Three-wavelength and four-angle polarization sensitive holographic imaging system

Abstract

According to the three-wavelength and four-angle polarization sensitive holographic imaging system provided by the invention, the imaging system has wavelength and polarization resolution capability by using a mode of increasing information channels and synchronously acquiring; distortion phase and scattering noise, caused by medium refractive index disturbance and smoke multiple scattering in a severe imaging environment, of a target light field are analyzed, and the complex and severe environment imaging problem similar to an energetic particle combustion field can be well solved. Furthermore, four-angle polarization is introduced to increase an independent channel, so that the anti-interference capability of holographic imaging is improved, and the information storage capacity of the system is improved by four times on the premise that the physical size does not need to be increased. The problem of extinction and speckle interference caused by high particle concentration number is solved by means of the thought of synchronous interpretation of multi-dimensional light field information, and polarization characteristic parameters capable of distinguishing particles from background areas are extracted from polarization channels by means of the advantage of more stable transmission of polarization characteristics of light so as to obtain three-dimensional position information of the particles.

Description

Technical Field

[0001] This invention relates to the field of optical imaging technology, and in particular to a three-wavelength and four-angle polarization-sensitive holographic imaging system. Background Technology

[0002] Energetic particles have high calorific value and energy density, and they have been widely used in solid propellants and high-energy explosives, which can significantly improve propulsion performance and damage effects. New aerospace propulsion systems based on high-energy powder fuels, such as ramjet engines and rotary detonation engines, are also developing rapidly. Conducting basic research on the combustion of energetic particles and obtaining accurate and reliable particle combustion characteristics is of great scientific significance for combustion mechanism research and technology optimization.

[0003] Existing particle field three-dimensional imaging measurement methods include optical field imaging, tomographic imaging, and digital holographic imaging. Among them, digital holography is gradually being promoted and applied in the field of particle combustion diagnosis because it can achieve quantitative three-dimensional imaging and phase imaging with a single-frame, single-wavelength, and single-view optical path system, and can quantitatively characterize the particle characteristics of the combustion field. Although digital holography has flexible digital refocusing capabilities and can significantly extend the detection depth, it also has limitations in the imaging of energetic particle combustion fields: non-uniform media such as flue gas in the optical path will cause irregular optical phase distortion and random multiple scattering, interfering with the accurate recording of holographic fringes, and thus leading to image degradation and distortion during diffraction inversion reconstruction; as the particle number concentration increases, the aliasing of holographic fringes and multiple scattering between particles will gradually become significant, and even axial spatial occlusion will occur, which will seriously reduce the particle identification accuracy and measurement precision, and the system's measurable concentration threshold will be insufficient.

[0004] In imaging applications involving energetic particles and similar harsh environments, dynamic perturbations of the refractive index and multiple scattering of smoke often lead to severe image distortion. Deciphering the holographic light field transmission mechanism of particles in these heterogeneous, strongly scattering, and high-concentration environments is crucial for achieving high-fidelity numerical reconstruction. Traditional single-wavelength imaging modes, lacking sufficient physical information, are ill-suited for interpretation and cannot meet the needs of quantitative measurement research on high-concentration particle fields in heterogeneous, strongly scattering combustion environments.

[0005] In view of the above, this application is hereby submitted. Summary of the Invention

[0006] This invention discloses a three-wavelength and four-angle polarization-sensitive holographic imaging system, which aims to solve the problem that single-wavelength or single-polarization holographic imaging is prone to phase distortion, speckle noise and reconstruction distortion in heterogeneous, strongly scattering, and high-concentration particle environments such as combustion fields, making it difficult to accurately measure the three-dimensional position, particle size and number concentration of particles.

[0007] This invention provides a three-wavelength and four-angle polarization-sensitive holographic imaging system, comprising: a holographic illumination source module, a lens group, an image acquisition module, and a data processing module; The holographic illumination source module is used to generate laser beams of three independent wavelengths—red, green, and blue—and combine them into a coaxial composite beam via a beam splitter. The lens group is used to spatially filter and expand the composite beam, and after the polarization state is adjusted by a linear polarizer and a quarter-wave plate, it forms a composite circularly polarized light to illuminate the target particle field. The target particle field under test generates an interference light field containing particle scattering information under the illumination of the composite circularly polarized light. The image acquisition module is located at the end of the optical path and is used to record the interference light field and map the original multi-channel data into holographic stripe information in digital form by channel, and transmit it to the data processing module in real time. The data processing module is configured to execute a computer program stored therein to perform the following steps: The original holographic data of three wavelength channels and the intensity information of four polarization channels are acquired. Diffraction inversion is performed on the holographic data of each wavelength channel to obtain the complex amplitude distribution at each wavelength, and the polarization parameters are calculated based on the intensity information of the four polarization channels. Based on the complex amplitude distribution at each wavelength and the polarization parameters, the phase distortion of the target particle field under test is compensated and speckle noise is suppressed, and the quantization parameters of the target particle field under test are generated based on the enhanced image.

[0008] Preferably, after acquiring the original holographic data of three wavelength channels and the intensity information of four polarization channels, diffraction inversion is performed on the holographic data of each wavelength channel. The diffraction inversion is based on the Fresnel-Kirchhoff diffraction theory, and the holographic fringes formed by the interference of the object light field and the reference light field are represented as follows:

[0009] in, Represents the convolution operator. For the inverse Fourier transform of the diffraction nucleus G, For object light field, For the reference light field, the diffraction nucleus G is represented as: ),in,( () represents spatial frequency. The wavelength of coherent light. The axial propagation distance, The imaginary unit; Camera on the imaging surface The recorded holographic light intensity is: The target light field is reproduced using digital refocusing methods:

[0010] Where * denotes conjugation. For the inverse Fourier transform of the conjugate term of the diffraction nucleus G.

[0011] Preferably, the compensation for the phase distortion of the target particle field based on the complex amplitude distribution at each wavelength and the polarization parameters specifically involves: The differences in phase distortion across different wavelengths are analyzed, and compensation is performed based on the diffraction distance and polarization state. This compensation is achieved through frequency-domain constrained iterative inversion, correcting the distorted phase during the iteration process. Specifically: Based on synchronously acquired four-angle polarization holograms ( Transform it into a complex amplitude hologram with phase information: ,in, For reference optical amplitude, The intensity of the hologram recorded when the linear polarization direction is 0° is the light intensity. The light intensity of the hologram recorded when the linear polarization direction is 45° is shown. The light intensity of the hologram recorded when the linear polarization direction is 90° is shown. The intensity of the hologram is recorded when the linear polarization direction is 135°. Then, numerical reconstruction is performed based on the target light field to interpret the phase and amplitude of the target light field. Iterative inversion is performed through frequency domain constraints to correct the distorted phase.

[0012] Preferably, speckle noise is suppressed after phase distortion compensation, wherein speckle noise is suppressed by wavelength synthesis based on information from three wavelength channels, and the expression is as follows:

[0013] in, The wavelength of red light The wavelength of green light It is the wavelength of blue light. The equivalent wavelength for the combination of red and green light. The equivalent wavelength for the combination of red and blue light. The equivalent wavelengths of green and blue light are used to synthesize the image. By weighted fusion of the three single-wavelength reconstruction results and the three synthesized wavelength reconstruction results, a multi-wavelength fusion reconstruction image with enhanced signal-to-noise ratio is obtained.

[0014] Preferably, it further includes: performing multi-sample noise reduction on the scattered photons' random characteristics of the multi-wavelength fused reconstructed image, specifically: The multi-channel information is randomly rearranged and combined in the spatial or frequency domain to generate multiple independent samples. Random scattering noise is suppressed by averaging the reconstruction results of these multiple samples. The expression for the multi-channel information rearrangement is as follows:

[0015] in, For all wavelength channels, Summing over all polarization angles, A random coefficient matrix of wavelength dimension. The random coefficient matrix is ​​the polarization dimension. This is the original hologram corresponding to the wavelength and polarization channel. These are synthetic hologram samples generated by random rearrangement.

[0016] Preferably, the quantization parameters for generating the target particle field based on the enhanced image are specifically as follows: By using the recorded color polarization holograms of dense particle fields, multiple image features of particles are extracted from multi-channel information. The evolution of particle image similarity (F1), gradient sparsity (F2), light intensity uniformity (F3), edge sharpness (F4), and background consistency (F5) along the reconstruction depth direction is mainly interpreted. These feature expressions are:

[0017] in, Let be the set of pixels of a certain particle in the first frame of the image. Let N be the set of pixels to be matched in the second frame image, and n and m be the pixel indices of sets N and M, respectively. Let be the joint probability distribution of the two pixel sets. Marginal probability distribution of set N Let M be the marginal probability distribution of the set, and i and j be the pixel coordinate indices within the particle image. Let (i,j) be the image gradient magnitude. The light intensity value at position (i,j) inside the particle. Let (i,j) be the gradient magnitude at position (i,j) on the particle edge region. Let be the light intensity value at position (i,j) in the background region surrounding the particle; By fusing effective features through weighted allocation, an integrated identification criterion for single and overlapping particles is established. Adaptive curve fitting and extreme value search strategies are used to achieve accurate particle edge segmentation and axial positioning. Combined with clear particle morphology, spatiotemporal constraints, and polarization response, the three-dimensional displacement parameters of all particles between two frames are obtained.

[0018] Preferably, the image acquisition module is a color polarization camera, and the photosensitive chip of the color polarization camera integrates an RGB Bayer array and a four-angle micro polarizer array.

[0019] This invention provides a three-wavelength and four-angle polarization-sensitive holographic imaging system. Compared to existing single-wavelength imaging technologies, by increasing information channels and synchronously acquiring data, the imaging system possesses wavelength and polarization resolution capabilities. It resolves the distortion phase and scattering noise of the target light field caused by medium refractive index disturbances and smoke multiple scattering in harsh imaging environments, effectively solving complex and harsh environment imaging problems such as energetic particle combustion fields. Furthermore, the introduction of four-angle polarization adds independent channels, not only improving the anti-interference capability of holographic imaging but also significantly increasing information storage efficiency, allowing the system to increase information storage capacity by four times without increasing physical size. Similarly, by employing the idea of ​​synchronous interpretation of multi-dimensional light field information, it solves the extinction and speckle interference problems caused by high particle concentration. Utilizing the more stable transmission advantage of light polarization characteristics, polarization feature parameters that can distinguish particles from background areas are extracted from the polarization channels to obtain the three-dimensional position information of the particles. Attached Figure Description

[0020] Figure 1 This invention provides a three-wavelength and four-angle polarization-sensitive holographic imaging system. Figure 2 This is a schematic diagram of the execution flow of the data processing module provided in an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In order to better understand the technical solutions of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] This invention discloses a three-wavelength and four-angle polarization-sensitive holographic imaging system, which aims to solve the problem that single-wavelength or single-polarization holographic imaging is prone to phase distortion, speckle noise and reconstruction distortion in heterogeneous, strongly scattering, and high-concentration particle environments such as combustion fields, making it difficult to accurately measure the three-dimensional position, particle size and number concentration of particles.

[0023] Please see Figure 1 This invention provides a three-wavelength and four-angle polarization-sensitive holographic imaging system, including: a holographic illumination source module 1, a lens group, an image acquisition module 6, and a data processing module 8; The holographic illumination source module 1 is used to generate laser beams of three independent wavelengths: red, green, and blue, which are then combined into a coaxial composite beam by the beam splitter 2. The lens group is used to spatially filter and expand the composite beam, and after the polarization state is adjusted by the linear polarizer 4 and the quarter-wave plate 5, it forms a composite circularly polarized light to illuminate the target particle field 6. The target particle field 6 under test generates an interference light field containing particle scattering information under the illumination of the composite circularly polarized light. The image acquisition module 6 is located at the end of the optical path and is used to record the interference light field and map the original multi-channel data into holographic stripe information in digital form by channel, and transmit it to the data processing module 8 in real time. In this embodiment, the holographic illumination source module 1 can employ three independently controllable semiconductor lasers, each outputting coherent laser beams with a red wavelength of 633nm, a green wavelength of 530nm, and a blue wavelength of 445nm. The output power of the three lasers can be independently adjusted according to actual needs, with a typical power range of 50mW to 500mW. A beam splitter 2 is installed in front of the output port of each laser to combine the three laser beams of different wavelengths into a coaxial composite laser beam. The combined composite beam maintains the coherence and monochromaticity of each wavelength of laser light.

[0024] After entering the lens group, the composite beam first undergoes beam shaping through a spatial filter and a beam-expanding lens group. The spatial filter employs a pinhole structure with a pinhole diameter typically ranging from 10 to 50 micrometers. A converging lens is positioned before the pinhole to focus the laser beam onto the pinhole, filtering out high-frequency noise and stray light. A beam-expanding lens is positioned after the pinhole to expand the clean beam passing through the pinhole to the desired diameter. The beam-expanding lens group uses a dual-lens or multi-lens combination structure, achieving a beam expansion ratio of 10 to 50 times. The expanded beam diameter can cover the entire particle field region under test, with an expansion diameter ranging from 50 mm to 200 mm. The expanded composite beam undergoes homogenization processing, resulting in a Gaussian or flat-top intensity distribution to ensure uniform illumination.

[0025] The linear polarizer 4 in the lens group is positioned after the beam expander lens. This linear polarizer 4 is a Glan prism or a polarizing beam splitter prism with an extinction ratio greater than 1000:1. It can convert the expanded natural light or partially polarized light into linearly polarized light. The polarization direction is adjusted by rotating the installation angle of the linear polarizer 4, usually set to a horizontal direction or a specific direction aligned with the experimental coordinate system. The linearly polarized light then enters a quarter-wave plate 5, whose fast axis is placed at a 45-degree angle to the vibration direction of the linearly polarized light. When the linearly polarized light passes through the quarter-wave plate 5, its horizontal and vertical components generate a quarter-wavelength phase difference, thereby converting the linearly polarized light into circularly polarized light. The quarter-wave plate 5 is a multi-stage waveplate or an achromatic waveplate, capable of simultaneously achieving a quarter-wavelength phase delay at three wavelengths: 633nm, 530nm, and 445nm, ensuring that the beams of all three wavelengths are converted into circularly polarized light. The converted composite circularly polarized light has rotational symmetry and is unaffected by the rotation angle of the target under test, enabling more stable recording of particle scattering information.

[0026] The composite circularly polarized light illuminates the target particle field 6 in the form of parallel light or weakly divergent light. The particle field is located in the measurement region at the center of the optical path, which can be realized through a transparent window or an open space. When the composite circularly polarized light penetrates the particle field, the particles scatter, diffract, and absorb the light beam, forming a modulated light field containing information about the particle position, size, morphology, and concentration. Due to the use of circularly polarized light for illumination, the scattered light retains polarization state information, and the polarization preservation property of circularly polarized light is better than that of linearly polarized light, maintaining certain polarization characteristics even after multiple scatterings. The scattered modulated light field interferes with the unscattered reference light during propagation, forming an interference light field containing amplitude and phase information. This interference light field propagates to the image acquisition module 6 in a coaxial holographic manner.

[0027] The image acquisition module 6 is located at the end of the optical path, and the distance between it and the particle field to be measured is the working distance, which typically ranges from 50mm to 500mm and is adjusted according to the size of the particle field and the imaging resolution requirements. The image acquisition module 6 uses a color polarization camera, which is an integrated design that integrates color imaging and polarization imaging functions into a single sensor. The core of the color polarization camera is a CMOS or CCD photosensitive chip, with an effective pixel count typically ranging from 2 million to 20 million, a single pixel size of 3 to 5 micrometers, a frame rate of 24fps to 60fps, and a shutter speed of 30 microseconds to 1 millisecond, which can adapt to the imaging needs of particle fields with different movement speeds. An RGB Bayer array filter layer and a four-angle micro-polarizer array are integrated on the front surface of the photosensitive chip. The two arrays are arranged in a superpixel structure, with each 2×2 pixel group forming a superpixel unit. The RGB Bayer array is distributed in the traditional RGGB pattern, and a micro-polarizer is superimposed in front of each color pixel. The polarization directions of the four micro-polarizers are 0 degrees, 45 degrees, 90 degrees, and 135 degrees, respectively. This integrated structure allows for the simultaneous recording of independent information from three wavelength channels (red, green, and blue) and four polarization angle channels (0°, 45°, 90°, and 135°) in a single exposure, totaling 12 independent channels. The photosensitive chip converts the received interference light field into an electrical signal, which is then converted into digital image data by an analog-to-digital converter. The data for each channel is stored in an independent matrix format, recording the intensity distribution of the holographic fringes for each channel.

[0028] The color polarization camera connects to the data processing module 8 via a high-speed data interface such as USB 3.0, GigE, or Camera Link to transmit raw multi-channel data in real time. The data received by the data processing module 8 contains holographic fringe information from 12 channels, with red, green, and blue wavelengths each corresponding to holograms at four polarization angles, totaling 12 holograms. Based on the arrangement rules of the RGB Bayer array and polarization array, the data processing module 8 demultiplexes the raw pixel data into 12 independent holographic matrices, each matrix corresponding to interference fringe information at a specific wavelength and polarization angle. Due to the synchronous acquisition method, the data from the 12 channels are completely consistent in time and precisely aligned in spatial position.

[0029] Please see Figure 2 The data processing module is configured to execute a computer program stored therein to perform the following steps: S101: Obtain the original holographic data of three wavelength channels and the intensity information of four polarization channels. Perform diffraction inversion on the holographic data of each wavelength channel to obtain the complex amplitude distribution at each wavelength, and calculate the polarization parameters based on the intensity information of the four polarization channels. In this embodiment, starting with the acquisition of raw multi-channel data from a color polarization camera, the data from 12 independent channels is first demultiplexed and preprocessed. The color polarization camera simultaneously records information from three wavelength channels in a single exposure: red (633nm), green (530nm), and blue (445nm). Each wavelength channel contains intensity data for four polarization directions: 0°, 45°, 90°, and 135°. After receiving the raw pixel data, the data processing module separates the interwoven pixel values ​​into 12 independent hologram matrices according to the spatial arrangement rules of the RGB Bayer array and the four-angle micro-polarizer array on the photosensitive chip. Four holograms are extracted from the red channel, four from the green channel, and four from the blue channel. Due to the superpixel structure of the Bayer array and polarization array, the effective resolution of each independent channel is one-quarter of the original sensor resolution. Therefore, during the separation process, interpolation algorithms such as bilinear interpolation or bicubic interpolation are needed to restore the images of each channel to their original resolution, ensuring consistent spatial sampling rates in subsequent processing.

[0030] After acquiring the raw holographic data for three wavelength channels, the data processing module performs diffraction inversion on each wavelength channel to reconstruct the target optical field. The diffraction inversion is based on the Fresnel-Kirchhoff diffraction theory, which describes the propagation of light waves in free space. During holographic recording, the object light field carries the amplitude and phase information of the field of the particle being measured, and its mathematical expression is: ,in For the amplitude distribution of the object light, Let i represent the phase distribution of the object light, where i is the imaginary unit. The reference light field is a known plane wave or spherical wave, expressed as: ,in For reference light amplitude distribution, The reference light phase distribution is shown. The object light field and the reference light field interfere during propagation, and the resulting complex amplitude distribution of the holographic fringes is represented as follows: ,in Represents the convolution operator. The inverse Fourier transform of the diffraction kernel G is given by: [Formula omitted for brevity] In the formula, z is the axial propagation distance from the object plane to the recording plane, and λ is the wavelength of the coherent light. and These are the spatial frequency components in the horizontal and vertical directions. The hologram recorded by the camera at the imaging plane z can only acquire light intensity information and loses phase information; the recorded light intensity distribution is the square of the modulus of the complex amplitude of the holographic fringes.

[0031] To reconstruct the target light field from the recorded light intensity information, the data processing module employs a digital refocusing method. This method utilizes a known reference light field to perform backward diffraction propagation of the recorded hologram. The formula for reconstructing the target light field is as follows: ,in Represents the complex conjugate of the reference light field. This is the inverse Fourier transform of the conjugate term of the diffraction nucleus G. In actual calculations, the recorded holographic light intensity is first... Conjugate with reference light Multiply the results, then transform them to the frequency domain using a Fast Fourier Transform (FFT), multiply by the conjugate term of the diffraction kernel G, and finally transform them back to the spatial domain using an Inverse Fast Fourier Transform (IFFT) to obtain the reconstructed optical fields at different propagation distances z. The reconstructed light field has a complex distribution, and its magnitude is | | represents the amplitude distribution of the reconstructed image, and its argument arg( The phase distribution of the reconstructed image is represented by ). By changing the value of the propagation distance z, the light field distribution of the target particle field at different depths can be reconstructed, achieving digital focusing. The above diffraction inversion process is performed on the red, green, and blue light wavelength channels respectively. Since the wavelength λ in the diffraction nuclei G of the three wavelengths is different, the diffraction effect at the same propagation distance z is different. Therefore, the three wavelength channels need to be calculated independently, and finally three sets of complex amplitude distributions are obtained.

[0032] While obtaining the complex amplitude distribution at each wavelength, the data processing module calculates polarization parameters based on the intensity information of the four-angle polarization channels. For each wavelength channel, the acquired four polarization angle holograms h0, h45, h90, and h135 contain complete information on the polarization state of the optical field. According to Stokes parameter theory, the four components S0, S1, S2, and S3 of the Stokes vector can be calculated from the four polarization intensities. The calculation formulas are S0 = h0 + h90, S1 = h0 - h90, S2 = h45 - h135, and S3 needs to be obtained through left-hand and right-hand circular polarization measurements, but can be approximately ignored in this system. Based on the Stokes parameters, the degree of polarization DOP and the polarization angle AOP can be further calculated. The degree of polarization DOP = √(S1 + h90) / AOP. 2 +S2 2 +S3 2 S0 represents the degree of polarization of the light field, ranging from 0 to 1. Completely unpolarized light has a polarization degree of 0, while completely polarized light has a polarization degree of 1. The polarization angle AOP = (1 / 2)arctan(S2 / S1) represents the angle between the vibration direction of linearly polarized light and the reference direction. These polarization parameters form a spatial distribution map, reflecting the modulation effect of the particle field on the polarization state of light, providing additional physical information for subsequent phase distortion compensation and feature extraction.

[0033] Based on the complex amplitude distribution and calculated polarization parameters at each wavelength, the data processing module compensates for the phase distortion of the target particle field caused by medium refractive index perturbation and multiple scattering from smoke. Phase distortion manifests as the phase distribution arg(E_h(x,y,z)) in the reconstructed image deviating from the true value, leading to image blurring or artifacts. Since different wavelengths of light produce different phase delays under the same medium perturbation, the phase distortion of the reconstructed images at the three wavelengths is wavelength-dependent. The data processing module first analyzes the differences in phase distortion among the three wavelengths of red, green, and blue light. By calculating the phase differences ΔφRG=φR-φG, ΔφRB=φR-φB, and ΔφGB=φG-φB, it identifies systematic phase error patterns. These phase differences are affected by the choice of diffraction distance z and the polarization state. By establishing a correlation model between the phase differences and these parameters, the true phase distribution can be estimated. The specific compensation strategy is implemented through frequency-domain constrained iterative inversion. In each iteration, the phase value is corrected according to physical constraints, gradually approximating the true phase.

[0034] The core of frequency-domain constrained iterative inversion is utilizing the additional phase information provided by four-angle polarization holograms. Traditional single-intensity holograms hc(x,y) can only estimate the phase using iterative algorithms such as the Gerchberg-Saxton algorithm, but this method has slow convergence and is prone to getting trapped in local extrema. This invention directly recovers the phase information through four-angle polarization measurements, avoiding the uncertainty of phase estimation. Specifically, the synchronously acquired four-angle polarization holograms h0, h45, h90, and h135 are transformed into complex amplitude holograms with phase information. The transformation formula is as follows: ,in The reference optical amplitude is usually a known constant or determined through calibration experiments. The physical meaning of this formula is to extract the real part (h0-h90) and imaginary part (h45-h135) of the optical field from four polarization measurements to reconstruct the complete complex amplitude. h0 is the intensity at 0 degrees of polarization, h90 is the intensity at 90 degrees of polarization, and the difference between them (h0-h90) is proportional to the real component of the optical field in the horizontal direction. h45 is the intensity at 45 degrees of polarization, h135 is the intensity at 135 degrees of polarization, and the difference between them (h45-h135) is proportional to the imaginary part of the optical field in the diagonal direction. By combining the real and imaginary parts into a complex number and dividing by the normalization coefficient 4a_R, the following is obtained: This is a complex amplitude hologram containing complete amplitude and phase information. Compared to simple light intensity recording hc(x,y), this complex amplitude hologram preserves the phase information of the interference fringes, making the reconstruction process more accurate.

[0035] Obtain complex amplitude hologram Then, the data processing module performs numerical reconstruction based on the target light field reconstruction formula. At this point, iterative phase estimation is no longer required; instead, the data is directly processed. Substituting into the forward propagation formula

[0036] Calculations are performed because The phase is already included, and the reconstructed result is... This method enables more accurate interpretation of the phase φh(x,y,z) and amplitude a_h(x,y,z) of the target light field. To further correct residual phase distortion, an iterative inversion method with frequency domain constraints is employed. This method imposes physical constraints in the frequency domain, such as a support domain constraint requiring the target object to exist only within a finite spatial region, and a frequency domain constraint requiring the light field's spectrum to conform to physical realizability. After multiple iterations, the phase of the light field gradually converges to the true value, and the phase distortion is effectively corrected. The above complex amplitude reconstruction and iterative correction process is performed on the three wavelength channels respectively, ultimately obtaining three sets of high-quality complex amplitude distributions after phase correction.

[0037] After phase distortion compensation, the data processing module further suppresses speckle noise in the reconstructed image. Speckle noise is a random interference pattern generated by multiple scattering of a particle field under coherent laser illumination. It manifests as high-frequency granular noise in the image, severely affecting the sharpness of particle edges and measurement accuracy. Since speckle noise of different wavelengths is spatially uncorrelated—that is, the speckle pattern generated by red light is random and independent of the speckle patterns generated by green and blue light in terms of position and intensity—this characteristic can be utilized to significantly reduce speckle noise through multi-wavelength fusion. This invention employs a wavelength synthesis method, the basic principle of which is to construct a virtual light source with an equivalent wavelength much larger than a single wavelength, thereby reducing phase sensitivity and speckle contrast. The expressions for wavelength synthesis are λ_RG=(λ_R×λ_G) / |λ_R-λ_G|, λ_RB=(λ_R×λ_B) / |λ_R-λ_B|, λ_GB=(λ_G×λ_B) / |λ_G-λ_B|, where λ_R=633nm is the red light wavelength, λ_G=530nm is the green light wavelength, λ_B=445nm is the blue light wavelength, λ_RG is the equivalent wavelength of the synthesized red and green light, λ_RB is the equivalent wavelength of the synthesized red and blue light, and λ_GB is the equivalent wavelength of the synthesized green and blue light. Calculations show that λ_RG is approximately 3263nm, λ_RB is approximately 1501nm, and λ_GB is approximately 2782nm. The synthesized wavelength is several times or even tens of times larger than the single wavelength, resulting in a corresponding decrease in phase sensitivity.

[0038] The specific implementation of wavelength synthesis involves performing a phase difference operation on the reconstructed images of two different wavelengths. The phase difference ΔφRG = φR - φG is equivalent to the phase under illumination at wavelength λ_RG, and the corresponding reconstructed image is a virtual reconstructed image based on the phase difference. The data processing module calculates the reconstructed images of the three wavelength combinations (red-green, red-blue, and green-blue) respectively, and adds them to the original three single-wavelength reconstructed images (red, green, and blue), resulting in a total of six sets of reconstruction results. These six sets of images are weighted and fused at the same diffraction distance z. The weight coefficients are dynamically adjusted according to the signal-to-noise ratio (SNR) or contrast of each wavelength image, assigning larger weights to images with high SNR and smaller weights to images with low SNR. Through weighted fusion, random speckle noise in a single-wavelength image is suppressed after averaging multiple images, while the structural information of the target particles is enhanced due to the consistency across wavelengths, ultimately resulting in a multi-wavelength fused reconstructed image with enhanced SNR.

[0039] To further improve image quality, the data processing module performs multi-sample noise reduction on the multi-wavelength fused reconstructed image based on the random characteristics of scattered photons. This method utilizes the difference between the randomness of scattering noise and the determinism of the target signal, suppressing random scattering noise by constructing multiple independent noise samples and averaging them. Specifically, it involves randomly rearranging and combining the acquired multi-wavelength and multi-polarization channel information in the spatial or frequency domain to generate multiple synthetic hologram samples. The mathematical expression for random rearrangement is as follows: ,in, For all wavelength channels, Sum over all polarization angles, A random coefficient matrix of wavelength dimension. The random coefficient matrix is ​​the polarization dimension. This is the original hologram corresponding to the wavelength and polarization channel. These are synthetic hologram samples generated by random rearrangement.

[0040] random coefficient matrix and The elements are random numbers between 0 and 1 or between -1 and 1, and are re-randomized each time a sample is generated to ensure that the noise distribution of different samples is independent. Because the scattering noise has weak correlation between different wavelengths and polarization channels, the synthesized hologram after random rearrangement... The noise distribution in the target particle changes, while the ballistic photon signal of the target particle maintains coherence and polarization preservation and is highly correlated between channels. Therefore, the target signal can still maintain superposition enhancement after random combination.

[0041] The data processing module generates N independent synthetic hologram samples, where N typically ranges from 10 to 100, with each sample corresponding to a set of random coefficient matrices. Diffraction inversion and reconstruction are performed on each sample to obtain N reconstructed images. Since the noise distribution of each sample is random and independent, while the target signal remains consistent, by arithmetically averaging the N reconstruction results, the power of random scattering noise decays at a rate of 1 / N, while the power of the target signal remains unchanged, resulting in a signal-to-noise ratio improvement of √N times. The averaged image exhibits effective suppression of random noise, resulting in clearer particle edges and a flatter background.

[0042] S102, based on the complex amplitude distribution at each wavelength and the polarization parameters, compensate for the phase distortion of the target particle field under test and suppress speckle noise, and generate the quantization parameters of the target particle field under test based on the enhanced image.

[0043] The process of generating quantization parameters begins with the extraction of multiple image features of the particles. Since the dense particle field hologram recorded by the color polarization camera contains three wavelength channels (red, green, and blue) and four polarization channels (0°, 45°, 90°, and 135°), totaling 12 independent channels, the data processing module extracts five types of image features from this multi-channel information: image similarity F1, gradient sparsity F2, light intensity uniformity F3, edge sharpness F4, and background consistency F5.

[0044] Image similarity feature F1 is used to quantify the matching degree of particles between adjacent frames, and is calculated based on mutual information theory. For the pixel set N of a particle in the first frame and the pixel set M of the particle to be matched in the second frame, their joint probability distribution p(n,m) and edge probability distributions p(n) and p(m) are calculated. The feature expression is as follows: , where n and m are the pixel indices of sets N and M, respectively. A higher F1 value indicates greater similarity between two particle images, and is used for inter-frame particle tracking.

[0045] The gradient sparsity feature F2 is used to determine whether a particle is in optimal focus. The gradient magnitude G(i,j) is calculated for each pixel position (i,j) in the particle image, and the feature expression is: F2 is the ratio of the L1 norm to the L2 norm of the gradient. When the particle is in the focal position, the edge is sharp and the gradient is concentrated, and F2 reaches its maximum value. The axial coordinate of the particle is determined by searching for the maximum value of F2 at different reconstruction depths.

[0046] The light intensity uniformity feature F3 is used to distinguish between the interior and edge regions of particles. The mean [I(i,j)] and standard deviation [I(i,j)] of the light intensity I(i,j) in the interior region of the particle are calculated, and the characteristic expression is: A higher F3 value indicates uniform light intensity within the particle, while a lower F3 value indicates significant light intensity variation at the edges or in out-of-focus areas. This is used for particle segmentation.

[0047] The edge sharpness feature F4 is used to evaluate the sharpness of edges. The gradient over the grain edge region... The characteristic expression for calculating the mean and standard deviation is: When focusing, the edge gradient is large and consistent, and F4 increases, serving as an auxiliary criterion for axial positioning.

[0048] Background consistency feature F5 is used to distinguish grains from the background. (The text then abruptly shifts to discussing the light intensity in the background region.) The characteristic expression for calculating the mean and standard deviation is: When the actual background light intensity is stable, F5 is close to 1; when there are large fluctuations in grainy or speckle areas, F5 is decreased to eliminate false background interference.

[0049] After extracting five types of features, the data processing module establishes a comprehensive evaluation function F_total = w_1·F1 + w_2·F2 + w_3·F3 + w_4·F4 + w_5·F5 through weighted fusion. The weight coefficients w_1 to w_5 are determined based on experimental calibration or machine learning methods. The spatial distribution and extreme value locations of F_total are used to achieve integrated identification of single and overlapping particles. For overlapping particles, the center of each particle is distinguished by searching for local extreme points.

[0050] After determining the candidate particle positions, edge segmentation is achieved using adaptive curve fitting. Edge pixels are located using gradient operators, and smooth boundaries are obtained by fitting elliptical or B-spline curves. Geometric parameters such as particle area, equivalent diameter, and aspect ratio are calculated. An extreme value search strategy is used to determine the axial position. Near the coarse positioning peak, a Gaussian function is used to fit the F2(z) curve; the peak position is the particle's axial coordinate z_particle, achieving sub-pixel-level positioning accuracy.

[0051] After obtaining the 3D positions of all particles in a single frame, tracking matching is established between adjacent frames by combining particle morphology features, spatiotemporal constraints, and polarization response. For particle i in the first frame and candidate particle j in the second frame, a comprehensive score is constructed by comprehensively calculating spatial distance, morphological similarity, polarization similarity, and image mutual information F1. The Hungarian algorithm is used to solve for the optimal match that maximizes the total score. After matching, the 3D displacement parameters are calculated and the 3D velocity vector is obtained by combining the time interval. The data processing module outputs a complete quantized parameter list containing 3D coordinates, displacement vector, velocity vector, size parameters, and polarization parameters, realizing end-to-end measurement from the original hologram to the 3D motion parameters of the particles.

[0052] This invention provides a three-wavelength and four-angle polarization-sensitive holographic imaging system. Compared to existing single-wavelength imaging technologies, by increasing information channels and synchronously acquiring data, the imaging system possesses wavelength and polarization resolution capabilities. It resolves the distortion phase and scattering noise of the target light field caused by medium refractive index disturbances and smoke multiple scattering in harsh imaging environments, effectively solving complex and harsh environment imaging problems such as energetic particle combustion fields. Furthermore, the introduction of four-angle polarization adds independent channels, not only improving the anti-interference capability of holographic imaging but also significantly increasing information storage efficiency, allowing the system to increase information storage capacity by four times without increasing physical size. Similarly, by employing the idea of ​​synchronous interpretation of multi-dimensional light field information, it solves the extinction and speckle interference problems caused by high particle concentration. Utilizing the more stable transmission advantage of light polarization characteristics, polarization feature parameters that can distinguish particles from background areas are extracted from the polarization channels to obtain the three-dimensional position information of the particles.

[0053] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A three-wavelength and four-angle polarization-sensitive holographic imaging system, characterized in that, include: Holographic illumination source module, lens group, image acquisition module, and data processing module; The holographic illumination source module is used to generate laser beams of three independent wavelengths—red, green, and blue—and combine them into a coaxial composite beam via a beam splitter. The lens group is used to spatially filter and expand the composite beam, and after the polarization state is adjusted by a linear polarizer and a quarter-wave plate, it forms a composite circularly polarized light to illuminate the target particle field. The target particle field under test generates an interference light field containing particle scattering information under the illumination of the composite circularly polarized light. The image acquisition module is located at the end of the optical path and is used to record the interference light field and map the original multi-channel data into holographic stripe information in digital form by channel, and transmit it to the data processing module in real time. The data processing module is configured to execute a computer program stored therein to perform the following steps: The original holographic data of three wavelength channels and the intensity information of four polarization channels are acquired. Diffraction inversion is performed on the holographic data of each wavelength channel to obtain the complex amplitude distribution at each wavelength, and the polarization parameters are calculated based on the intensity information of the four polarization channels. Based on the complex amplitude distribution at each wavelength and the polarization parameters, the phase distortion of the target particle field under test is compensated and speckle noise is suppressed, and the quantization parameters of the target particle field under test are generated based on the enhanced image.

2. The three-wavelength and four-angle polarization-sensitive holographic imaging system according to claim 1, characterized in that, After acquiring the original holographic data for three wavelength channels and the intensity information for four polarization channels, diffraction inversion is performed on the holographic data for each wavelength channel. The diffraction inversion is based on the Fresnel-Kirchhoff diffraction theory, and the holographic fringes formed by the interference of the object field and the reference field are represented as follows: in, Represents the convolution operator. For the inverse Fourier transform of the diffraction nucleus G, For object light field, For the reference light field, the diffraction nucleus G is represented as: ),in,( () represents spatial frequency. The wavelength of coherent light. The axial propagation distance, The imaginary unit; Camera on the imaging surface The recorded holographic light intensity is: The target light field is reproduced using digital refocusing methods: Where * denotes conjugation. For the inverse Fourier transform of the conjugate term of the diffraction nucleus G.

3. A three-wavelength and four-angle polarization-sensitive holographic imaging system according to claim 2, characterized in that, The compensation for the phase distortion of the target particle field based on the complex amplitude distribution at each wavelength and the polarization parameters is specifically as follows: The differences in phase distortion across different wavelengths are analyzed, and compensation is performed based on the diffraction distance and polarization state. This compensation is achieved through frequency-domain constrained iterative inversion, correcting the distorted phase during the iteration process. Specifically: Based on synchronously acquired four-angle polarization holograms ( Transform it into a complex amplitude hologram with phase information: ,in, For reference optical amplitude, The intensity of the hologram recorded when the linear polarization direction is 0° is the light intensity. The light intensity of the hologram recorded when the linear polarization direction is 45° is shown. The light intensity of the hologram recorded when the linear polarization direction is 90° is shown. The intensity of the hologram is recorded when the linear polarization direction is 135°. Then, numerical reconstruction is performed based on the target light field to interpret the phase and amplitude of the target light field. Iterative inversion is performed through frequency domain constraints to correct the distorted phase.

4. A three-wavelength and four-angle polarization-sensitive holographic imaging system according to claim 3, characterized in that, After phase distortion compensation, speckle noise is suppressed. Specifically, speckle noise is suppressed through wavelength synthesis based on information from three wavelength channels, and its expression is as follows: in, The wavelength of red light The wavelength of green light It is the wavelength of blue light. The equivalent wavelength for the combination of red and green light. The equivalent wavelength for the combination of red and blue light. The equivalent wavelengths of green and blue light are used to synthesize the image. By weighted fusion of the three single-wavelength reconstruction results and the three synthesized wavelength reconstruction results, a multi-wavelength fusion reconstruction image with enhanced signal-to-noise ratio is obtained.

5. A three-wavelength and four-angle polarization-sensitive holographic imaging system according to claim 4, characterized in that, Also includes: Multi-sample noise reduction is performed on the scattered photons' random characteristics in the multi-wavelength fused reconstructed image, specifically as follows: The multi-channel information is randomly rearranged and combined in the spatial or frequency domain to generate multiple independent samples. Random scattering noise is suppressed by averaging the reconstruction results of these multiple samples. The expression for the multi-channel information rearrangement is as follows: in, For all wavelength channels, Summing over all polarization angles, A random coefficient matrix of wavelength dimension. The random coefficient matrix is ​​the polarization dimension. This is the original hologram corresponding to the wavelength and polarization channel. These are synthetic hologram samples generated by random rearrangement.

6. A three-wavelength and four-angle polarization-sensitive holographic imaging system according to claim 4, characterized in that, The quantization parameters for generating the target particle field based on the enhanced image are specifically as follows: By using the recorded color polarization holograms of dense particle fields, multiple image features of particles are extracted from multi-channel information. The evolution of particle image similarity (F1), gradient sparsity (F2), light intensity uniformity (F3), edge sharpness (F4), and background consistency (F5) along the reconstruction depth direction is mainly interpreted. These feature expressions are: in, Let be the set of pixels of a certain particle in the first frame of the image. Let N be the set of pixels to be matched in the second frame image, and n and m be the pixel indices of sets N and M, respectively. Let be the joint probability distribution of the two pixel sets. Marginal probability distribution of set N Let M be the marginal probability distribution of the set, and i and j be the pixel coordinate indices within the particle image. Let (i,j) be the magnitude of the image gradient. The light intensity value at position (i,j) inside the particle. Let (i,j) be the gradient magnitude at position (i,j) on the particle edge region. Let be the light intensity value at position (i,j) in the background region surrounding the particle; By fusing effective features through weighted allocation, an integrated identification criterion for single and overlapping particles is established. Adaptive curve fitting and extreme value search strategies are used to achieve accurate particle edge segmentation and axial positioning. Combined with clear particle morphology, spatiotemporal constraints, and polarization response, the three-dimensional displacement parameters of all particles between two frames are obtained.

7. A three-wavelength and four-angle polarization-sensitive holographic imaging system according to claim 1, characterized in that, The image acquisition module is a color polarization camera, and the photosensitive chip of the color polarization camera integrates an RGB Bayer array and a four-angle micro polarizer array.

Images