Method for establishing coherent noise sample data set in microstructure measurement

Abstract

The invention relates to the technical field of microstructure measurement, in particular to a method for establishing a coherent noise sample data set in microstructure measurement, which comprises the following steps of: S1, initially setting system parameter definition; and S2, optical element modeling parameterization. A digital holographic microscopic simulation system is established based on a scalar diffraction theory, a complete transmission process of a whole light path is fully considered, parameters of a light source, a lens and a measured object are incorporated into the simulation system, a USAF1951 standard target is used as a standard substance, the USAF1951 standard target is simulated to generate a hologram and is reconstructed, and a holographic microscopic simulation result is obtained. Reconstruction results show that the reproduced three-dimensional structure basically coincides with a standard substance, main reasons for generating parasitic fringes in an experiment for measuring the microstructure under the transparent medium are analyzed by discussing that the overall resolution of the system and the resolution of a simulation system are consistent with theoretical analysis results and coincide with experimental results, and simulation is carried out on the parasitic fringes. Through comparison with parasitic fringes obtained by experiments, the two fringes are highly consistent in structure.

Description

Technical Field

[0001] The present invention relates to the technical field of microstructure measurement, and particularly to a method for establishing a coherent noise sample data set in microstructure measurement. Background Art

[0002] With the continuous development of modern science and technology, holographic microscopy technology has been increasingly widely used in the field of microstructure measurement. As a non-contact, high-resolution three-dimensional imaging technology, digital holographic microscopy technology is widely used in fields such as materials science, biomedicine, and optical manufacturing, and can accurately measure microstructures at the micron or even nanometer level. However, in practical applications, holographic microscopy measurement is often limited by coherent noise interference, such as parasitic fringes, speckle noise, etc. These noises will significantly reduce the quality of the reconstructed image and affect the accuracy and reliability of microstructure measurement. In recent years, the rapid development of deep learning technology has provided a new solution for holographic image denoising, and high-quality, standardized data sets are the key basis for deep learning model training.

[0003] Current digital holographic microscopy measurement systems still face many challenges in dealing with coherent noise. Traditional denoising methods usually rely on physical modeling and filtering algorithms. These methods lack in-depth analysis of the generation mechanism and characteristics of parasitic fringes, resulting in limited denoising effects and difficulty in dealing with complex and diverse noise structures. Most existing holographic image simulation technologies only focus on local or single system parameters, and do not comprehensively consider the overall transmission process of the optical path and the comprehensive influence of multiple factors such as light sources, lenses, and measured objects. The generated simulation data lacks high consistency with real experimental data. The data scale generated by existing systems is small, making it difficult to meet the needs of deep learning model training for large-scale, high-quality sample data, and at the same time, the adaptability and scalability to different holographic microscopy systems are low. Therefore, we provide a method for establishing a coherent noise sample data set in microstructure measurement. Summary of the Invention

[0004] Aiming at the above-mentioned shortcomings of the prior art, the first object of the present invention is to provide a method for establishing a coherent noise sample data set in microstructure measurement to solve the problems in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for establishing a coherent noise sample data set in microstructure measurement, comprising the following steps:

[0007] S1. Initial setting of system parameter definition;

[0008] S2. Parametrization of optical element modeling;

[0009] S3. Generation of light field for light source modeling;

[0010] S4. Object reference light field superposition;

[0011] S5. Parasitic fringe mechanism simulation and analysis;

[0012] S6. Dataset generation and management;

[0013] S7. Experimental platform construction and data acquisition;

[0014] S8. Dataset verification and application.

[0015] The present invention is further configured such that in the step S1, system parameter definition and initial setting:

[0016] S1.1. Select a helium-neon (He-Ne) laser with a wavelength of 632.8 nm, accurately measure and record the power, coherence length, and divergence angle of the light source;

[0017] S1.2. Select a microscope objective with a numerical aperture (NA) of 0.25 and a focal length set to 50 mm. The selection of the microscope objective affects the resolution and depth of field of the system;

[0018] S1.3. Select a CCD camera with a resolution of 2048×2048 pixels and a single pixel size of 0.801 µm. The quantum efficiency, dynamic range, and noise characteristics of the CCD need to be recorded in detail;

[0019] S1.4. According to the propagation distance and the size of the optical elements in the system, preliminarily set the width of the sampling window to 0.1 m and the sampling interval to 0.05 m;

[0020] S1.5. Build a Michelson digital holographic microscopy system in an optical laboratory to ensure the stability of the optical path. The optical table should be equipped with anti-vibration devices and carried out in a dust-free environment.

[0021] The present invention is further configured such that in the step S2, optical element modeling and parameterization:

[0022] S2.1. Use the complex amplitude transmittance formula to model the lens in the system. Set the transmittance of the lens to 1 inside the aperture and 0 outside the aperture, with a focal length of 50 mm, a numerical aperture NA = 0.25, and a radius set to 25 mm. The complex amplitude transmittance of the lens can be expressed as:

[0023]

[0024] where: τ is the transmittance of the lens, P(x, y) is the pupil function, which is 1 inside the aperture and 0 outside the aperture, f is the focal length of the lens, and the numerical aperture of the lens is NA = r / f, where r is the radius of the lens;

[0025] S2.2. Taking the USAF 1951 standard resolution target as the measurement object, construct its three-dimensional height matrix, convert the image of the standard target into a grayscale matrix of 1560×1560, and assign an appropriate height value (less than half the wavelength of the light source).

[0026] S2.3. Precisely model the reflection and transmission characteristics of the beam splitter (BS) and the mirror (M). The beam splitter divides the incident beam into a reference beam and an object beam. It is necessary to ensure that its reflectivity and transmittance meet the experimental requirements, usually set to a 50:50 splitting ratio.

[0027] S2.4. The beam expander BE is used to expand the incident beam, and the aperture limits the beam size. Set the expansion multiple of the beam expander to 10 times, and the aperture diameter of the aperture to 5 mm.

[0028] S2.5. Use the simulation software COMSOL Multiphysics to preliminarily verify the modeled optical components to ensure that the transfer function and phase modulation of each component meet the theoretical expectations.

[0029] The present invention is further set as follows: In the step S3, light source modeling and light field generation:

[0030] S3.1. Use the fundamental mode Gaussian beam model to simulate the laser light source. Set the central beam waist radius to 0.05 m, and the Rayleigh range to π*(0.05)^2 / (632.8e-9)≈3.93 m. Ensure that the beam is approximately parallel within the Rayleigh range. The light field of the fundamental mode Gaussian beam can be expressed as:

[0031]

[0032] In the formula: w0 is the central beam waist radius, λ is the wavelength of the light source, k = 2π / λ is the wave number (k mentioned in the following text is all the wave number), z R = 1 / 2kw0 is the Rayleigh range, is the beam width when the light wave propagates to z, R = z R (z / z R + z R / z) represents the radius of curvature of the equiphase surface, Ψ represents the phase factor. The propagation of the Gaussian light source within the Rayleigh range can be approximately considered as parallel light. It can be clearly seen from the formula that the larger the beam width, the longer the Rayleigh range, that is, the weaker the beam divergence;

[0033] S3.2. Use the Fresnel diffraction integral to simulate the propagation of the beam from the light source to the microscope objective lens under the paraxial approximation, and use the S-FFT method (based on a single Fourier transform) for fast calculation.

[0034] S3.3. During the propagation of the light beam, considering the phase and amplitude modulation of the light beam by the lens and the object, the light beam is adjusted accordingly through the complex amplitude transmittance function, simulating the focusing effect of the lens and the scattering of the object on the light field. The object is regarded as a material with a certain refractive index and specific thickness at different positions, and its complex amplitude transmittance can be expressed as:

[0035] t o (x,y) = exp[jkh(x,y)]·τ

[0036] In the formula: L(x,y) = n o h(x,y) is the optical path brought by the object height matrix h, n o is the refractive index, τ is the transmittance or reflectivity. Considering that the research in this paper is on a reflective holographic microscopy system, so τ is the reflectivity in this paper;

[0037] S3.4. Dynamically adjust the sampling window and sampling interval according to the propagation distance to ensure the information integrity of the light field at each propagation stage, and use the SBLAS method (linear convolution of the band-limited angular spectrum method) to selectively scale the sampling area;

[0038] S3.5. Compare the light field generated by simulation with the theoretical expectation to verify the accuracy of the phase and amplitude during the propagation of the light beam.

[0039] The present invention is further set as: In the step S4, object reference light field superposition:

[0040] S4.1. Respectively simulate the propagation paths of the object light and the reference light. After the object light is reflected by the object, obtain its light field distribution; the reference light is directly reflected by the reflector without passing through the object to maintain coherence;

[0041] S4.2. Consider the optical path difference between the object light and the reference light, especially in an off-axis system, the phase difference caused by the object-reference angle. Use the lens focal length and matrix coordinates to calculate the optical path difference at each point to form an optical path difference matrix. Since the simulated system is an off-axis system, the object light and the reference light are not completely symmetric, and there is a certain object-reference angle. Therefore, it is necessary to consider the optical path difference brought by this angle θ to accurately reflect the optical characteristics of the off-axis system. The optical path difference caused by the angle θ can be characterized by the superposition of the phase differences caused by horizontal tilt and vertical tilt:

[0042]

[0043] The phase modulation t brought by it tilt can be expressed as:

[0044]

[0045] where x and y are the coordinates of the observation plane, and θ is the angle between the object light and the reference light;

[0046] S4.3. Superimpose the object light field and the reference light field, calculate their interference intensity, and generate a hologram to ensure the accurate transmission of phase information during the superimposition process to reflect the three-dimensional structure of the object;

[0047] S4.4. Use the Fourier transform method to calculate the wrapped phase of the hologram to provide the necessary phase information for subsequent three-dimensional reconstruction;

[0048] S4.5. Verify whether the generated hologram accurately reflects the three-dimensional structure of the object by comparing it with the theoretical hologram.

[0049] The present invention is further configured as follows: In the step S5, simulation and analysis of the parasitic fringe mechanism:

[0050] S5.1. Based on the scalar diffraction theory, simulate the multiple reflection interference when the light beam passes through the transparent medium. Set the refractive index of the transparent medium to 1.5, and the thickness range to 0.5 mm to 4 mm to study its influence on the parasitic fringes;

[0051] S5.2. Calculate the optical path difference between the reflected light on the surface of the transparent medium and the object light. Set the focal length of the lens to 50 mm, and the thicknesses of the transparent medium to 1 mm, 2 mm, and 4 mm, and correspondingly generate the optical path difference matrices at different thicknesses. The optical path difference matrix can be expressed as:

[0052]

[0053] where f is the focal length of the lens, δ0 is the thickness of the transparent medium, and x and y are the coordinates on the observation plane of the lens;

[0054] S5.3. Introduce the optical path difference matrix into the hologram generation process, superimpose the influence of the parasitic fringes, and use the SBLAS method to accurately simulate the interference effect of the parasitic fringes to generate a hologram containing parasitic fringes;

[0055] S5.4. Reconstruct the generated hologram containing parasitic fringes, observe the fringe distribution and density changes in the reconstructed image, and analyze the formation law of the parasitic fringes and their relationship with the thickness of the transparent medium;

[0056] S5.5. Compare the parasitic fringes generated by simulation with the actual experimental results, verify the accuracy of the simulation model, and adjust the simulation parameters according to the comparison results to ensure the high consistency between the simulation and experimental data.

[0057] The present invention is further configured as follows: In the step S6, dataset generation and management:

[0058] S6.1. During the simulation process, systematically adjust key parameters such as the thickness of the transparent medium, the incident angle, and the object height to generate diverse hologram samples. Each group of samples corresponds to different coherent noise conditions, covering all possible situations in actual measurements.

[0059] S6.2. Introduce systematic noise and environmental noise, such as light source intensity fluctuations, CCD noise, and environmental vibrations, into the generated holograms. By adjusting the noise parameters, simulate the common types and intensities of coherent noise in actual measurements.

[0060] S6.3. Add detailed label information to each generated hologram sample, including the thickness of the transparent medium, the incident angle, the type and intensity of the noise, etc. Classify the data according to different noise conditions for subsequent deep learning model training.

[0061] S6.4. Organize and store the generated hologram samples using the efficient HDF5 data storage format, and establish an indexing and metadata management system to ensure the traceability and accessibility of the data, supporting the rapid retrieval and processing of large-scale datasets.

[0062] S6.5. Conduct quality checks on the generated dataset to ensure the integrity and consistency of the hologram samples. Eliminate samples with obvious simulation errors or abnormal noise to ensure the high quality and reliability of the dataset.

[0063] The present invention is further configured as follows: In step S7, during the data acquisition of the experimental platform setup:

[0064] S7.1. According to the design in step one, build a Michelson digital holographic microscopy system, ensuring that the positions of all optical elements (He-Ne laser, beam expander, aperture, beam splitter, microscopic objective, mirror, and CCD camera) are accurately aligned and the optical path is stable without jitter.

[0065] S7.2. Conduct optical calibration on the experimental platform, adjust the angles of the beam splitter and the mirror so that stable interference fringes are formed between the reference light and the object light on the CCD, and use a standard resolution target (USAF1951) for alignment.

[0066] S7.3. Conduct data acquisition in a laboratory environment with constant temperature, no vibration, and no dust. Use an anti-vibration table and a temperature control system to reduce the influence of the external environment on the experimental results.

[0067] S7.4. Use the set experimental platform to acquire holograms of the USAF1951 resolution target, and record hologram samples at different thicknesses of the transparent medium (0.5 mm, 1 mm, 2 mm, and 4 mm).

[0068] S7.5. Preprocess the collected holograms, including noise filtering, background subtraction, and image alignment, to ensure that the experimental data and simulation data are compared under the same standard.

[0069] The present invention is further configured as follows in step S8, the dataset verification application:

[0070] S8.1. Compare and analyze the holograms generated by simulation with the holograms collected experimentally, and evaluate the accuracy and reliability of the simulation model through visual comparison and quantitative indicators;

[0071] S8.2. Use the generated hologram dataset with coherent noise to train a convolutional neural network deep learning model for holographic denoising, and set training parameters and optimization algorithms;

[0072] S8.3. Use an independent experimental dataset to verify and test the trained deep learning model, evaluate the performance of the model in actual denoising tasks, and ensure that it has good generalization ability and robustness;

[0073] S8.4. According to the model training and test results, further expand and optimize the dataset, introduce more diverse noise types and measurement conditions, and improve the coverage of the dataset and the adaptability of the model;

[0074] S8.5. Apply the established dataset and the trained deep learning model to an actual digital holographic microscopy system, improve the quality and measurement accuracy of the holograms, provide technical support and training, and promote the wide use of the dataset in related research and applications.

[0075] Beneficial effects

[0076] Adopting the technical solution provided by the present invention, compared with the known public technology, it has the following beneficial effects:

[0077] 1. The present invention establishes a digital holographic microscopy simulation system based on the scalar diffraction theory, fully considering the complete transmission process of the entire optical path, and incorporating the parameters of the light source, lens, and the object to be measured into the simulation system. Using the USAF1951 standard target as the standard substance, holograms are simulated and reconstructed. The reconstruction results show that the reproduced three-dimensional structure is basically consistent with the standard substance. By discussing the overall resolution of the system, the resolution of the simulation system is consistent with the theoretical analysis results and also coincides with the experimental results. The main reasons for the generation of parasitic fringes in the experiment of measuring microstructures under transparent media are analyzed and simulated. By comparing with the parasitic fringes captured in the experiment, it can be seen that they are highly consistent in structure.

[0078] 2. The simulation system established by the present invention can quickly generate a large number of holograms with parasitic fringes, which provides a rich and efficient training data set for the deep learning model during the training process, thus greatly promoting the application and development of deep learning technology in the field of holographic denoising.

[0079] 3. By adjusting system parameters such as the light source structure, controlling the imaging distance, and setting optical elements, the simulation system of the present invention can also be extended to systems with other structures, such as dual-wavelength digital holographic microscopy systems, in-line digital holographic microscopy systems, pre-amplified digital holographic microscopy systems, etc., providing prior guidance and technical support for the development of holographic technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 Schematic diagram of the optical path structure of the Michelson digital holographic microscopy system for the method of establishing a coherent noise sample data set in the microstructure measurement of the present invention;

[0081] Figure 2 Schematic diagram of the three-dimensional structure of the reference substance for the method of establishing a coherent noise sample data set in the microstructure measurement of the present invention;

[0082] Figure 3 Schematic diagram of the simulated hologram and reconstruction result for the method of establishing a coherent noise sample data set in the microstructure measurement of the present invention;

[0083] Figure 4 Schematic diagram of the experimental platform of the digital holographic microscopy system for the method of establishing a coherent noise sample data set in the microstructure measurement of the present invention;

[0084] Figure 5 Schematic diagram of the experimental hologram and reconstruction result for the method of establishing a coherent noise sample data set in the microstructure measurement of the present invention;

[0085] Figure 6 Schematic diagram of the holographic simulation with parasitic fringes for the method of establishing a coherent noise sample data set in the microstructure measurement of the present invention;

[0086] Figure 7 Schematic diagram of the experimental result of the hologram with parasitic fringes for the method of establishing a coherent noise sample data set in the microstructure measurement of the present invention;

[0087] Figure 8 Schematic diagram of the comparison between the experimental and simulation results of parasitic fringes for the method of establishing a coherent noise sample data set in the microstructure measurement of the present invention;

[0088] Figure 9 Schematic diagram of the comparison between the experimental and simulation results under different thicknesses of transparent media for the method of establishing a coherent noise sample data set in the microstructure measurement of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0089] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0090] The following further describes the present invention with reference to embodiments.

[0091] Embodiment 1

[0092] As Figure 1 - Figure 9 shown, the present invention provides a solution: a method for establishing a coherent noise sample data set in microstructure measurement, including the following steps:

[0093] S1. Initial setting of system parameter definition;

[0094] S2. Parametric modeling of optical elements;

[0095] S3. Generation of light field in light source modeling;

[0096] S4. Superposition of object reference light fields;

[0097] S5. Simulation and analysis of parasitic fringe mechanism;

[0098] S6. Generation and management of data set;

[0099] S7. Building an experimental platform for data acquisition;

[0100] S8. Verification and application of data set;

[0101] In the step S1, initial setting of system parameter definition:

[0102] S1.1. Select a helium-neon (He-Ne) laser with a wavelength of 632.8 nm, accurately measure and record the power, coherence length, and divergence angle of the light source;

[0103] S1.2. Select a microscope objective lens with a numerical aperture (NA) of 0.25, set the focal length to 50 mm. The selection of the microscope objective lens affects the resolution and depth of field of the system;

[0104] S1.3. Select a CCD camera with a resolution of 2048×2048 pixels, and the size of a single pixel is 0.801 μm. The quantum efficiency, dynamic range, and noise characteristics of the CCD need to be recorded in detail;

[0105] S1.4. According to the propagation distance and the size of the system optical elements, preliminarily set the width of the sampling window to 0.1 m and the sampling interval to 0.05 m;

[0106] S1.5. Set up a Michelson digital holographic microscopy system in an optical laboratory to ensure the stability of the optical path. The optical table should be equipped with an anti-vibration device and operate in a clean room environment;

[0107] In the step S2, parameterization of optical element modeling:

[0108] S2.1. Model the lens in the system using the complex amplitude transmittance formula. Set the transmittance of the lens to 1 inside the aperture and 0 outside the aperture, with a focal length of 50 mm, a numerical aperture NA = 0.25, and a radius set to 25 mm. The complex amplitude transmittance of the lens can be expressed as:

[0109]

[0110] where: τ is the transmittance of the lens, P(x,y) is the pupil function, which is 1 inside the aperture and 0 outside the aperture, f is the focal length of the lens, and the numerical aperture of the lens is NA = r / f, where r is the radius of the lens;

[0111] S2.2. Take the USAF 1951 standard resolution target as the measurement object, construct its three-dimensional height matrix, convert the image of the standard target into a grayscale matrix of 1560×1560, and assign appropriate height values (less than half the wavelength of the light source);

[0112] S2.3. Accurately model the reflection and transmission characteristics of the beam splitter (BS) and the mirror (M). The beam splitter divides the incident beam into a reference beam and an object beam. It is necessary to ensure that its reflectivity and transmittance meet the experimental requirements, usually set to a 50:50 splitting ratio;

[0113] S2.4. The beam expander BE is used to expand the incident beam, and the aperture limits the beam size. Set the expansion multiple of the beam expander to 10 times, and the aperture diameter of the aperture to 5 mm;

[0114] S2.5. Conduct a preliminary verification of the modeled optical elements through the simulation software COMSOL Multiphysics to ensure that the transfer function and phase modulation of each element meet the theoretical expectations;

[0115] In the step S3, light field generation in light source modeling:

[0116] S3.1. Use the fundamental mode Gaussian beam model to simulate the laser light source. Set the central beam waist radius to 0.05 m and the Rayleigh range to π*(0.05)^2 / (632.8e-9) ≈ 3.93 m to ensure that the beam is approximately parallel within the Rayleigh range. The light field of the fundamental mode Gaussian beam can be expressed as:

[0117]

[0118] where: w0 is the central beam waist radius, λ is the wavelength of the light source, k = 2π / λ is the wave number (k mentioned in the following text is all wave number), z R = 1 / 2kw0 is the Rayleigh range, is the beam width when the light wave propagates to z, R = z R (z / z R + z R / z) represents the radius of curvature of the equiphase surface, Ψ represents the phase factor. The propagation of the Gaussian light source within the Rayleigh range can be approximately considered as parallel light. It can be clearly seen from the formula that the larger the beam width, the longer the Rayleigh range, that is, the weaker the beam divergence;

[0119] S3.2. Use the Fresnel diffraction integral to simulate the propagation of the light beam from the light source to the microscope objective under the paraxial approximation, and use the S-FFT method (based on a single Fourier transform) for fast calculation;

[0120] S3.3. During the propagation of the light beam, consider the phase and amplitude modulation of the light beam by the lens and the object, and make corresponding adjustments to the light beam through the complex amplitude transmittance function to simulate the focusing effect of the lens and the scattering of the object on the light field. The object is regarded as a material with a certain refractive index and a specific thickness at different positions, and its complex amplitude transmittance can be expressed as:

[0121] t o (x, y) = exp[jkh(x, y)]·τ

[0122] where: L(x, y) = n o h(x, y) is the optical path brought by the object height matrix h, n o is the refractive index, τ is the transmittance or reflectivity. In view of the fact that this paper studies a reflection holographic microscopy system, so τ is the reflectivity in this paper;

[0123] S3.4. Dynamically adjust the sampling window and sampling interval according to the propagation distance to ensure the information integrity of the light field at each propagation stage, and use the SBLAS method (linear convolution of the band-limited angular spectrum method) to selectively scale the sampling area;

[0124] S3.5. Compare the light field generated by the simulation with the theoretical expectation to verify the accuracy of the phase and amplitude during the propagation of the light beam;

[0125] In the step S4, object-reference light field superposition:

[0126] S4.1. Respectively simulate the propagation paths of the object light and the reference light. After the object light is reflected by the object, obtain its light field distribution; the reference light is directly reflected by the reflector without passing through the object to maintain coherence;

[0127] S4.2. Consider the optical path difference between the object beam and the reference beam. Especially in an off-axis system, due to the phase difference caused by the angle between the object and the reference, calculate the optical path difference at each point using the lens focal length and matrix coordinates to form an optical path difference matrix. Since the simulated system is an off-axis system, the object beam and the reference beam are not completely symmetric, and there is a certain angle between the object and the reference. Therefore, it is necessary to consider the optical path difference brought by this angle θ to accurately reflect the optical characteristics of the off-axis system. The optical path difference caused by the angle θ can be characterized by the superposition of the phase differences caused by horizontal tilt and vertical tilt:

[0128]

[0129] The resulting phase modulation t tilt can be expressed as:

[0130]

[0131] where x and y are the coordinates of the observation plane, and θ is the angle between the object beam and the reference beam;

[0132] S4.3. Superimpose the object beam optical field and the reference beam optical field, calculate their interference intensity, and generate a hologram ( Figure 3 (a)), ensuring the accurate transmission of phase information during the superposition process to reflect the three-dimensional structure of the object;

[0133] S4.4. Use the Fourier transform method to calculate the wrapped phase of the hologram ( Figure 3 (b)) to provide the necessary phase information for subsequent three-dimensional reconstruction;

[0134] S4.5. By comparing with the theoretical hologram, verify whether the generated hologram accurately reflects the three-dimensional structure of the object;

[0135] In the above-mentioned step S5, the simulation analysis of the parasitic fringe mechanism:

[0136] S5.1. Based on the scalar diffraction theory, simulate the multiple reflection interference when the light beam passes through a transparent medium ( Figure 6 (a)), set the refractive index of the transparent medium to 1.5, and the thickness range to 0.5 mm to 4 mm to study its influence on the parasitic fringes;

[0137] S5.2. Calculate the optical path difference between the reflected light on the surface of the transparent medium and the object beam. Set the lens focal length to 50 mm, and the thicknesses of the transparent medium to 1 mm, 2 mm, and 4 mm, and correspondingly generate optical path difference matrices at different thicknesses. The optical path difference matrix can be expressed as:

[0138]

[0139] Among them, f is the focal length of the lens, δ0 is the thickness of the transparent medium, and x and y are the coordinates on the lens observation plane;

[0140] S5.3. Introduce the optical path difference matrix into the hologram generation process, superimpose the influence of parasitic fringes, and use the SBLAS method to accurately simulate the interference effect of parasitic fringes to generate a hologram containing parasitic fringes ( Figure 6 (a));

[0141] S5.4. Reconstruct the generated hologram containing parasitic fringes, and observe the fringe distribution and density changes in the reconstructed image ( Figure 6 (b) and Figure 6 (c)), analyze the formation law of parasitic fringes and their relationship with the thickness of the transparent medium;

[0142] S5.5. Compare the parasitic fringes generated by simulation with the actual experimental results ( Figure 8 ), verify the accuracy of the simulation model, and adjust the simulation parameters according to the comparison results to ensure high consistency between the simulation and experimental data;

[0143] In the step S6, dataset generation and management:

[0144] S6.1. During the simulation process, systematically adjust key parameters such as the thickness of the transparent medium, the incident angle, and the object height to generate diverse hologram samples. Each group of samples corresponds to different coherent noise conditions, covering various possible situations in actual measurements;

[0145] S6.2. Introduce system noise and environmental noise, such as light source intensity fluctuations, CCD noise, and environmental vibrations, into the generated holograms. By adjusting the noise parameters, simulate the common types and intensities of coherent noise in actual measurements;

[0146] S6.3. Add detailed label information to each generated hologram sample, including the thickness of the transparent medium, the incident angle, the noise type and intensity, etc. Classify the data according to different noise conditions for subsequent deep learning model training;

[0147] S6.4. Organize and store the generated hologram samples using the efficient HDF5 data storage format, establish an indexing and metadata management system to ensure the traceability and accessibility of the data, and support the rapid retrieval and processing of large-scale datasets;

[0148] S6.5. Conduct quality inspection on the generated dataset to ensure the integrity and consistency of the hologram samples, eliminate samples with obvious simulation errors or abnormal noise, and ensure the high quality and reliability of the dataset;

[0149] In the step S7, experimental platform construction and data acquisition:

[0150] S7.1. Set up a Michelson digital holographic microscopy system according to the design in Step 1, ensuring that the positions of all optical elements (He-Ne laser, beam expander, aperture, beam splitter, microscope objective, mirror, and CCD camera) are accurately aligned and the optical path is stable without jitter.

[0151] S7.2. Perform optical calibration on the experimental platform, adjust the angles of the beam splitter and the mirror so that stable interference fringes are formed on the CCD by the reference light and the object light, and use a standard resolution target (USAF1951) for alignment.

[0152] S7.3. Conduct data acquisition in a laboratory environment with constant temperature, no vibration, and no dust. Use an anti-vibration table and a temperature control system to reduce the influence of the external environment on the experimental results.

[0153] S7.4. Use the set experimental platform to collect holograms of the USAF1951 resolution target, and record hologram samples at different transparent medium thicknesses (0.5 mm, 1 mm, 2 mm, and 4 mm).

[0154] S7.5. Preprocess the collected holograms, including noise filtering, background subtraction, and image alignment, to ensure that the experimental data and the simulation data are compared under the same standard.

[0155] In the application of step S8, dataset verification:

[0156] S8.1. Compare and analyze the holograms generated by simulation with those collected experimentally, and evaluate the accuracy and reliability of the simulation model through visual comparison and quantitative indicators.

[0157] S8.2. Use the generated hologram dataset with coherent noise to train a convolutional neural network deep learning model for holographic denoising, and set the training parameters and optimization algorithms.

[0158] S8.3. Use an independent experimental dataset to verify and test the trained deep learning model, evaluate the performance of the model in actual denoising tasks, and ensure that it has good generalization ability and robustness.

[0159] S8.4. According to the model training and test results, further expand and optimize the dataset, introduce more diverse noise types and measurement conditions, and improve the coverage of the dataset and the adaptability of the model.

[0160] S8.5. Apply the established dataset and the trained deep learning model to the actual digital holographic microscopy system, improve the quality and measurement accuracy of the holograms, provide technical support and training, and promote the wide use of the dataset in related research and applications.

[0161] In this embodiment, during the superposition process of the object and the reference light field, the optical path difference and phase modulation in the off-axis system are considered, accurately reflecting the three-dimensional structure of the object. Through the mechanism simulation and analysis of parasitic fringes, various types of coherent noise are systematically introduced and controlled, generating hologram samples with real noise characteristics. The dataset generation and management steps use efficient data storage and classification methods to ensure the diversity and high quality of the data, providing rich training samples for the training of deep learning models. In the construction of the experimental platform and data acquisition, the optical elements are precisely aligned and experimental data are acquired in a controlled environment, further verifying the accuracy of the simulation model. By comparing and analyzing the simulation and experimental data and applying a convolutional neural network for holographic denoising, the quality and measurement accuracy of the hologram are significantly improved.

[0162] The present invention,

[0163] The initial settings of the system parameter definitions ensure the stability and repeatability of the experimental conditions, including the precise parameter settings of key optical elements such as a helium-neon laser with a specific wavelength, a microscope objective lens with a numerical aperture of 0.25, and a high-resolution CCD camera. The optical element modeling is parameterized, and the complex amplitude transmittance formula is used to accurately model optical elements such as lenses, beam splitters, and mirrors, and their transfer functions and phase modulation characteristics are verified through simulation software to ensure that the optical performance of the system meets the theoretical expectations. The light source modeling and light field generation use the fundamental mode Gaussian beam model to simulate the laser light source, and the phase and amplitude modulation during the beam propagation process are simulated through the Fresnel diffraction integral and the S-FFT method. The sampling window is dynamically adjusted to maintain the integrity of the light field information, thereby generating an accurate light field distribution.

[0164] During the process of generating and managing the dataset, the object-reference light field superposition generates a hologram reflecting the three-dimensional structure of the object by simulating the interference between the object light and the reference light, and the wrapped phase information is extracted using the Fourier transform method to support subsequent three-dimensional reconstruction. The mechanism simulation and analysis of parasitic fringes are based on the scalar diffraction theory, simulating the parasitic fringe effect caused by the multiple reflections and interference of the transparent medium. By adjusting the thickness of the transparent medium and the optical path difference, holograms with parasitic fringes under different conditions are generated, and the accuracy of the simulation model is verified by comparing with the actual experimental results. The dataset generation and management systematically adjust the key parameters and introduce various types of system noise and environmental noise, and use the HDF5 data storage format to efficiently organize and classify the hologram samples to ensure the diversity and high quality of the dataset, providing rich and reliable training samples for the training of deep learning models.

[0165] The experimental platform is built, and data acquisition is carried out by precisely aligning optical components and collecting data in a controlled environment to obtain experimental holograms at different transparent medium thicknesses. Through preprocessing, the consistency between experimental data and simulation data is ensured. The dataset verification application verifies the accuracy of the simulation model by comparing and analyzing the holograms generated by simulation with those collected experimentally, and uses the generated hologram dataset with coherent noise to train a convolutional neural network for holographic denoising. After the verification and testing of the independent experimental dataset, it is ensured that the deep learning model has good generalization ability and robustness. Finally, the dataset and the trained model are applied to the actual digital holographic microscopy system, significantly improving the quality of holograms and measurement accuracy, and promoting the wide use of related research and applications.

[0166] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements will not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. Method for establishing coherent noise sample data set in microstructure measurement, characterized in that, It includes the following steps: S1. Initial setting of system parameter definition; S2. Parametric modeling of optical elements; S3. Generation of light field in light source modeling; S4. Superposition of object reference light field; S5. Simulation and analysis of parasitic fringe mechanism; S6. Generation and management of data sets; S7. Construction of experimental platform for data acquisition; S8. Verification and application of data sets.

2. The method for establishing a coherent noise sample data set in microstructure measurement according to claim 1, wherein: In the step S1, initial setting of system parameter definition: S1.

1. Select a helium-neon (He-Ne) laser with a wavelength of 632.8 nm, and accurately measure and record the power, coherence length, and divergence angle of the light source; S1.

2. Select a microscope objective with a numerical aperture (NA) of 0.25, set the focal length to 50 mm. The selection of the microscope objective affects the resolution and depth of field of the system; S1.

3. Select a CCD camera with a resolution of 2048×2048 pixels, and the size of a single pixel is 0.801 μm. The quantum efficiency, dynamic range, and noise characteristics of the CCD need to be recorded in detail; S1.

4. According to the propagation distance and the size of the system optical elements, initially set the width of the sampling window to 0.1 m and the sampling interval to 0.05 m; S1.

5. Build a Michelson digital holographic microscopy system in an optical laboratory to ensure the stability of the optical path. The optical table should be equipped with anti-vibration devices and carried out in a clean room environment.

3. The method for establishing a coherent noise sample data set in microstructure measurement according to claim 1, wherein: In the step S2, parametric modeling of optical elements: S2.

1. Use the complex amplitude transmittance formula to model the lens in the system. Set the transmittance of the lens to 1 inside the aperture and 0 outside the aperture, the focal length to 50 mm, the numerical aperture NA = 0.25, and the radius to 25 mm. The complex amplitude transmittance of the lens can be expressed as: Where: τ is the transmittance of the lens, P(x,y) is the pupil function, which is 1 inside the aperture and 0 outside the aperture, f is the focal length of the lens, and the numerical aperture of the lens is NA = r / f, where r is the radius of the lens; S2.

2. Take the USAF1951 standard resolution target as the measurement object, construct its three-dimensional height matrix, convert the image of the standard target into a gray matrix of 1560×1560, and assign appropriate height values (less than half the wavelength of the light source); S2.

3. Accurately model the reflection and transmission characteristics of the beam splitter (BS) and the mirror (M). The beam splitter divides the incident beam into a reference beam and an object beam, and it is necessary to ensure that its reflectivity and transmittance meet the experimental requirements, usually set to a 50:50 splitting ratio; S2.

4. The beam expander BE is used to expand the incident beam, and the aperture limits the beam size. Set the expansion multiple of the beam expander to 10 times, and the aperture diameter of the aperture to 5 mm; S2.

5. Conduct a preliminary verification of the modeled optical elements through the simulation software COMSOL Multiphysics to ensure that the transfer function and phase modulation of each element meet the theoretical expectations.

4. The method for establishing a coherent noise sample data set in microstructure measurement according to claim 1, characterized in that: In the step S3, generation of light field in light source modeling: S3.

1. Use the fundamental mode Gaussian beam model to simulate the laser light source. Set the central beam waist radius to 0.05 m and the Rayleigh distance to π*(0.05)^2 / (632.8e-9)≈3.93 m to ensure that the beam is approximately parallel within the Rayleigh distance. The light field of the fundamental mode Gaussian beam can be expressed as: Where: w0 is the central waist radius, λ is the wavelength of the light source, k = 2π / λ is the wave number (k mentioned in the following text is all wave number), z R = 1 / 2kw0 is the Rayleigh range, is the beam width at z where the light wave propagates, R = z R (z / z R + z R / z) represents the radius of curvature of the equiphase surface, Ψ represents the phase factor, and the propagation of the Gaussian light source within the Rayleigh range can be approximately considered as parallel light. It can be clearly seen from the formula that the larger the beam width, the longer the Rayleigh range, that is, the weaker the beam divergence; S3.

2. Use the Fresnel diffraction integral to simulate the propagation of the light beam from the light source to the microscope objective under the paraxial approximation, and perform fast calculations using the S-FFT method (based on a single Fourier transform). S3.

3. During the propagation of the light beam, consider the phase and amplitude modulation of the light beam by the lens and the object, and adjust the light beam accordingly through the complex amplitude transmittance function to simulate the focusing effect of the lens and the scattering of the object on the light field. The object is regarded as a material with a certain refractive index and a specific thickness at different positions, and its complex amplitude transmittance can be expressed as: t o (x,y) = exp[jkh(x,y)]·τ where: L(x, y) = n o h(x, y) is the optical path brought by the object height matrix h, n o is the refractive index, τ is the transmittance or reflectance. Since this paper studies a reflection holographic microscopy system, τ is the reflectance in this paper; S3.

4. Dynamically adjust the sampling window and sampling interval according to the propagation distance to ensure the integrity of the light field information at each propagation stage, and use the SBLAS method (linear convolution of the band-limited angular spectrum method) to selectively scale the sampling area. S3.

5. Compare the light field generated by the simulation with the theoretical expectation to verify the accuracy of the phase and amplitude during the propagation of the light beam.

5. The method for establishing a coherent noise sample data set in microstructure measurement according to claim 1, characterized in that: In step S4, the superposition of the object reference light field: S4.

1. Simulate the propagation paths of the object light and the reference light respectively. After the object light is reflected by the object, obtain its light field distribution; the reference light is directly reflected by the mirror without passing through the object to maintain coherence. S4.

2. Consider the optical path difference between the object light and the reference light. Especially in an off-axis system, due to the phase difference caused by the angle between the object and the reference, calculate the optical path difference at each point using the focal length of the lens and matrix coordinates to form an optical path difference matrix. Since the simulated system is an off-axis system, the object light and the reference light are not completely symmetric, and there is a certain angle between the object and the reference. Therefore, it is necessary to consider the optical path difference brought by this angle θ to accurately reflect the optical characteristics of the off-axis system. The optical path difference caused by the angle θ can be characterized by the superposition of the phase differences caused by horizontal tilt and vertical tilt: The resulting phase modulation t tilt can be expressed as: where x and y are the coordinates of the observation plane, and θ is the angle between the object light and the reference light. S4.

3. Superimpose the object light field and the reference light field, calculate their interference intensity, generate a hologram, and ensure the accurate transmission of the phase information during the superposition process to reflect the three-dimensional structure of the object. S4.

4. Use the Fourier transform method to calculate the wrapped phase of the hologram to provide the necessary phase information for subsequent three-dimensional reconstruction. S4.

5. Compare with the theoretical hologram to verify whether the generated hologram accurately reflects the three-dimensional structure of the object.

6. The method for establishing a coherent noise sample data set in microstructure measurement according to claim 1, wherein In step S5, the simulation and analysis of the mechanism of parasitic fringes: S5.

1. Based on the scalar diffraction theory, simulate the multiple reflection interference when the light beam passes through the transparent medium. Set the refractive index of the transparent medium to 1.5 and the thickness range to 0.5 mm to 4 mm to study its influence on the parasitic fringes. S5.

2. Calculate the optical path difference between the reflected light on the surface of the transparent medium and the object light. Set the focal length of the lens to 50 mm and the thickness of the transparent medium to 1 mm, 2 mm, and 4 mm, and generate the optical path difference matrices corresponding to different thicknesses. The optical path difference matrix can be expressed as: where f is the focal length of the lens, δ0 is the thickness of the transparent medium, and x and y are the coordinates on the lens observation plane. S5.

3. Introduce the optical path difference matrix into the hologram generation process, superimpose the influence of the parasitic fringes, and use the SBLAS method to accurately simulate the interference effect of the parasitic fringes to generate a hologram containing parasitic fringes. S5.

4. Reconstruct the generated hologram containing parasitic fringes, observe the fringe distribution and density change in the reconstructed image, and analyze the formation law of the parasitic fringes and their relationship with the thickness of the transparent medium. S5.

5. Compare the parasitic fringes generated by the simulation with the actual experimental results to verify the accuracy of the simulation model, and adjust the simulation parameters according to the comparison results to ensure the high consistency between the simulation and the experimental data.

7. The method for establishing a coherent noise sample data set in microstructure measurement according to claim 1, wherein, In step S6, the generation and management of the dataset: S6.

1. During the simulation process, systematically adjust key parameters such as the thickness of the transparent medium, the incident angle, and the object height to generate diverse hologram samples. Each group of samples corresponds to different coherent noise conditions, covering all possible situations in actual measurements. S6.

2. Introduce systematic noise and environmental noise, such as light source intensity fluctuations, CCD noise, and environmental vibrations, into the generated holograms. By adjusting the noise parameters, simulate the common types and intensities of coherent noise in actual measurements. S6.

3. Add detailed label information to each generated hologram sample, including the thickness of the transparent medium, the incident angle, the noise type and intensity, etc. Classify the data according to different noise conditions to facilitate subsequent training of the deep learning model. S6.

4. Organize and store the generated hologram samples using the efficient HDF5 data storage format, and establish an indexing and metadata management system to ensure the traceability and accessibility of the data, and support the rapid retrieval and processing of large-scale datasets. S6.

5. Conduct a quality check on the generated dataset to ensure the integrity and consistency of the hologram samples. Eliminate samples with obvious simulation errors or abnormal noise to ensure the high quality and reliability of the dataset.

8. The method for establishing a coherent noise sample data set in microstructure measurement according to claim 1, wherein In step S7, during the data acquisition of the experimental platform setup: S7.

1. According to the design in step one, set up a Michelson digital holographic microscopy system to ensure the precise alignment of the positions of all optical elements (He-Ne laser, beam expander, aperture, beam splitter, microscopic objective, mirror, and CCD camera), and the optical path is stable without jitter. S7.

2. Conduct optical calibration on the experimental platform, adjust the angles of the beam splitter and the mirror so that stable interference fringes are formed on the CCD by the reference light and the object light, and use a standard resolution target (USAF1951) for alignment. S7.

3. Conduct data acquisition in a laboratory environment with constant temperature, no vibration, and no dust. Use an anti-vibration table and a temperature control system to reduce the influence of the external environment on the experimental results. S7.

4. Use the set experimental platform to acquire holograms of the USAF1951 resolution target, and record hologram samples at different thicknesses of the transparent medium (0.5 mm, 1 mm, 2 mm, and 4 mm). S7.

5. Preprocess the acquired holograms, including noise filtering, background subtraction, and image alignment, to ensure that the experimental data and the simulation data are compared under the same standard.

9. The method for establishing a coherent noise sample data set in microstructure measurement according to claim 1, characterized in that In step S8, during the dataset verification and application: S8.

1. Compare and analyze the holograms generated by simulation with those acquired experimentally, and evaluate the accuracy and reliability of the simulation model through visual comparison and quantitative indicators. S8.

2. Use the generated hologram dataset with coherent noise to train a convolutional neural network deep learning model for holographic denoising, and set the training parameters and optimization algorithms. S8.

3. Use an independent experimental dataset to verify and test the trained deep learning model, and evaluate the performance of the model in actual denoising tasks to ensure its good generalization ability and robustness. S8.

4. Further expand and optimize the dataset based on the model training and test results, introduce more diverse noise types and measurement conditions, and improve the coverage of the dataset and the adaptability of the model; S8.

5. Apply the established dataset and the trained deep learning model to the actual digital holographic microscopy system, improve the quality of holograms and measurement accuracy, provide technical support and training, and promote the wide use of the dataset in related research and applications.

Images