A method and system for reconstructing three-dimensional light-thermal parameters of a combustion process

Abstract

The application discloses a three-dimensional light-thermal parameter reconstruction method and system for a combustion process, and belongs to the field of combustion diagnosis.The application solves the problems of low measurement accuracy in a high-temperature environment and poor accuracy of extracting effective signals in a complex background existing in the prior art light field parameter measurement.The application adopts a polarization spectrum camera and a light field camera to synchronously perform multi-angle data acquisition on a combustion area, and acquires polarization spectrum data and light field data of the combustion area;light field depth information is extracted from the light field data, a three-dimensional structure of the combustion area is constructed by combining the light field depth information with a light field reconstruction algorithm, spectral analysis is performed on the polarization spectrum data, and polarization degree and polarization angle information of the polarization spectrum are acquired;the three-dimensional structure is combined with a CT reconstruction algorithm; a three-dimensional distribution field of particulate matters in the combustion process is acquired; and based on the polarization spectrum information and the preprocessed light field data, the three-dimensional temperature field and the concentration field distribution of gases in the combustion area are calculated by using the optimized three-dimensional structure.The application is mainly used for combustion product parameter acquisition.

Description

Technical Field

[0001] The invention belongs to the field of combustion diagnosis. Background Art

[0002] Combustion is a fundamental process that converts chemical energy into heat and light, and is widely used in many fields, including industrial production, energy conversion, and environmental protection. Accurate diagnosis and control of the combustion process not only improves energy efficiency but also effectively reduces pollutant emissions, thereby achieving energy conservation and emission reduction. Combustion diagnostic technology is primarily used to monitor various physical and chemical phenomena generated during the combustion process, such as temperature, concentration, and flow fields, to achieve real-time monitoring and optimized control of the combustion process. With the continuous advancement of science and technology, combustion diagnostic technology is also developing, gradually moving towards high-precision, high-resolution, and multi-parameter measurement.

[0003] Polarization spectral imaging is an emerging technology that combines polarization measurement and spectral imaging, capable of simultaneously acquiring both spectral and polarization information about a target object. The polarization state of light carries a wealth of useful information about the microstructure and surface properties of a material. By analyzing this polarization information, parameters such as the material's refractive index, surface roughness, thickness, and stress distribution in composite materials can be extracted. Spectral imaging acquires images of an object at different wavelengths and analyzes its spectral characteristics to derive the composition, concentration, and distribution of the material. Combining polarization measurement with spectral imaging enables the simultaneous acquisition of multi-parameter, multi-dimensional information in complex scenes. Polarization spectral imaging offers unique advantages in combustion diagnostics. Combustion produces a large amount of soot and particulate matter. The polarization characteristics of these particles can reveal their size distribution and shape. Analyzing this polarization information provides a more accurate understanding of the formation and evolution of particulate matter during combustion. Spectral imaging can also measure the temperature and concentration distributions during combustion, enabling comprehensive monitoring of the combustion process.

[0004] A light field camera is an imaging device that can record the spatial distribution of light. Unlike traditional cameras that can only record light intensity, light field cameras add a microlens array in front of the sensor to record the direction and position of light, thereby achieving stereoscopic imaging and depth information acquisition of three-dimensional scenes. The emergence of light field camera technology has provided a new approach for high-precision three-dimensional imaging and full light field reconstruction. In the field of combustion diagnosis, light field cameras can be used to obtain three-dimensional structural information of the combustion area. By recording light information from different directions, light field cameras can reconstruct the three-dimensional temperature and concentration field distribution of the combustion area, providing important data support for in-depth analysis of the combustion process. In addition, light field cameras can also achieve real-time monitoring of dynamic combustion processes. By capturing instantaneous light field information, the dynamic evolution of flames and smoke during the combustion process can be reconstructed.

[0005] Combining polarization spectral imaging technology with a light-field camera enables the reconstruction of three-dimensional photothermal parameters of the combustion process. Specifically, polarization spectral imaging captures the polarization characteristics and spectral information of particles during combustion, thereby determining parameters such as particle size, shape, and composition. Light-field cameras record the three-dimensional light distribution within the combustion region and reconstruct the three-dimensional temperature and concentration fields within the combustion region. This approach offers the advantage of simultaneous acquisition and high-precision reconstruction of multi-parameter and multi-dimensional information, providing strong technical support for comprehensive combustion process diagnosis. Accurate measurement of three-dimensional photothermal parameters during combustion provides a deeper understanding of combustion mechanisms, optimizes combustion conditions, improves combustion efficiency, and reduces pollutant emissions, thereby achieving energy conservation and emission reduction goals. Currently, polarization spectral imaging and light-field camera technologies have established a substantial research foundation in the field of combustion diagnosis. While polarization spectral imaging has achieved significant results in particle measurement and material analysis, its application in high-temperature, high-velocity combustion environments still faces numerous challenges, such as high-temperature background interference and weak signal strength. Light-field cameras offer unique advantages in three-dimensional imaging and depth information acquisition, but their spatial resolution and data processing complexity still require further improvement. The 3D photothermal parameter reconstruction method based on polarization spectral imaging technology and light field cameras still needs to overcome a series of technical challenges. For example, how to achieve high-precision polarization spectral measurement in high-temperature combustion environments, how to extract effective signals in complex backgrounds, and how to efficiently process and analyze massive amounts of light field data are all issues that require in-depth research. Summary of the Invention

[0006] The present invention aims to solve the problems of low measurement accuracy in high-temperature environments and poor accuracy in extracting effective signals under complex backgrounds in existing light field parameter measurements. A method and system for reconstructing three-dimensional photothermal parameters of a combustion process are now provided.

[0007] The method for reconstructing three-dimensional photothermal parameters of a combustion process according to the present invention comprises:

[0008] Step 1: Use a polarization spectrum camera and a light field camera to synchronously collect data from multiple angles of the burning area to obtain polarization spectrum data and light field data of the burning area;

[0009] Step 2: Preprocessing the polarization spectrum data and the light field data respectively to obtain preprocessed polarization spectrum data and preprocessed light field data;

[0010] Step 3: Extracting light field depth information from the light field data, using the light field depth information in combination with a light field reconstruction algorithm to construct a three-dimensional structure of the combustion area, introducing the pre-processed polarization spectrum information into the three-dimensional structure of the combustion area, and optimizing the three-dimensional structure of the combustion area; obtaining the optimized three-dimensional structure;

[0011] Step 4: Perform spectral analysis on the pre-processed polarization spectrum data to obtain polarization degree and polarization angle information of the polarization spectrum; use the polarization degree and polarization angle information to extract the size and shape characteristics of solid particles during the combustion process; use the optimized three-dimensional structure in combination with a CT reconstruction algorithm to obtain a three-dimensional distribution field of particulate matter during the combustion process;

[0012] Based on the pre-processed polarization spectrum information and the pre-processed light field data, the three-dimensional temperature field and concentration field distribution of the gas in the combustion area are calculated using the optimized three-dimensional structure.

[0013] Furthermore, in the present invention, the frame rate of the polarization spectrum camera is ≥100fps, the spectral range is: 400-1000nm, covering visible light to near-infrared light, the polarization resolution is ≤0.1°, and the spectral resolution is ≤1nm.

[0014] Furthermore, in the present invention, the high spatial resolution of the light field camera is ≥4K, and the dynamic range is ≥120dB.

[0015] Furthermore, in the present invention, in step 2, the polarization spectrum data and the light field data are preprocessed separately by:

[0016] The polarization spectrum data is sequentially subjected to dark current correction, flat field correction, spectrum calibration, polarization calibration, denoising and background interference elimination to obtain pre-processed polarization spectrum data;

[0017] The light field data is subjected to denoising and microlens array calibration in sequence to obtain preprocessed light field data.

[0018] Furthermore, in the present invention, in step 3, the method for constructing the three-dimensional structure of the combustion area by using the light field depth information combined with the light field reconstruction algorithm is:

[0019] Step 31: Extract the 4D light field function from the light field data, and use the local plane approximation algorithm of the structure tensor to calculate the depth information of each point in the combustion area; use the depth information to obtain a depth map of the combustion area;

[0020] Step 32: Applying a shift-and-add algorithm in combination with Gaussian weighting to synthesize a fully focused image based on the depth map;

[0021] Step 3. Use the depth map and the fully focused image to generate an initial point cloud, and use the Poisson surface reconstruction algorithm to generate a continuous surface from the initial point cloud to obtain the three-dimensional structure of the burning area.

[0022] Furthermore, in the present invention, in step 3, the specific method for obtaining the optimized three-dimensional structure is:

[0023] Step 3: Extract feature points from the full-focus image and polarization spectrum data respectively, and match the feature points using a geometric consistency method;

[0024] Step 3 and 4: Use the Markov random field (MRF) model to construct the matched polarization spectrum data as the depth prior, refine the depth map, use barycentric coordinate interpolation for texture mapping, and map the refined depth map to the continuous surface to obtain the optimized three-dimensional structure.

[0025] Furthermore, in the present invention, in step 4, the specific method for obtaining the polarization degree and polarization angle information of the polarization spectrum is:

[0026] Step 41: Extract the measured intensity data and Stokes vector from the polarization spectrum data, and obtain the Mueller matrix through least squares optimization;

[0027] Step 42: Decompose the Mueller matrix using a least squares method-based Mueller matrix decomposition algorithm to obtain degree of polarization (DoP) information of the polarization spectrum, and use the degree of polarization (DoP) information to obtain the polarization angle of the polarization spectrum.

[0028] Furthermore, in the present invention, in step 4, the method for extracting the size and shape characteristics of solid particles during the combustion process using polarization degree and polarization angle information is:

[0029] Step 43: Construct a forward model based on Mie scattering theory, and use the T-matrix method to expand the forward model to obtain an expanded model that can identify particles of different shapes and sizes;

[0030] Step 44: Using the extended model and the polarization degree and polarization angle information, the Levenberg-Marquardt algorithm is used to analyze the pre-processed polarization spectrum data to identify the size and shape characteristics of the solid particles in the combustion process.

[0031] Furthermore, in the present invention, step 4 further includes the steps of identifying particulate matter during the combustion process and calculating the concentrations of different types of solid particulate matter, specifically including:

[0032] The continuous wavelet transform (CWT) algorithm is used to perform peak detection on the polarization spectrum data and light field data of the combustion area to obtain characteristic peak data. The characteristic peak data is compared with the data in the existing combustion product spectrum database to obtain the type of particulate matter during the combustion process.

[0033] The Beer-Lambert law is used to calculate the concentration of different types of particulate matter during the combustion process.

[0034] Furthermore, in the present invention, in step 4, the specific method for calculating the three-dimensional temperature field distribution of the gas in the combustion area is:

[0035] Step M1: using the pre-processed polarization spectrum information, the Fresnel equation and the Kramers-Kronig relationship are used to estimate the emissivity of the gas combustion products; and calculating the complex refractive index using the estimated emissivity of the gas combustion products;

[0036] Step M2: Use the complex refractive index to calculate the projection matrix A, and use the projection matrix A to establish a linear equation AT=b, where A is the projection matrix, T is the temperature field, and b is the observation value; use algebraic reconstruction technology (ART) to solve the linear equation AT=b to obtain a 3D temperature field of multi-angle temperature projection.

[0037] Furthermore, in the present invention, in step 4, the specific method for calculating the concentration field distribution of the gas in the combustion area is:

[0038] Step N1: Establish a temperature radiation transfer equation (RTE), use polarization spectrum data, and solve the RTE using the discrete coordinate method (DOM) to obtain the radiation intensity distribution of each point in the combustion area;

[0039] Step N2: applying the Tikhonov regularization method to solve the inverse problem of the radiation intensity distribution at each point in the combustion area, and deriving the initial concentration distribution of the gas in the combustion area;

[0040] Step N3: Use a non-negative matrix factorization (NMF) algorithm to separate and analyze the initial concentration distribution of the gas in the combustion area to obtain the final concentration field distribution of the gas in the combustion area.

[0041] A three-dimensional photothermal parameter reconstruction system for the combustion process, including a polarization spectrum camera, a light field camera, a synchronous control system, a rotating stage, and a computer;

[0042] The polarization spectrum camera is used to collect polarization spectrum data of the burning area at multiple angles and send it to the computer, and the light field camera is used to collect spectrum data of the burning area at multiple angles and send it to the computer;

[0043] The rotating stage is used to control the polarization spectrum camera and the light field camera to rotate along the combustion area;

[0044] The synchronous control system synchronizes the image acquisition moments of the polarization spectrum camera and the light field camera;

[0045] The computer uses polarization spectrum data and light field data to construct the three-dimensional structure of the combustion area and optimize it. Using the optimized three-dimensional structure, combined with the CT reconstruction algorithm, the three-dimensional distribution field of particulate matter during the combustion process is obtained, and the three-dimensional temperature field and concentration field distribution of the gas in the combustion area are calculated.

[0046] Furthermore, in the present invention, the polarization spectrum camera includes a lens 1, a polarization optical element, a tunable filter, a spectrum splitting system and a photoelectric detection system 1;

[0047] The lens I is aimed at the combustion area and is used to capture an image of the combustion area. The image of the combustion area is filtered by a polarization optical element and then transmitted to an adjustable filter. The light after wavelength selection filtering by the adjustable filter is incident on a spectral spectrometer system (104). The spectral spectrometer system (104) separates the spectrum of the filtered light, and the separated spectrum is converted into an electrical signal by a photoelectric detection system (105).

[0048] In this embodiment, the photoelectric detection system I105 includes a high-resolution CMOS sensor, a filter array and a high-speed data acquisition card. The spectrum separated by the spectral spectrometer system (104) is incident on the CMOS sensor through the filter array. The CMOS sensor is used to convert the optical signal passing through the filter array into an electrical signal, which is then transmitted to the computer through the high-speed data acquisition card.

[0049] Furthermore, in the present invention, the light field camera includes a lens II, a microlens array, and a photoelectric detection system II;

[0050] Lens II is aimed at the burning area to capture the image of the burning area and transmit the image of the burning area to the microlens array; the microlens array is used to enhance the quality of the image of the burning area by focusing, and send the image of the burning area with enhanced quality to the photoelectric detection system II.

[0051] This method enables high-precision 3D reconstruction and simultaneous multi-parameter measurement. Combining polarization spectral imaging and light field camera technology, it achieves high-precision 3D reconstruction of the combustion process, providing more detailed and accurate information about the combustion zone than traditional methods. It simultaneously captures multiple optical and thermal parameters within the combustion zone, including the temperature field, concentration distribution, scattering coefficient, and absorption coefficient, providing rich data support for a comprehensive understanding of the combustion process. Optical imaging methods enable non-contact and non-invasive measurement of the combustion process, avoiding the interference of traditional measurement methods. Light field camera technology achieves high-spatial-resolution 3D reconstruction, enabling precise location and analysis of the fine structure of the combustion zone. Combining multi-angle polarization spectral data provides more comprehensive information about combustion products, facilitating their accurate identification and quantification. Through optimized algorithms and parallel computing, this method enables near-real-time data processing and parameter reconstruction, making it suitable for real-time monitoring and control applications. Compared to traditional multi-sensor combinations, this method integrates multiple measurement functions, reducing overall system complexity and cost. This provides new observational tools and data sources for combustion science research, deepening our understanding of complex combustion phenomena. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a flow chart of the method of the present invention;

[0053] Figure 2 It is a principle block diagram of the system of the present invention;

[0054] Figure 3 This is the principle block diagram of the polarization spectrum camera and light field camera;

[0055] Figure 4 It is a block diagram of the system principle described in a specific embodiment. DETAILED DESCRIPTION

[0056] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other in the absence of conflict.

[0057] Specific implementation method 1: refer to Figure 1 Specifically describing this embodiment, the method for reconstructing three-dimensional photothermal parameters of a combustion process described in this embodiment includes:

[0058] Step 1: Use a polarization spectrum camera and a light field camera to synchronously collect data from multiple angles of the burning area to obtain polarization spectrum data and light field data of the burning area;

[0059] Step 2: Preprocessing the polarization spectrum data and the light field data respectively to obtain preprocessed polarization spectrum data and preprocessed light field data;

[0060] Step 3: Extracting light field depth information from the light field data, using the light field depth information in combination with a light field reconstruction algorithm to construct a three-dimensional structure of the combustion area, introducing the pre-processed polarization spectrum information into the three-dimensional structure of the combustion area, and optimizing the three-dimensional structure of the combustion area; obtaining the optimized three-dimensional structure;

[0061] Step 4: Perform spectral analysis on the pre-processed polarization spectrum data to obtain polarization degree and polarization angle information of the polarization spectrum; use the polarization degree and polarization angle information to extract the size and shape characteristics of solid particles during the combustion process; use the optimized three-dimensional structure in combination with a CT reconstruction algorithm to obtain a three-dimensional distribution field of particulate matter during the combustion process;

[0062] Based on the pre-processed polarization spectrum information and the pre-processed light field data, the three-dimensional temperature field and concentration field distribution of the gas in the combustion area are calculated using the optimized three-dimensional structure.

[0063] Furthermore, in this embodiment, the frame rate of the polarization spectrum camera is ≥100 fps, the spectral range is: 400-1000 nm, covering visible light to near-infrared light, the polarization resolution is ≤0.1°, and the spectral resolution is ≤1 nm.

[0064] Furthermore, in this embodiment, the high spatial resolution of the light field camera is ≥4K, and the dynamic range is ≥120dB.

[0065] Furthermore, in this embodiment, in step 2, the polarization spectrum data and the light field data are preprocessed separately by:

[0066] The polarization spectrum data is sequentially subjected to dark current correction, flat field correction, spectrum calibration, polarization calibration, denoising and background interference elimination to obtain pre-processed polarization spectrum data;

[0067] The light field data is subjected to denoising and microlens array calibration in sequence to obtain preprocessed light field data.

[0068] Furthermore, in this embodiment, in step 3, the method of constructing the three-dimensional structure of the combustion area by using the light field depth information combined with the light field reconstruction algorithm is as follows:

[0069] Step 31: Extract the 4D light field function from the light field data, and use the local plane approximation algorithm of the structure tensor to calculate the depth information of each point in the combustion area; use the depth information to obtain a depth map of the combustion area;

[0070] Step 32: Applying a shift-and-add algorithm in combination with Gaussian weighting to synthesize a fully focused image based on the depth map;

[0071] Step 3. Use the depth map and the fully focused image to generate an initial point cloud, and use the Poisson surface reconstruction algorithm to generate a continuous surface from the initial point cloud to obtain the three-dimensional structure of the burning area.

[0072] Furthermore, in this embodiment, in step 3, the specific method for obtaining the optimized three-dimensional structure is:

[0073] Step 3: Extract feature points from the full-focus image and polarization spectrum data respectively, and match the feature points using a geometric consistency method;

[0074] Step 3 and 4: Use the Markov random field (MRF) model to construct the matched polarization spectrum data as the depth prior, refine the depth map, use barycentric coordinate interpolation for texture mapping, and map the refined depth map to the continuous surface to obtain the optimized three-dimensional structure.

[0075] Furthermore, in this embodiment, in step 4, the specific method for obtaining the polarization degree and polarization angle information of the polarization spectrum is:

[0076] Step 41: Extract the measured intensity data and Stokes vector from the polarization spectrum data, and obtain the Mueller matrix through least squares optimization;

[0077] Step 42: Decompose the Mueller matrix using a least squares method-based Mueller matrix decomposition algorithm to obtain degree of polarization (DoP) information of the polarization spectrum, and use the degree of polarization (DoP) information to obtain the polarization angle of the polarization spectrum.

[0078] Furthermore, in this embodiment, in step 4, the method for extracting the size and shape characteristics of solid particles during the combustion process using polarization degree and polarization angle information is:

[0079] Step 43: Construct a forward model based on Mie scattering theory, and use the T-matrix method to expand the forward model to obtain an expanded model that can identify particles of different shapes and sizes;

[0080] Step 44: Using the extended model and the polarization degree and polarization angle information, the Levenberg-Marquardt algorithm is used to analyze the pre-processed polarization spectrum data to identify the size and shape characteristics of the solid particles in the combustion process.

[0081] Furthermore, in this embodiment, step 4 further includes steps of identifying particulate matter during the combustion process and calculating the concentrations of different types of solid particulate matter, specifically including:

[0082] The continuous wavelet transform (CWT) algorithm is used to perform peak detection on the polarization spectrum data and light field data of the combustion area to obtain characteristic peak data. The characteristic peak data is compared with the data in the existing combustion product spectrum database to obtain the type of particulate matter in the combustion process.

[0083] The Beer-Lambert law is used to calculate the concentration of different types of particulate matter during the combustion process.

[0084] Furthermore, in this embodiment, in step 4, the specific method for calculating the three-dimensional temperature field distribution of the gas in the combustion area is:

[0085] Step M1: using the pre-processed polarization spectrum information, the Fresnel equation and the Kramers-Kronig relationship are used to estimate the emissivity of the gas combustion products; and calculating the complex refractive index using the estimated emissivity of the gas combustion products;

[0086] Step M2: Use the complex refractive index to calculate the projection matrix A, and use the projection matrix A to establish a linear equation AT=b, where A is the projection matrix, T is the temperature field, and b is the observation value; use algebraic reconstruction technology (ART) to solve the linear equation AT=b to obtain a 3D temperature field of multi-angle temperature projection.

[0087] Furthermore, in this embodiment, in step 4, the specific method for calculating the concentration field distribution of the gas in the combustion area is:

[0088] Step N1: Establish a temperature radiation transfer equation (RTE), use polarization spectrum data, and solve the RTE using the discrete coordinate method (DOM) to obtain the radiation intensity distribution of each point in the combustion area;

[0089] Step N2: applying the Tikhonov regularization method to solve the inverse problem of the radiation intensity distribution at each point in the combustion area, and deriving the initial concentration distribution of the gas in the combustion area;

[0090] Step N3: Use a non-negative matrix factorization (NMF) algorithm to separate and analyze the initial concentration distribution of the gas in the combustion area to obtain the final concentration field distribution of the gas in the combustion area.

[0091] Specific implementation method 2: Figure 2 and Figure 3 This embodiment describes a three-dimensional photothermal parameter reconstruction system for a combustion process, comprising a polarization spectrum camera 1, a light field camera 2, a synchronous control system 3, a rotating stage 4, and a computer 5.

[0092] The polarization spectrum camera 1 is used to collect polarization spectrum data of the burning area at multiple angles and send it to the computer 5. The light field camera 2 is used to collect spectrum data of the burning area at multiple angles and send it to the computer 5.

[0093] The rotating stage is used to control the polarization spectrum camera 1 and the light field camera 2 to rotate along the combustion area;

[0094] The synchronous control system 3 performs synchronous control on the image acquisition moments of the polarization spectrum camera and the light field camera;

[0095] Computer 5 uses polarization spectrum data and light field data to construct a three-dimensional structure of the combustion area and optimizes it. Using the optimized three-dimensional structure, combined with a CT reconstruction algorithm, it obtains the three-dimensional distribution field of particulate matter during the combustion process and calculates the three-dimensional temperature field and concentration field distribution of the gas in the combustion area.

[0096] Furthermore, in this embodiment, the polarization spectrum camera includes a lens 1101, a polarization optical element 102, a tunable filter 103, a spectrum splitting system 104 and a photoelectric detection system 1105;

[0097] The lens I101 is aimed at the combustion area and is used to capture the image of the combustion area. The image of the combustion area is filtered by the polarization optical element 102 and transmitted to the tunable filter 103. The light after wavelength selection filtering by the tunable filter 103 is incident on the spectral spectrometry system 104. The spectral spectrometry system 104 separates the spectrum of the filtered light, and the separated spectrum is converted into an electrical signal by the photoelectric detection system (105).

[0098] In this embodiment, the photoelectric detection system I105 includes a high-resolution CMOS sensor, a filter array and a high-speed data acquisition card. The spectrum separated by the spectral spectrometry system 104 is incident on the CMOS sensor through the filter array. The CMOS sensor is used to convert the optical signal passing through the filter array into an electrical signal, which is then transmitted to the computer through the high-speed data acquisition card.

[0099] Furthermore, in this embodiment, the light field camera includes a lens II 201 , a microlens array 202 , and a photoelectric detection system II 203 ;

[0100] The lens II201 is aimed at the burning area to capture the image of the burning area and transmit the image of the burning area to the microlens array 202; the microlens array 202 is used to enhance the quality of the image of the burning area by focusing, and send the image of the burning area after quality enhancement to the photoelectric detection system II203.

[0101] The specific process is as follows Figure 1 As shown in the preceding steps, this method leverages the advantages of polarization spectral imaging technology and light field cameras to achieve high-precision reconstruction of the three-dimensional photothermal parameters of the combustion process. This method not only provides more comprehensive and accurate combustion diagnostic information but also provides an important basis for combustion process optimization and control.

[0102] The polarization spectrum camera described in step 1 is used to obtain polarization and spectrum information during the combustion process. Figure 1-3 As shown, the main components of the polarization spectroscopic camera 1 are the optical system, imaging sensor, spectral splitting system, photoelectric detection system, synchronization control system, and data acquisition and processing system. The optical system consists of a main lens 101, a tunable filter 103, and a polarization optical element 102. The main lens 101 covers the required spectral range and uses a liquid crystal tunable filter (LCTF) (or acousto-optic tunable filter (AOTF)) to select a spectrum with a specific wavelength. A polarizer and a quarter-wave plate are installed in front of or behind the lens for polarization state adjustment and analysis. The spectral splitting system 104 consists of a prism or grating and a slit. The prism or grating is used to decompose the incident light into a spectrum of different wavelengths, and the slit is used to define the light propagation direction and spectral resolution. The photoelectric detection system 1 includes a high-resolution CMOS sensor and a filter array 105. The high-resolution CMOS sensor 105 has a resolution of at least 4K (4096 x 3072 pixels), a frame rate ≥100 fps, a dynamic range ≥14 bits, and a peak quantum efficiency ≥80%. A filter array is installed in front of the sensor to capture images of different polarization states and wavelengths. A photomultiplier tube (PMT) or a high-sensitivity detector is then used to detect the spectral signal after splitting and convert the spectral signal into an electrical signal; the synchronous control system 3 includes a high-precision clock source and a trigger system. The high-precision clock source is used to synchronize the switching of the polarizer and the filter, and the trigger system realizes the synchronous triggering of the sensor, polarizer and filter. In the present invention, the data after photoelectric conversion is also collected by a high-speed data acquisition card. The high-speed data acquisition card is connected to the computer via PCIe or other high-speed interfaces, and the high-performance computer 5 processes and stores the collected image data, and uses software for real-time data acquisition, polarization analysis, spectral analysis and data fusion.

[0103] The light field camera described in step 1 is used to acquire three-dimensional light field information of the combustion process. The device is selected to have a high spatial resolution (≥4K) to obtain detailed light field data, a large aperture (f / 2.0 or larger) to improve imaging quality in low-light environments, a microlens array design to record light direction information, and a high dynamic range (≥120dB) to cope with the wide brightness range during the combustion process. The light field camera structure mainly includes lens II 201 and microlens array 202. Lens II 201 covers the required spectral range, and microlens array 202 is installed behind lens II to capture light from different angles. The imaging sensor is a high-resolution CMOS sensor 203 with a resolution of at least 4K (4096 x 3072 pixels), a frame rate ≥100 fps, a dynamic range ≥14 bits, and a peak quantum efficiency ≥80%. The specific data acquisition process involves using a high-precision rotating stage 6, mounting the polarization spectral camera on the stage, ensuring that the optical axis is aligned with the center of the combustion area 5. The camera parameters (exposure time, frame rate, wavelength range) are set using the camera software. Multi-angle acquisition is achieved by controlling the rotating stage. The combustion process is scanned 360 degrees, with data collected every 15 degrees, for a total of 24 angles. Synchronous control system 3 ensures data synchronization between polarization camera 1 and light field camera 2. The device features a high-precision clock (error < 1 μs), multi-channel trigger output, programmable delay and exposure time control, and more. The collected raw data is saved in ENVI format for easy processing.

[0104] The polarization spectrum data preprocessing described in step 2 includes dark current correction, flat field correction, spectral calibration, polarization calibration, denoising and background interference elimination processes. The dark current correction process is to obtain an image without light (dark field image) under the same exposure time and temperature conditions, collect multiple dark field images (≥20 images), calculate the average value as the dark current benchmark, and subtract the benchmark dark field image from the original data; the flat field correction process is to use a uniform light source (such as an integrating sphere) to obtain a flat field image to correct the unevenness of the optical system. First, collect multiple flat field images (≥20 images), calculate the average value, subtract the dark current, normalize, and correct the original data; the spectral calibration process is to use a standard light source with known emission lines (such as a mercury argon lamp) for wavelength calibration. The steps are to identify the characteristic peak position of the standard light source, establish a corresponding relationship between the pixel position and the wavelength, and usually use polynomial fitting. Combine and map the raw data onto the calibrated wavelength axis; the polarization calibration process uses a standard polarizer with a known polarization state for calibration. The steps are to place the standard polarizer at different angles (0°, 45°, 90°, 135°), collect images, calculate the difference between the actual Stokes parameters and the theoretical values, establish a calibration matrix, and correct the polarization information of the raw data; the denoising process includes spectral dimension denoising using Savitzky-Golay filters, spatial dimension denoising using anisotropic diffusion filters, polarization dimension processing using polarization consistency constraints, outlier detection and processing, etc. This multi-dimensional, multi-step denoising method can better protect the characteristics of polarization spectral data while effectively removing noise. The background interference elimination process uses time series analysis combined with principal component analysis (PCA) method. The steps are to perform PCA decomposition on multiple consecutive frames of images, identify the principal component representing the background (usually the component with the least change), and subtract the background principal component from the raw data;

[0105] The light field data preprocessing in step 2 includes denoising and microlens array calibration. The non-local means (NLM) denoising algorithm is used to preserve edge and texture information for denoising, and a whiteboard image is used for geometric calibration of the microlens array.

[0106] Step three obtains the optimized three-dimensional structure, which specifically includes the following processes: first, reconstruct the 4D light field function using light field data, perform depth estimation and full-focus image synthesis, and then generate a point cloud and perform surface reconstruction; and fuse polarization spectrum information to perform operations such as feature point matching, geometric consistency optimization, depth map refinement and texture mapping to achieve high-precision and high-resolution three-dimensional structure reconstruction of the combustion area.

[0107] A 4D light field function Δf = f1 - f2L(u, v, s, t) is extracted from the original image using a microlens lookup table (LUT), where u and v represent spatial coordinates and s and t represent angular coordinates. A multi-view depth estimation algorithm is implemented using a local plane approximation algorithm based on a structured tensor. A shift-and-add algorithm, combined with Gaussian weighting, is used to synthesize an all-in-focus image based on the estimated depth map. Back-projection is used to project each pixel into 3D space, and an initial point cloud is generated using the depth map and all-in-focus image. A Poisson surface reconstruction algorithm is used to generate a continuous surface from the point cloud. An octree data structure is used to optimize computational efficiency. Polarization and spectral information fusion, including feature point matching, geometric consistency optimization, depth map refinement, and texture mapping, is employed to achieve 3D structure acquisition. The SIFT (Scale-Invariant Feature Transform) algorithm is used to extract and match feature points. Feature point matching is performed between the all-in-focus image and the polarization and spectral image, and RANSAC is used to eliminate false matches. The matched feature points are used to optimize the relative pose between the light field camera and the polarization and spectral camera for geometric consistency. A Markov random field (MRF) model is constructed, using the degree of polarization as a depth prior and leveraging polarization information to refine the depth map. Texture mapping is performed using barycentric coordinate interpolation to map the polarization spectrum information onto the 3D model surface. Through systematic reconstruction and optimization methods, the light field data and polarization spectrum information are fully utilized to achieve high-precision, high-resolution 3D structural reconstruction of the burned area.

[0108] Step 4 also includes the process of identifying combustion products. The analysis of polarization spectrum data, extraction of particulate matter characteristics, and identification of combustion products specifically include the following process:

[0109] 1) Perform polarization characteristic analysis, including Mueller matrix calculation and polarization characteristic parameter extraction. Using the least squares-based Mueller matrix decomposition algorithm, the complete 4x4 Mueller matrix is ​​calculated using the acquired multi-angle polarization spectrum data. The formula is I = M*S, where I is the measured intensity, M is the Mueller matrix, and S is the Stokes vector. Polarization characteristic parameter extraction is performed by extracting the degree of polarization (degree of linear polarization (DoLP) and degree of circular polarization (DoCP)) from the Mueller matrix.

[0110] 2) Extract particle size and shape characteristics; construct a forward model based on Mie scattering theory to calculate the theoretical polarization characteristics of particles of different sizes and shapes; use the T-matrix method to expand the model to handle non-spherical particles. Apply the Levenberg-Marquardt algorithm for nonlinear fitting and invert the particle size distribution. The objective function is to minimize the difference between the theoretical model and the measured data. The constraint condition is to use the lognormal distribution as a priori information to improve the inversion stability. The non-sphericity of the particles is estimated by using the variation of the polarization degree with the scattering angle. The shape factor χ is introduced, which is defined as the volume ratio of the equivalent sphere to the actual particle. By comparing the theoretical curves and the measured data under different χ values, the best fitting shape factor is determined.

[0111] 3) Identify and estimate combustion products. A continuous wavelet transform (CWT) algorithm is implemented for peak detection, and peak shape parameters (such as half-width and asymmetry) are used for peak screening. A combustion product spectral database is constructed, containing the locations and relative intensities of characteristic peaks of common combustion products. A spectral matching algorithm based on cosine similarity is implemented, and machine learning methods (such as support vector machines (SVM)) are used to improve identification accuracy. Concentration estimation is performed using a modified Beer-Lambert law, accounting for multicomponent mixing effects. The formula is: A(λ) = Σ[εi(λ)*ci*l]+S(λ), where A(λ) is absorbance, εi(λ) is the molar absorption coefficient, ci is the concentration, l is the optical path length, and S(λ) is the scattering term. A concentration prediction model is developed using partial least squares regression (PLSR) to address spectral collinearity.

[0112] 4) Conduct a joint polarization-spectral analysis, construct a two-dimensional polarization-spectral correlation diagram, analyze the relationship between particle characteristics and chemical composition, and use singular value decomposition (SVD) to extract key characteristic patterns. Extract spatiotemporal distribution features, apply empirical mode decomposition (EMD) to analyze the temporal evolution of particle characteristics and concentrations, and use Moran's I index to assess the spatial autocorrelation of concentration distributions.

[0113] 5) Conduct uncertainty analysis. First, perform a Monte Carlo simulation, subjecting the measured data to multiple random perturbations, repeat the analysis, and calculate confidence intervals for key parameters (such as average particle size and main product concentration). Then, perform a sensitivity analysis using the Sobol index to assess the impact of different input parameters on the final results, identify key factors, and provide a basis for experimental optimization.

[0114] 6) Validate the results. First, perform cross validation and use k-fold cross validation to evaluate the stability of the concentration prediction model and calculate the root mean square error (RMSE) and determination coefficient (R 2) as a performance indicator for the model. Independent sample testing was then conducted, using pre-prepared samples of known concentrations for blind testing to assess the accuracy and reliability of the method. This systematic analysis method enabled the extraction of detailed characteristics of particulate matter during combustion from polarization spectral data, and the accurate identification of combustion products and their concentration distribution.

[0115] Step 4, which involves calculating the three-dimensional photothermal parameters, includes the temperature field distribution, concentration distribution, and scattering and absorption coefficients. Specifically, the following steps are involved: The three-dimensional temperature field distribution calculation includes spectral radiation temperature calculation, emissivity correction, and three-dimensional temperature field reconstruction. Using a nonlinear least-squares fitting algorithm based on Planck's blackbody radiation law, the radiation temperature is calculated from multi-wavelength spectral data. The Fresnel equation and the Kramers-Kronig relationship are applied to estimate the emissivity of the combustion products using polarization information, and the complex refractive index n+ik is calculated, where k is related to the emissivity ε. The linear equation system AT=b is iteratively solved, where A is the projection matrix, T is the temperature field, and b is the observed value. The algebraic reconstruction technique (ART) is used to reconstruct the 3D temperature field from multi-angle temperature projections. Total variation regularization is applied to improve reconstruction stability. The three-dimensional concentration distribution calculation includes solving the radiative transfer equation (RTE), solving the inverse problem, and analyzing multi-component concentration separation. The RTE is solved using the discrete ordinate method (DOM), discretized using the finite volume method, and solved using the SIMPLE algorithm. The inverse problem is solved using the Tikhonov regularization method, and the optimal regularization parameter α is selected using the L-curve method. The non-negative matrix factorization (NMF) algorithm is used to separate the multicomponent concentrations. The spectral data matrix V is decomposed into V ≈ WH, where W is the basis matrix (representing the pure component spectra) and H is the coefficient matrix (representing the concentrations). The NMF problem is solved using the alternating least squares (ALS) method. (3) Calculation of scattering and absorption coefficients involves the application of Mie scattering theory, inversion of the radiative transfer equation, and multiscale analysis. The scattering contributions of particles of different sizes are integrated, and the theoretical scattering coefficients are calculated using Mie scattering theory based on the previously obtained particle size distribution. Using the known temperature field and concentration distribution, the scattering and absorption coefficients in the RTE are inverted, and the conjugate gradient method is used to minimize the objective function. The discrete wavelet transform (DWT) is used, and the Daubechies wavelet basis is selected. The scattering and absorption coefficients are decomposed at multiple scales using the wavelet transform. The variations in scattering and absorption characteristics at different scales are analyzed, revealing the multiscale structure of the combustion process.

[0116] After obtaining the three-dimensional distribution field, three-dimensional temperature field, and three-dimensional concentration field of particulate matter during combustion, the present invention also performs multi-scale fusion and optimization to improve the accuracy and completeness of the reconstruction results. It also uses an iterative optimization algorithm to minimize the error between the reconstruction results and the observed data. The specific process includes multi-scale data construction, parameter fusion and scale transfer analysis, and iterative optimization algorithm analysis. ① Multi-scale data construction includes initial database construction and multi-scale three-dimensional model construction. A multi-resolution database of light field data and polarization spectral data is constructed. The images are smoothed using a Gaussian filter and then downsampled to form images of different resolutions. Gaussian filtering is applied to the original images to obtain low-frequency components. The filtered images are then downsampled, and the above steps are repeated to construct the database. Three-dimensional reconstruction is performed on the images at each resolution level to generate three-dimensional models of different scales. A light field reconstruction algorithm is used to reconstruct the three-dimensional structure at each resolution level, and the three-dimensional structure at each level is optimized based on the polarization spectral information. ② Parameter fusion and scale transfer analysis include multi-scale parameter initialization and inter-scale parameter fusion. Initialize parameters (temperature, concentration, scattering coefficient, and absorption coefficient) at the maximum resolution level. Based on the initial 3D reconstruction results, estimate the photothermal parameters at the high-resolution level. Use spectral matching to calibrate the initial parameters. Parameters are transferred and optimized step by step, fusing parameters at different scales from coarse to fine. Bilinear interpolation is used to transfer the low-resolution parameters to the high-resolution level. Local refinement methods (such as local search optimization) are used to precisely adjust the parameters at the high-resolution level. The iterative optimization algorithm includes objective function construction, optimization algorithm selection, and total variation regularization analysis. The objective function is constructed to minimize the error between the reconstruction and the observed data. Regularization terms are introduced to control overfitting. Iterative optimization is performed using the Levenberg-Marquardt algorithm or a variational method. The reconstruction results are estimated based on the current parameters. The error between the reconstruction and the observed data is calculated, and the parameters to be optimized are updated to minimize the error. A total variation regularization term is introduced to maintain the smoothness and edge-preserving properties of the reconstruction results.

[0117] Building on the steps already taken, an adaptive parameter adjustment mechanism needs to be established to improve the robustness of the reconstruction results. This involves optimizing the data acquisition strategy in real time, particularly increasing the sampling density in key areas. The specific processes of adaptive parameter adjustment and real-time optimization include adaptive parameter adjustment, feedback mechanisms and real-time optimization, and multi-objective optimization. ① Adaptive parameter adjustment includes error analysis and feedback, dynamic learning rate adjustment, and intelligent hyperparameter tuning. The error between the reconstruction results and the validation data is calculated, and parameter adjustments are made based on the error feedback. The mean squared error (MSE) between the reconstruction results and the validation data is calculated, the error distribution is analyzed, and areas with large errors are identified. This error information is fed back to the optimization algorithm to adjust the model parameters. The learning rate is dynamically adjusted using an adaptive optimization algorithm (such as Adam or RMSprop). Hyperparameter tuning is also performed using Bayesian optimization or genetic algorithms. ② Feedback mechanisms and real-time optimization include data acquisition strategy optimization, a real-time feedback loop, and online learning and model updates. Based on the current reconstruction results and error distribution, the data acquisition strategy is adjusted in real time to increase the sampling density in key areas. The error distribution in the current reconstruction results is analyzed to identify high-error areas and optimize the sampling strategy. For example, the sampling density in high-error areas is increased. The acquisition parameters of the polarization and light field cameras are adjusted in real time to improve data quality in these high-error areas. A feedback-based optimization loop is established to dynamically adjust data acquisition and model parameters. Acquisition and reconstruction parameters are initialized, initial data acquisition and 3D reconstruction are performed, and the error between the reconstruction result and the validation data is calculated. A feedback loop is then entered to adjust data acquisition and reconstruction parameters based on error feedback. New data acquisition and reconstruction are performed, and these steps are repeated until the error converges to a small value. An online learning algorithm (such as Online Gradient Descent) is used to update the model in real time to improve reconstruction accuracy. Real-time online learning is performed using the currently acquired data to update model parameters. The online learning rate is adjusted based on error feedback to adapt to data changes. Continuous iteration of online learning and acquisition strategy adjustments ensures that the model maintains high accuracy and robustness. (③) Multi-objective optimization involves trade-offs between multiple objectives and fuzzy control processes. Multi-objective optimization algorithms (such as Pareto frontier analysis) are used to find the optimal balance between reconstruction accuracy and computational resources. Multiple objective functions (such as reconstruction error and computational resource consumption) are defined, and Pareto frontier analysis is used to find the optimal balance between the two. The reconstruction algorithm is then optimized to ensure a balance between high accuracy and low resource consumption. Fuzzy control theory is introduced to address the uncertainty and nonlinearity of the system. Fuzzy rules and membership functions are defined to describe the input-output relationship of the system. A fuzzy inference system is applied for parameter adjustment and optimization. Within the fuzzy control system, acquisition and reconstruction parameters are dynamically adjusted to ensure robustness and adaptability. Through these steps, a system for adaptive parameter adjustment and real-time optimization is established.The system not only automatically adjusts parameters based on the current reconstruction results but also dynamically adjusts data acquisition strategies for key areas, significantly improving the reconstruction results and system robustness. The introduction of a feedback mechanism ensures the system's ability to self-correct and optimize, making the reconstruction of the combustion process more accurate and reliable. This approach is suitable for dynamic process analysis in complex scenarios.

[0118] The advantages of the present invention are: (1) It can perform high-precision three-dimensional reconstruction and multi-parameter synchronous measurement: by combining polarization spectrum imaging and light field camera technology, high-precision three-dimensional reconstruction of the combustion process is achieved, providing more detailed and accurate information about the combustion area than traditional methods. At the same time, a variety of photothermal parameters such as the temperature field, concentration distribution, scattering coefficient and absorption coefficient of the combustion area are obtained, providing rich data support for a comprehensive understanding of the combustion process. (2) By adopting optical imaging methods, non-contact and non-invasive measurement of the combustion process is achieved, avoiding the interference of traditional measurement methods on the combustion process. By using multi-scale fusion technology, macro- and micro-scale combustion phenomena can be analyzed simultaneously, providing a comprehensive understanding of the combustion process. Through adaptive parameter adjustment and feedback mechanism, the system can automatically optimize the reconstruction algorithm and data acquisition strategy according to real-time data, improving the robustness and adaptability of the method. (3) By using light field camera technology, high-spatial-resolution three-dimensional reconstruction is achieved, which can accurately locate and analyze the fine structure of the combustion area. Combined with multi-angle polarization spectrum data, more comprehensive information on combustion products is provided, which helps to accurately identify and quantify combustion products. Through optimization algorithms and parallel computing, this method can achieve near-real-time data processing and parameter reconstruction, and is suitable for real-time monitoring and control applications. Compared to traditional multi-sensor combinations, this method integrates multiple measurement functions, reducing overall system complexity and cost. It provides new observational tools and data sources for combustion science research, helping to deepen our understanding of complex combustion phenomena.

[0119] Example 1

[0120] like Figure 4As shown, this embodiment uses a polarization light field camera 11 that integrates polarization spectroscopy and light field imaging functions to achieve simultaneous acquisition of polarization spectroscopy and light field data in a single device, reducing the number of devices and synchronization errors. The system described in this embodiment comprises a camera body and heat dissipation system, constructed from high-strength, low thermal expansion coefficient materials (such as carbon fiber composites) to ensure structural stability and thermal performance. Built-in high-efficiency heat dissipation devices, such as heat pipes and micro fans, maintain a constant temperature environment within the camera. The polarization light field camera 11 consists of a main lens 112, a microlens array 113, a polarization filter array 114, and an adjustable spectral filter 115. A sensor system 116 is used for noise processing and consists of a high-resolution image sensor 117 and a sensor cooling system 118. The main lens 112 is a high-quality wide-angle lens with a large aperture and excellent color reproduction. The microlens array 113 is located between the main lens and the high-resolution image sensor 117 and is used for light field imaging. The polarization filter array 114 is integrated into the microlens array or closely attached to the sensor to acquire polarization information. The tunable spectral filter 115 uses a liquid crystal tunable filter or a linear variable filter to achieve spectrally selective imaging. Calibration is performed using a spectral calibration unit system 7, which consists of a built-in spectral calibration source and an automatic white balance system. The built-in spectral calibration source is used for real-time spectral calibration, and the automatic white balance system ensures color accuracy under various lighting conditions. The high-resolution image sensor uses a large-size, high-pixel density CMOS or CCD sensor. The sensor cooling system uses thermoelectric cooling technology to reduce sensor noise. The data processing system 6 consists of a high-performance processor, a large-capacity cache, and an FPGA. The high-performance processor is used for real-time data processing and image reconstruction, the large-capacity cache supports fast data read and write, and the FPGA is used for parallel processing of large amounts of light field and polarization data. The interface and storage system consists of a high-speed data interface and a built-in solid-state drive. High-speed data interfaces such as USB 3.1, Thunderbolt, or 10GbE are used for data transmission, and the built-in solid-state drive is used for high-speed data caching and storage. A synchronization system is also included, consisting of a high-precision clock and an external trigger interface. The high-precision clock is used to precisely control exposure time and synchronize multiple devices, and the external trigger interface is used for synchronization with other devices. The camera also includes a mechanical system consisting of a precision focusing mechanism and an anti-shake device. The precision focusing mechanism is an electric or manual fine-adjustment focusing system, and the anti-shake device is an optical or mechanical anti-shake system to improve image stability. Power is supplied by a power system consisting of a high-capacity lithium battery and an external power interface. The lithium battery supports long-term operation, and the external power supply is used for continuous power supply or battery charging.

[0121] A polarization light field camera is used to image the combustion process from multiple angles, acquiring polarization spectrum and light field data. Preprocessing steps remain the same, including denoising and calibration of the polarization light field data. By integrating the polarization spectrum and light field data, the size and shape characteristics of the particles are extracted and the three-dimensional structure is reconstructed. The three-dimensional temperature field distribution is calculated based on the reconstructed three-dimensional structure and polarization spectrum information, along with the concentration distribution and scattering and absorption coefficients. The incorporation of multi-scale information improves the accuracy and completeness of the reconstruction results and minimizes errors. Parameters are automatically adjusted based on the reconstruction results and verification data, optimizing the data acquisition strategy.

[0122] Example 2

[0123] In this embodiment, using two light field cameras 2 to collect light field data of the combustion area from different angles is an effective method for improving the accuracy of three-dimensional reconstruction. Using two light field cameras to collect light field data of the combustion area from two different angles, respectively, improves the resolution and accuracy of three-dimensional reconstruction. The two light field cameras work in conjunction with the polarization spectrum camera 1 to simultaneously collect data, enabling multi-angle light field imaging and polarization spectrum data acquisition. Data preprocessing such as denoising and calibration is performed on the two sets of light field data and polarization spectrum data. Data analysis and reconstruction are performed on the data, and dual light field data is used to perform higher-precision three-dimensional structure reconstruction, combining the polarization spectrum data to optimize the results. The three-dimensional temperature field distribution and concentration distribution, as well as photothermal parameters such as scattering and absorption coefficients, are calculated. Multi-scale fusion of data from different sources improves reconstruction accuracy. Adaptive parameter adjustment is performed to dynamically adjust data acquisition and reconstruction parameters.

[0124] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely illustrative of the principles and applications of the invention. It should be understood that many modifications may be made to the illustrative embodiments, and that other arrangements may be devised, without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in ways other than those described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be used in conjunction with other described embodiments.

Claims

1. A method for reconstructing three-dimensional photothermal parameters of a combustion process, characterized in that: include: Step 1: Use a polarization spectrum camera and a light field camera to synchronously collect data from multiple angles of the burning area to obtain polarization spectrum data and light field data of the burning area; Step 2: Preprocessing the polarization spectrum data and the light field data respectively to obtain preprocessed polarization spectrum data and preprocessed light field data; Step 3: Extracting light field depth information from the light field data, using the light field depth information in combination with a light field reconstruction algorithm to construct a three-dimensional structure of the combustion area, bringing the pre-processed polarization spectrum data into the three-dimensional structure of the combustion area, and optimizing the three-dimensional structure of the combustion area; obtaining the optimized three-dimensional structure; Step 4: Spectral analysis is performed on the pre-processed polarization spectrum data to obtain polarization degree and polarization angle information of the polarization spectrum; the size and shape characteristics of the solid particles in the combustion process are extracted using the polarization degree and polarization angle information; The optimized three-dimensional structure is used in combination with a CT reconstruction algorithm to obtain a three-dimensional distribution field of particulate matter during combustion; Based on the pre-processed polarization spectrum data and the pre-processed light field data, the optimized three-dimensional structure is used to calculate the three-dimensional temperature field and concentration field distribution of the combustion area gas; In step 3, the method of constructing the three-dimensional structure of the burning area by using the light field depth information combined with the light field reconstruction algorithm is as follows: Step 31: Extract the 4D light field function from the light field data, and use the local plane approximation algorithm of the structure tensor to calculate the depth information of each point in the combustion area; use the depth information to obtain a depth map of the combustion area; Step 32: Applying a shift-add algorithm in combination with Gaussian weighting to synthesize a fully focused image based on the depth map; Step 3: Generate an initial point cloud of the burning area using the depth map and the fully focused image, and use the Poisson surface reconstruction algorithm to generate a continuous surface from the initial point cloud to obtain the three-dimensional structure of the burning area; In step 3, the specific method for obtaining the optimized three-dimensional structure is: Step 3 and 4: extract feature points from the fully focused image and the pre-processed polarization spectrum data respectively, and use the geometric consistency method to match the feature points of the fully focused image and the pre-processed polarization spectrum data; Step 35: Construct a Markov random field model, use the matched polarization spectrum data as the depth prior, refine the depth map, use the barycentric coordinate interpolation method for texture mapping, and map the refined depth map to the continuous surface; obtain the optimized three-dimensional structure.

2. A method for reconstructing three-dimensional photothermal parameters of a combustion process according to claim 1, characterized in that: In step 2, the method for preprocessing the polarization spectrum data and the light field data is: The polarization spectrum data is sequentially subjected to dark current correction, flat field correction, spectrum calibration, polarization calibration, denoising and background interference elimination to obtain pre-processed polarization spectrum data; The light field data is subjected to denoising and microlens array calibration in sequence to obtain preprocessed light field data.

3. The method for reconstructing three-dimensional photothermal parameters of a combustion process according to claim 1, characterized in that: In step 4, the specific method for obtaining the polarization degree and polarization angle information of the polarization spectrum is: Step 41: Extract the measured intensity data and Stokes vector from the polarization spectrum data, and obtain the Mueller matrix through least squares optimization; Step 42: Decompose the Mueller matrix using a least squares method-based Mueller matrix decomposition algorithm to obtain polarization degree information of the polarization spectrum, and use the polarization degree information to obtain the polarization angle of the polarization spectrum.

4. The method for reconstructing three-dimensional photothermal parameters of a combustion process according to claim 3, characterized in that: In step 4, the method for extracting the size and shape characteristics of solid particles during combustion using polarization degree and polarization angle information is as follows: Step 43: Construct a forward model based on Mie scattering theory, and use the T-matrix method to expand the forward model to obtain an expanded model that can identify particles of different shapes and sizes; Step 4: Utilize the extended model and polarization degree and polarization angle information, and adopt the Levenberg-Marquardt algorithm to analyze the pre-processed polarization spectrum data to identify the size and shape characteristics of solid particles in the combustion process.

5. The method for reconstructing three-dimensional photothermal parameters of a combustion process according to claim 4, characterized in that: In step 4, the specific method for calculating the three-dimensional temperature field distribution of the gas in the combustion area is: Step M1: using the pre-processed polarization spectrum information, the Fresnel equation and the Kramers-Kronig relationship are used to estimate the emissivity of the gas combustion products; and calculating the complex refractive index using the estimated emissivity of the gas combustion products; Step M2: Calculate the projection matrix A using the complex refractive index, and establish a linear equation using the projection matrix A, where is the projection matrix, is the temperature field, and is the observed value; solve the linear equation using algebraic reconstruction technology to obtain the three-dimensional temperature field distribution.

6. The method for reconstructing three-dimensional photothermal parameters of a combustion process according to claim 5, characterized in that: In step 4, the specific method for calculating the concentration field distribution of the gas in the combustion area is: Step N1: Establish a temperature radiation transfer equation, use polarization spectrum data, and solve the radiation transfer equation using the discrete coordinate method to obtain the radiation intensity distribution of each point in the combustion area; Step N2: applying the Tikhonov regularization method to solve the inverse problem of the radiation intensity distribution at each point in the combustion area, and deriving the initial concentration distribution of the gas in the combustion area; Step N3: Use a non-negative matrix decomposition algorithm to separate and analyze the initial concentration distribution of the gas in the combustion area to obtain the final concentration field distribution of the gas in the combustion area.

7. A three-dimensional photothermal parameter reconstruction system for a combustion process, implemented based on the method of any one of claims 1 to 6, characterized in that: It includes a polarization spectrum camera (1), a light field camera (2), a synchronization control system (3), a rotating stage (4) and a computer (5); The polarization spectrum camera (1) is used to collect polarization spectrum data of the combustion area at multiple angles and send it to the computer (5); the light field camera (2) is used to collect spectrum data of the combustion area at multiple angles and send it to the computer (5); The rotating stage is used to control the polarization spectrum camera (1) and the light field camera (2) to rotate along the combustion area; The synchronous control system (3) synchronously controls the image acquisition moments of the polarization spectrum camera and the light field camera; The computer (5) uses the polarization spectrum data and the light field data to construct the three-dimensional structure of the combustion area and optimizes it. The optimized three-dimensional structure is combined with the CT reconstruction algorithm to obtain the three-dimensional distribution field of the particles during the combustion process and calculate the three-dimensional temperature field and concentration field distribution of the gas in the combustion area.

8. The combustion process three-dimensional photothermal parameter reconstruction system according to claim 7, characterized in that: The polarization spectrum camera includes a lens I (101), a polarization optical element (102), a tunable filter (103), a spectrum splitting system (104) and a photoelectric detection system I (105); The lens I (101) is aimed at the combustion area and is used to capture an image of the combustion area. The image of the combustion area is filtered by the polarization optical element (102) and transmitted to the tunable filter (103). The wavelength is selected and filtered by the tunable filter (103). The filtered light is incident on the spectral spectrometry system (104). The spectral spectrometry system (104) separates the spectrum of the filtered light according to different wavelength bands. The separated spectrum is converted into an electrical signal by the photoelectric detection system I (105), and the electrical signal is transmitted to the computer (5).

Images