: The incoherent full-color polarization edge imaging technology proposed in this invention, based on reverse-designed photonic crystals, has the core advantage of overcoming the trade-offs between spectral bandwidth, coherence, and efficiency in existing technologies. It employs a one-dimensional photonic crystal structure stacked from all dielectric materials, optimized using a gradient descent algorithm based on backpropagation of errors. This allows for stable operation across the entire visible light spectrum (400nm-700nm) under arbitrary incoherent illumination conditions, without requiring special light sources or additional conditions. It not only achieves high-contrast full-color edge imaging but also innovatively integrates polarization edge imaging functionality, introducing more dimensional information for edge detection. Furthermore, the device can be fabricated with a large aperture (up to meter-level), combining compact structure, strong integration potential, high computational efficiency, and fast convergence. It exhibits excellent performance in various scenarios, including microscopic imaging, telescopic imaging, dynamic scene capture, and biomedical imaging, with a spatial resolution greater than 2µm. Experimental results are highly consistent with numerical simulations, providing a robust and feasible technical path for the practical application of large-aperture, broadband analog optical processors. Attached Figure Description

[0021] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0022] Figure 1 This is a schematic diagram of the structure of an incoherent full-color polarization edge imaging system provided in an embodiment of the present invention; Figure 2 This is a characterization of the optical response of a polarization-dependent one-dimensional photonic crystal and a verification of its full-color optical spatial differential performance. (a) and (b) represent the numerical simulation results of transmittance as a function of wavelength and incident angle under TE polarization and TM polarization incident conditions, respectively; (c) and (d) represent the experimental measurement results of transmittance as a function of wavelength and incident angle under TE polarization and TM polarization incident conditions, respectively; (e) and (f) represent the numerical simulation and experimental measurement results of the optical transfer function difference (DOTF) in the wavelength range of 400 nm to 700 nm, respectively; (g) and (h) represent the line scan curves of the numerical simulation and experimental measurement results of DOTF at seven discrete wavelengths selected at 50 nm intervals in the wavelength range of 400 nm to 700 nm, respectively. Figure 3 The original image of the barcode pattern and its edge imaging results under 400nm wavelength conditions are presented. Figure 4 Numerical simulation results of barcode pattern edge imaging under wavelength conditions of 450nm and 500nm; Figure 5 Numerical simulation results of barcode pattern edge imaging under wavelength conditions of 550nm and 600nm; Figure 6 Numerical simulation results of barcode pattern edge imaging under wavelength conditions of 650nm and 700nm; Figure 7 This paper presents the characterization, microscopic imaging system, and full-color optical edge imaging experimental results of polarization-correlated one-dimensional photonic crystal (PDPC) devices. (a) shows the optical path and structure of the PDPC-based optical microscopic edge imaging experimental device under incoherent illumination conditions; (b) shows a large-size polarization-correlated one-dimensional photonic crystal physical sample prepared using electron beam evaporation; (c) shows the cross-sectional scanning electron microscope (SEM) morphology of the prepared PDPC; (d) and (e) show the original bright-field imaging results of various biological cell samples under TE and TM polarization illumination, respectively; and (f) shows the full-color optical edge extraction image of the biological cell sample obtained after differential operation on the orthogonal polarized bright-field image. Figure 8 The results show the application performance verification of the polarization differential optical edge imaging system based on polarization-correlated one-dimensional photonic crystal (PDPC) in outdoor far-field imaging scenarios; where (b1) is the original bright-field image of the scene, and (b2) is the corresponding generated full-color edge-enhanced image. Figure 9 This is a verification of the polarization edge imaging application of the PDPC-based polarization differential edge imaging system in dynamic scenes, where (a) represents the bright field image of the moving object; (b) represents the polarization edge image of the object obtained through the 0° and 90° polarization channels; and (c) represents the polarization edge image of the object obtained through the 45° and 135° polarization channels.

[0023] Explanation of reference numerals in the accompanying drawings: 10, Incoherent light source module; 20, Polarization control module; 30, Polarization-correlated one-dimensional photonic crystal device; 40, Imaging detection module; 50, Image processing module; 100, Object to be imaged. Detailed Implementation

[0024] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0025] In incoherent optical systems, the frequency domain correlation between the output image and the input object follows a linear modulation law, and its mathematical model satisfies formula (1): (1) in, The spatial angular frequency describes the rate of spatial change in light intensity distribution; Optical transfer function (OTF) characterizes the ability of a system to transfer different spatial frequency components; The frequency domain light intensity distribution of the input object. This represents the frequency domain light intensity distribution of the output image.

[0026] According to Fourier optics theory, due to Amplitude Transfer Function (ATF) The autocorrelation function of is expressed by formula (2): (2) Where * denotes the complex conjugate operation, and N is the normalization coefficient. This can be derived from the Cauchy–Schwarz inequality. Always non-negative and at the zero spatial frequency point ( The inherent characteristic of achieving a global maximum conflictes with the core requirement of "complete suppression of zero-frequency components" for edge enhancement imaging, making it impossible for traditional single-channel OTFs to perform optical spatial differentiation operations under incoherent illumination modes.

[0027] To overcome the physical limitations of optical spatial differentiation under incoherent illumination, this invention proposes a polarization differential modulation technique. A first-order spatial differentiation mechanism is constructed by calculating the difference in optical transfer functions of orthogonal polarization states. Its frequency domain modulation model is shown in equation (3): (3) Where abs is the absolute value operation. and These are the OTFs of arbitrary linearly polarized light and its orthogonally polarized light, respectively. It is the polarization difference optical transfer function (DOTF), which is the absolute difference between the two polarization states OTF.

[0028] The key to achieving first-order optical spatial differentiation lies in the polarization difference optical transfer function. It needs to satisfy a relationship that is proportional to the spatial angular frequency k, as shown in formula (4): (4)

[0029] The target DOTF defined in the above formula (4) can be realized through a one-dimensional photonic crystal structure, which has comprehensive technical advantages in terms of precise control of polarization-angle response, adaptability to large aperture, and compatibility with batch micro-nano fabrication. However, traditional intelligent optimization algorithms (particle swarm optimization, genetic algorithm) have multi-objective optimization bottlenecks and cannot simultaneously meet the dual design indicators of visible full-band broadband response and linear correlation between orthogonal polarization OTF difference and incident angle, making it difficult to match the performance requirements of the imaging system.

[0030] To address the aforementioned challenges in structural design and algorithm adaptation, this invention discloses an incoherent illumination full-color polarization edge imaging system, such as... Figure 1 As shown, its optical path structure includes, in sequence according to the incoherent light propagation order: incoherent light source module 10, polarization control module 20, polarization-correlated one-dimensional photonic crystal device 30, imaging detection module 40, and image processing module 50. The incoherent light source module 10 is used to provide incoherent illumination light in the target visible light band (400nm~700nm in this embodiment). It can be an indoor fluorescent lamp, outdoor natural light or tungsten halide light source (such as the Thorlabs SLS201L model) and can work stably without the need for a special coherent light source. The polarization control module 20 is a linear polarizer, which is set in the imaging optical path between the incoherent light source module 10 and the object to be imaged 100. It is used to generate linearly polarized light in a specified direction and output its orthogonal linearly polarized light, such as TE / TM polarization state, to provide dual polarization channel input for subsequent polarization difference calculation. The polarization-dependent one-dimensional photonic crystal device 30 is disposed between the object to be imaged 100 and the imaging detection module 40, and is composed of different all-dielectric materials stacked along the direction perpendicular to the light propagation direction. The imaging detection module 40 is used to acquire image data after passing through the polarization-correlated one-dimensional photonic crystal device under different polarization states. It can select a microscopic imaging component (long working distance objective lens + tube lens + CCD, such as Mitutoyo MY20X-804 objective lens + AVT GT3400CCCD) or a telescopic imaging component (industrial camera + telescope lens, such as DaHengMER-1810-21U3C camera + LOTS SV-2514V lens) according to the application scenario. The image processing module 50 is used to perform differential operations on the image data under different polarization states. Through differential modulation of polarization information, the absolute difference of orthogonal polarization OTF satisfies the first-order spatial differential condition, and finally outputs an edge-enhanced image. Among them, the thickness of each layer of the polarization-dependent one-dimensional photonic crystal device 30 is used as a trainable parameter, and an iterative optimization algorithm based on error backpropagation is used to perform optimization. The process includes: Within the target visible light band, N discrete sampling wavelengths (set to 15 sampling points in this embodiment) are selected according to a preset step size of 20 nm, and a target differential optical transfer function is set for each discrete sampling wavelength. : , and Let represent the optical transfer functions of arbitrary linearly polarized light and its orthogonally polarized light, respectively. This indicates the absolute value operation; The ReLU activation function is introduced as a boundary constraint operator to limit the thickness of a single-layer medium to no less than a preset minimum value (5nm), thus avoiding the limits of micro-nano fabrication technology. The global loss function is defined as shown in Equation (5) to minimize the deviation between the absolute difference of orthogonal polarization transmittance at each sampling wavelength and the linear response of the incident angle: (5), in, The sampling wavelength within the target visible light band. The angle of incidence of the incident light; and To represent the optimization process, at the th... Each sampling wavelength and angle of incidence Under certain conditions, the optical transfer function (OTF) of the first linearly polarized light and the second linearly polarized light orthogonal to it can be optionally defined as follows: the first linearly polarized light is TE polarized light and the second linearly polarized light is TM polarized light. To match the sampling wavelength The corresponding adaptive coefficients; Indicates the sampling wavelength number.

[0031] Through multiple rounds of iterative convergence, the absolute difference in transmittance of orthogonally linearly polarized light in the PDPC within the full wavelength range of 400nm~700nm and the ±30° incident angle range is strictly linearly related to the incident angle, satisfying the target DOTF requirement defined by formula (4), and thus obtaining the optimal thickness of each layer of medium in the polarization-dependent one-dimensional photonic crystal device 30.

[0032] Optionally, the polarization-dependent one-dimensional photonic crystal device 30 may use GaN, Al2O3, HfO2, SiN, SiC, TiO2, SiO2, etc. as stacked dielectric materials. These dielectric materials all have high transmittance and negligible intrinsic absorption loss in the visible light band of 400nm~700nm.

[0033] To characterize the angle-resolved transmission characteristics of the designed polarization-dependent one-dimensional photonic crystal device 30 nm, numerical simulations of the transmission spectra of transverse electric (TE) and transverse magnetic (TM) polarization modes in the visible light band of 400–700 nm were performed based on rigorous coupled-wave analysis (RCWA). The results are as follows: Figure 2 (a) Figure 2 As shown in (b). Given that the PDPC structure is isotropic in-plane, the TE and TM polarization modes can be equivalent to arbitrary orthogonal linear polarization components, laying the physical foundation for constructing the polarization difference optical transfer function (DOTF).

[0034] To verify the accuracy of the theoretical model, numerical simulations of the transmittance as a function of wavelength and incident angle under TE and TM polarization states were performed based on rigorous coupled-wave analysis (RCWA). The results are as follows: Figure 2 (a) Figure 2 As shown in (b). Using an Agilent Cary 7000 UV-Vis-NIR spectrophotometer, the angle-resolved transmission spectra under TE and TM polarization states were measured in the same visible light band with an angular sampling step of 1°. The corresponding results are shown in (b). Figure 2 (c) Figure 2 (d) Comparative analysis shows that the measured transmission spectrum and the numerical simulation results are in high agreement, which fully verifies the reliability and accuracy of optical modeling and numerical calculation.

[0035] Based on the simulated and measured angle-resolved transmission data, a two-dimensional distribution of the polarization difference optical transfer function (DOTF) was obtained numerically over the entire visible light band (400 nm to 700 nm) and a wide incident angle range. The numerical simulation results are as follows: Figure 2 As shown in (e), the experimental measurement results are as follows: Figure 2 As shown in (f).

[0036] To quantify its linear response characteristics, seven discrete characteristic wavelengths were selected in the range of 400nm to 700nm at 50nm intervals. The DOTF variation curves with incident angle at each wavelength were extracted. The numerical simulation line scan curves are shown below. Figure 2 As shown in (g), the experimental measurement line scan curve is as follows: Figure 2 As shown in (h). The linearity of the above curves was quantitatively characterized by linear fitting. The results show that within the ±30° incident angle range (shaded area in the figure), the simulated and measured DOTFs corresponding to the seven discrete characteristic wavelengths all exhibit excellent linear response relationships, and this linear characteristic completely covers the entire visible light operating band.

[0037] Furthermore, based on the constructed DOTF, optical spatial differentiation simulation reconstruction of target images containing standard barcode patterns is performed in the 400nm–700nm full visible light band. Figures 3 to 6Numerical simulation results of edge imaging under different monochromatic wavelengths are presented. The reconstruction process is achieved by performing a spatial convolution operation between the input original image and the DOTF of the PDPC, successfully obtaining a first-order spatial differential image of the target in the entire visible light band. The edge contours are clear and the contrast is excellent, verifying the theoretical feasibility and broadband applicability of this structure to achieve optical edge enhancement under broadband and incoherent illumination conditions.

[0038] This study employed electron beam evaporation deposition (EBD) to fabricate PDPC devices at the micro- and nano-scale levels. During deposition, an online optical reflectivity monitoring module was used to calibrate and control the thickness of each monolayer in real time, ensuring that the physical thickness of each functional layer was highly consistent with the target parameters of the reverse optimization design.

[0039] This fabrication process successfully enabled the fabrication of large-area PDPC elements with an effective light transmission aperture of 170 mm, such as... Figure 7 As shown in (b) to (c), its effective light transmission scale far exceeds the fabrication limit of conventional metasurface structures, highlighting the engineering application advantages of this invention in large-aperture, lightweight optical imaging systems. Figure 7 Image (c) shows the cross-sectional morphology of the fabricated PDPC device using a scanning electron microscope (SEM). The high-contrast bright areas in the image correspond to high-refractive-index films, while the low-contrast dark areas correspond to low-refractive-index dielectric films. The periodically stacked structure has clear interfaces, uniform film thickness, and excellent overall structural integrity, meeting the requirements for fabricating high-performance optical devices.

[0040] To verify the feasibility of this PDPC device achieving incoherent full-color first-order optical spatial differentiation in the full visible light band of 400nm~700nm, a system was constructed as follows: Figure 7 The incoherent polarization differential microscopy imaging optical path system is shown in (a). The system uses a Thorlabs SLS201L tungsten halide white light source as the incoherent illumination source and achieves precise gating of the working band through an adjustable filter component. The full-color imaging experiment uses a combination of FELH0400 long-pass filter and FESH0700 short-pass filter to achieve full coverage of the continuous visible light band from 400 to 700 nm. The monochromatic spot frequency imaging experiment uses narrow-band pass filters (models FLH633-5, FLH532-10, and FBH450-10) with center wavelengths of 633 nm, 532 nm, and 450 nm respectively to achieve independent testing of the red, green, and blue primary color bands.

[0041] The filtered collimated beam is focused onto the sample surface by an achromatic lens (Thorlabs AC508-080-A-ML), and the polarization state of the incident light is continuously controlled by a rotating high-precision linear polarizer (Thorlabs LPVISE200-A). The sample (biological cell slices, etc.) is placed in the object-side focal plane region of a Mitutoyo MY20X-804 long working distance microscope objective, with a PDPC device integrated in the conjugate optical path between the sample and the objective. The transmitted light field is focused by a Daheng Optoelectronics GCO-210012 imaging tube lens, and the image is digitized and stored by an AVT GT3400C color charge-coupled device (CCD), constructing a complete incoherent full-color polarization edge imaging link.

[0042] To expand application scenarios and verify the adaptability of biomedical label-free imaging, label-free edge feature extraction experiments were carried out on various biological samples based on the system. The test samples included various colored biological tissues such as Rhizopus spore cells, pancreatic cells, and onion epidermal cells. Figure 7 (d) and Figure 7 Image (e) shows the original bright-field images of cells acquired by the TE and TM polarization channels under conventional broadband white light illumination. After performing spatial difference operations on the two sets of orthogonal polarization images, the full-color edge enhancement result is obtained as shown below. Figure 7 As shown in (f), without external fluorescent labeling, special coherent light sources, or complex digital post-processing, the system can directly and clearly extract cell outlines, boundaries, and fine internal structural features, demonstrating its outstanding practical application potential in scenarios such as label-free bioimaging, rapid observation of pathological sections, and in vivo microscopic detection.

[0043] Furthermore, the optical edge imaging scheme based on polarization-dependent photonic crystal (PDPC) proposed in this embodiment is directly compatible with and can be extended to conventional photography systems and telescopic far-field imaging systems. The imaging acquisition unit uses a Daheng MER-1810-21U3C industrial camera, paired with a LOTSSV-2514V optical lens, directly integrating the PDPC device into the object-side front end of the imaging lens to complete the construction of a compact polarization differential imaging module.

[0044] Figure 8 In the image, (b1) is the original bright-field color image of the outdoor far-field scene, and (b2) is the full-color edge-enhanced image obtained by the corresponding orthogonal polarization difference operation. Experimental results show that weak texture structures such as trees and clouds in natural scenes, as well as the fine geometric contours of buildings, can be highlighted with high contrast and accurately resolved in the edge image. Moreover, the color information of the original scene is preserved without distortion and with high fidelity. This verifies the practicality and color edge fidelity capability of this scheme in engineering scenarios such as outdoor far-field incoherent imaging, natural scene observation, vehicle vision, and security monitoring.

[0045] Figure 7 The device shown in (a) can simultaneously perform real-time polarization state analysis and single-exposure edge imaging, making it suitable for video acquisition in high-speed dynamic scenes. To verify this performance, a PDPC device was integrated into the front end of the lens of a FLIR BFS-U3-51S5P-C polarization camera with four linear polarization detection channels at 0°, 90°, 45°, and 135°, covering a working wavelength range of 400–900 nm, enabling direct recording of dynamic target videos.

[0046] Figure 9 (a) shows typical bright-field raw frame images of a toy car moving on a smooth reflective substrate in a dynamic scene at times 1s, 3s and 7s of a video sequence. Figure 9 Image (b) shows the edge enhancement image obtained after polarization difference calculation of images acquired through the 0° and 90° orthogonal polarization channels; using the same difference principle, the edge imaging result calculated based on the 45° and 135° orthogonal polarization channel images is shown below. Figure 9 As shown in (c). Experimental results show that the edge images obtained by selecting different orthogonal polarization basis vectors have significant feature differences. The virtual image of the toy car corresponding to the specular reflection formed by the smooth substrate is... Figure 9 In (b), the area clearly distinguishable is marked by the dashed box, while... Figure 9 In (c), the reflection artifact is almost completely suppressed.

[0047] The aforementioned phenomenon stems from the significant linear polarization characteristics of the specular reflection component of a smooth interface, with different orthogonal polarization groups exhibiting marked differences in their responses to this type of polarization-sensitive reflection component. The experimental results confirm that the PDPC polarization differential imaging method proposed in this invention can simultaneously extract target geometric edges and interpret polarization features, possessing polarization-edge dual-mode collaborative imaging capabilities. It can effectively suppress specular reflection interference and improve the robustness of target recognition in complex scenes.

[0048] Based on the above simulation and experimental results, the PDPC device and imaging system proposed in this invention can stably achieve incoherent full-color first-order optical spatial differentiation and high-contrast edge enhancement under any incoherent illumination conditions in the full visible light band of 400nm~700nm. It exhibits excellent performance in terms of broadband spectral adaptability, angular linear response, spatial resolution, large-aperture fabrication, integration level, and environmental robustness. It effectively overcomes the bottlenecks of traditional technologies in terms of coherence limitation, narrow bandwidth, aperture limitation, and alignment sensitivity, and provides a stable, reliable, and engineerable physical implementation scheme for incoherent optical simulation computing, real-time machine vision, biomedical imaging, far-field observation and other fields.

[0049] Based on the same inventive concept as the aforementioned incoherent full-color polarization edge imaging system, this invention also provides an incoherent full-color polarization edge imaging method based on reverse-designed photonic crystals, utilizing the aforementioned incoherent full-color polarization edge imaging system, comprising the following steps: S1: The imaging detection module 40 performs incoherent illumination on the surface of the target object and performs imaging detection to acquire at least one set of intermediate light intensity images corresponding to each of the orthogonal linear polarization states (such as TE polarization state / TM polarization state). S2: Perform spatial domain difference calculation on the intermediate light intensity image under the orthogonal polarization state. Utilize the polarization difference characteristics of PDPC to ensure that the absolute difference of the orthogonal polarization OTF satisfies the first-order spatial differential condition, and output the first-order spatial differential edge enhancement image of the object surface under test. Its frequency domain mathematical representation is as follows: (6), in, This represents the frequency domain light intensity distribution of the output image after first-order optical spatial differentiation. It represents the original light intensity frequency domain distribution of the object surface under test, and fully describes the frequency domain modulation mechanism of edge enhancement imaging under incoherent illumination.

[0050] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0051] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0052] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0053] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0054] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. An incoherent full-color polarization edge imaging system based on reverse-designed photonic crystals, characterized in that, include: Incoherent light source module, used to provide incoherent illumination light in the visible light band of the target; A polarization-dependent one-dimensional photonic crystal device, wherein the polarization-dependent one-dimensional photonic crystal device is composed of different all-dielectric materials stacked along a direction perpendicular to the light propagation direction, and is used to generate differentiated optical transfer functions for different linearly polarized light; The imaging detection module is used to acquire light intensity image data after passing through the polarization-correlated one-dimensional photonic crystal device when arbitrary linearly polarized light and its orthogonally linearly polarized light are incident. And an image processing module, used to perform differential operations on a set of light intensity image data corresponding to orthogonal polarization states, and output a first-order spatial differential edge enhancement image; In this process, the thickness of each dielectric layer in the polarization-dependent one-dimensional photonic crystal device is used as a trainable parameter. An iterative optimization algorithm based on error backpropagation is used to perform optimization. The optimization process introduces a nonlinear activation function as a boundary constraint operator to limit the thickness of a single dielectric layer to no less than a preset minimum value, thereby obtaining the optimal thickness of each dielectric layer.

2. The incoherent full-color polarization edge imaging system based on reverse-designed photonic crystals according to claim 1, characterized in that: To obtain the optimal thickness of each dielectric layer, including: Within the target visible light band, multiple discrete sampling wavelengths are selected according to a preset wavelength step size, and a target differential optical transfer function is constructed for each discrete sampling wavelength, so that the target differential optical transfer function satisfies a preset linear proportional relationship with the spatial angular frequency in the spatial frequency domain. Based on the gradient descent optimization algorithm, the optimization objective is to minimize the deviation between the actual differential optical transfer function and the corresponding target differential optical transfer function of the polarization-dependent one-dimensional photonic crystal device at each discrete sampling wavelength. The trainable parameters are iteratively updated, and the optimal thickness of each dielectric layer in the polarization-dependent one-dimensional photonic crystal device is determined through multiple rounds of convergence calculation.

3. The incoherent full-color polarization edge imaging system based on reverse-designed photonic crystals according to claim 2, characterized in that: The loss function for iteratively optimizing the trainable parameters is as follows: ， in, The sampling wavelength within the target visible light band. The angle of incidence of the incident light; and To represent the optimization process, at the th... Each sampling wavelength and angle of incidence Under the given conditions, the optical transfer functions of the first linearly polarized light and the second linearly polarized light, which are orthogonal to it, are... To match the sampling wavelength The corresponding adaptive coefficients; Indicates the sampling wavelength sequence number; This indicates the absolute value operation; N is the total number of discrete sampling wavelengths within the target visible light band.

4. The incoherent full-color polarization edge imaging system based on reverse-designed photonic crystals according to claim 3, characterized in that: The first linearly polarized light is TE-polarized light, and the second linearly polarized light is TM-polarized light.

5. The incoherent full-color polarization edge imaging system based on reverse-designed photonic crystals according to claim 2, characterized in that: The target differential optical transfer function In the spatial frequency domain, it is related to the spatial angular frequency. The preset linear proportional relationship is satisfied, as follows: ， in, It represents the sign of direct proportion.

6. The incoherent full-color polarization edge imaging system based on reverse-designed photonic crystals according to claim 5, characterized in that: The target differential optical transfer function The expression is as follows: ， in, and Let represent the optical transfer functions corresponding to arbitrary linearly polarized light and its orthogonally polarized light, respectively. These two represent a set of optical transfer functions corresponding to orthogonal polarization states. This indicates the operation of taking the absolute value.

7. The incoherent full-color polarization edge imaging system based on reverse-designed photonic crystals according to claim 1, characterized in that: The nonlinear activation function is the ReLU function.

8. The incoherent full-color polarization edge imaging system based on reverse-designed photonic crystals according to claim 1, characterized in that: The system also includes a polarization control module, which is disposed in the imaging optical path between the incoherent light source module and the object to be imaged, and is used to generate linearly polarized light in a specified direction and synchronously output its orthogonally linearly polarized light.

9. A method for incoherent full-color polarization edge imaging based on reverse-designed photonic crystals, characterized in that, The method, using the incoherent full-color polarization edge imaging system as described in any one of claims 1 to 8, comprises the following steps: S1: Perform incoherent illumination on the surface of the target object to be imaged, and acquire at least one set of intermediate light intensity images corresponding to each of the orthogonal linear polarization states; S2: Perform spatial domain difference calculation on the intermediate light intensity image under the orthogonal polarization state. Through the differential modulation of polarization information, make the absolute difference of the optical transfer function corresponding to the orthogonal linearly polarized light satisfy the first-order spatial differential condition, and output the first-order spatial differential edge enhancement image of the surface under test.

10. The incoherent full-color polarization edge imaging method based on reverse-designed photonic crystals according to claim 9, characterized in that: The mathematical representation of the first-order spatial differential edge enhancement image in the frequency domain is as follows: ， in, It is the spatial angular frequency. This represents the frequency domain light intensity distribution of the edge-enhanced image output after the first-order optical spatial differentiation operation. This represents the frequency domain light intensity distribution of the target surface to be imaged. This is the differential optical transfer function.

Images