Three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging

Abstract

The invention discloses a three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging, and belongs to the technical field of flame combustion detection. The invention aims to solve the problem of low reconstruction precision caused by limited spontaneous radiation information and poor low-temperature region signal sensitivity in the existing combustion flow field temperature field reconstruction. Comprising the following steps: obtaining a flame two-dimensional infrared temperature distribution image and a flame hyperspectral light field image; a two-dimensional blackbody radiation intensity curve and a three-dimensional blackbody radiation intensity curve are obtained through calibration according to the blackbody radiation law; combining the corresponding image gray values to obtain two-dimensional infrared radiation intensity, three-dimensional visible radiation intensity and three-dimensional infrared radiation intensity; determining a temperature threshold value of the high-temperature area; constructing a multispectral radiation source term matrix equation, and reconstructing a three-dimensional temperature field; obtaining a separated flame hyperspectral light field image, and reconstructing a low-temperature region and a reconstructed temperature field; and combining with a high-temperature region reconstruction temperature field to obtain a whole temperature field. The method is used for three-dimensional combustion multi-parameter field inversion.

Description

Technical Field

[0001] This invention relates to a three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging, belonging to the field of flame combustion detection technology. Background Technology

[0002] Optical diagnostic methods based on emission spectra typically employ charge-coupled devices (CCDs) as detection equipment and include various measurement methods such as single-camera imaging, multi-angle imaging, multi-camera imaging, and fiber optic imaging. Single-camera imaging systems can achieve three-dimensional measurements of temperature, soot component concentration, and nanofluid concentration distribution, but their applicability is limited to steady-state combustion flow fields with axisymmetric structures, failing to fully characterize transient three-dimensional features. Multi-angle imaging systems reduce the cost of detection equipment by acquiring more projection information, but due to reliance on mechanical movement, the temporal resolution is significantly reduced, still making it difficult to meet the research needs of actual three-dimensional turbulent combustion behavior. Multi-camera measurement systems have advantages in resolution and coverage and have been successfully applied to the three-dimensional temperature field measurement of large coal-fired boilers; however, their system complexity, calibration difficulties, and the need for numerous optical channels limit their application in small-scale combustion chambers or engines. Fiber optic imaging technology improves temporal resolution by sacrificing spatial resolution, offering unique advantages in measuring combustion flow field parameters within confined spaces. In addition, multispectral or hyperspectral imaging technology can simultaneously detect emission spectra in multiple wavelength ranges, making up for the shortcomings of single-wavelength diagnosis, and can simultaneously obtain information such as the concentration of multiple components and flame temperature.

[0003] In combustion flow field measurement, spectral imaging technology typically employs two detection methods: single-camera multispectral imaging and multi-camera multispectral imaging. Single-camera imaging systems are simple in structure and highly flexible, suitable for studying two-dimensional or three-dimensional axisymmetric combustion flow fields; while multi-camera imaging systems are better suited for in-depth analysis of the complex three-dimensional spatial structure of combustion flow fields, but their cost and complexity also increase accordingly. In recent years, light field imaging technology has developed rapidly due to its flexibility, portability, and three-dimensional reconstruction capabilities. Unlike traditional cameras, light field cameras, by placing a microlens array (MLA) between the main lens and the light sensor, can capture the entire field of radiation information in a single exposure, including radiation intensity, position, and angle, thereby achieving estimation of the three-dimensional structure and depth information of the target object. Light field imaging technology has been widely used in target classification and recognition, computer vision, biomedicine, and other fields, and has shown great potential in combustion flow field diagnosis. Its high spatiotemporal resolution three-dimensional imaging capability is expected to overcome the limitations of traditional cameras in terms of unsteady, non-axisymmetric photothermal characteristics, as well as multi-camera information acquisition and collaboration. Currently, light field imaging technology has been successfully applied to the visualization research of velocity field, temperature field, soot concentration and three-dimensional structure in scenarios such as the internal flow field of compressor blades, supersonic wind tunnel, and laminar flame.

[0004] In the reconstruction of the combustion flow field temperature field based on emission spectrum, the measurement signal sensitivity in the low temperature region is poor due to the limited spontaneous emission information measured, resulting in low reconstruction accuracy. Summary of the Invention

[0005] To address the problem of low reconstruction accuracy in existing combustion flow field temperature field reconstructions due to limited spontaneous emission information and poor signal sensitivity in low-temperature regions, this invention provides a three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging.

[0006] The present invention provides a three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging, comprising:

[0007] A two-dimensional infrared camera was used to capture the combustion flow field of the target flame region, resulting in a two-dimensional infrared temperature distribution image of the flame; simultaneously, a hyperspectral light field camera was used to capture the combustion flow field of the target flame region, resulting in a hyperspectral light field image of the flame.

[0008] Blackbody radiation calibration was performed on two-dimensional infrared temperature distribution images and hyperspectral light field images of flames using the blackbody radiation law, resulting in two-dimensional and three-dimensional blackbody radiation intensity curves. Two-dimensional spectral gray values ​​were extracted from the two-dimensional infrared temperature distribution images of flames and combined with the two-dimensional blackbody radiation intensity curves to obtain the two-dimensional infrared radiation intensity. Three-dimensional spectral gray values ​​were extracted from the hyperspectral light field images of flames and combined with the three-dimensional blackbody radiation intensity curves to obtain the three-dimensional visible radiation intensity and three-dimensional infrared radiation intensity.

[0009] The temperature threshold of the high-temperature region is determined based on the two-dimensional infrared radiation intensity, and the two-dimensional infrared temperature distribution image is segmented to obtain the two-dimensional high-temperature region.

[0010] The hyperspectral light field image of the flame is discretized to establish a spatial grid model. Then, the multispectral radiation source term matrix equation is constructed by combining the three-dimensional visible radiation intensity and the three-dimensional infrared radiation intensity. The three-dimensional temperature field is reconstructed by combining the topological constraint prior information provided by the two-dimensional high-temperature region, and the initial solution of the three-dimensional temperature field is calculated. The temperature field of the high-temperature region is reconstructed by combining the temperature threshold of the high-temperature region and using adaptive threshold segmentation iterative regularization to extract the temperature field of the high-temperature region from the initial solution of the three-dimensional temperature field.

[0011] Based on the reconstructed temperature field of the high-temperature region, the radiation contribution generated by the high-temperature region is separated from the hyperspectral light field image of the flame to obtain the separated hyperspectral light field image of the flame; then, based on the separated hyperspectral light field image of the flame, the multispectral radiation source term matrix equation of the low-temperature region is established, and the temperature field of the low-temperature region is reconstructed; the temperature field of the low-temperature region is optimized by an iterative optimization algorithm to obtain the reconstructed temperature field of the low-temperature region; combined with the reconstructed temperature field of the high-temperature region, the entire temperature field is obtained.

[0012] The three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging according to the present invention further includes:

[0013] The absorption coefficients of soot particles and the gas absorption coefficients are solved by using the multispectral radiation source matrix equation obtained from the hyperspectral light field image of the flame and the characteristic absorption peak wavelength of the gas.

[0014] Based on Mie scattering theory, the concentration of black smoke particles is inverted according to the absorption coefficient of black smoke particles.

[0015] Using a statistical narrowband model, the gas concentration is derived from the gas absorption coefficient.

[0016] According to the three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging of the present invention, the hyperspectral light field camera includes an integrated visible and infrared light field camera and a wavelength control element. The wavelength control element is disposed in front of the lens of the integrated visible and infrared light field camera and is used to control the light wavelength of the flame.

[0017] According to the three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging of the present invention, the wavelength control element is a liquid crystal wavelength modulator or a grating.

[0018] According to the three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging of the present invention, the hyperspectral light field camera includes a visible hyperspectral light field camera and an infrared hyperspectral light field camera that are set independently. A liquid crystal wavelength modulator or a grating is respectively set in front of the lens of the visible hyperspectral light field camera and the infrared hyperspectral light field camera.

[0019] According to the three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging of the present invention, the visible hyperspectral light field camera is a visible hyperspectral light field camera array formed by multiple cameras, and the infrared hyperspectral light field camera is an infrared hyperspectral light field camera array formed by multiple cameras; respectively used to acquire corresponding flame images from different angles;

[0020] The two-dimensional infrared camera is a two-dimensional infrared camera array formed by multiple cameras, used to acquire two-dimensional images of flames from different angles.

[0021] According to the three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging of the present invention, the calculation method of the initial solution of the three-dimensional temperature field is as follows:

[0022] By minimizing the objective function Calculate the radiation source term of a three-dimensional flame and the initial solution of the three-dimensional temperature field :

[0023] ,

[0024] In the formula The source term distribution in the low-temperature region of the hyperspectral light field image. For the objective function The regularization parameter, is the projection matrix of the hyperspectral light field image of the flame. For emission spectrum observation models, It is an L1 norm;

[0025] The method for obtaining the high-temperature region in the hyperspectral light field image is as follows:

[0026] The two-dimensional infrared temperature distribution image is segmented and the pixels of the two-dimensional high-temperature region are extracted based on the temperature threshold of the high-temperature region. These pixels are then mapped onto the hyperspectral light field image of the flame to obtain the high-temperature region of the hyperspectral light field image and the low-temperature region of the hyperspectral light field image.

[0027] The radiation source term of the three-dimensional flame was solved. Estimating the initial solution of the three-dimensional temperature field .

[0028] The three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging according to the present invention minimizes the objective function. Update the source term distribution in the low-temperature region of the hyperspectral light field image. The method is as follows:

[0029] ,

[0030] In the formula For the objective function The regularization parameter, For hyperparameters, This is the projection matrix of the low-temperature region in the hyperspectral light field image. This is the projection matrix of the high-temperature region in the hyperspectral light field image. The source term distribution in the high-temperature region of the hyperspectral light field image. This is the regularization matrix.

[0031] The three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging according to the present invention minimizes the objective function. get , and Value:

[0032] ,

[0033] In the formula These are measurements from the emission spectrum observation model. These are the observation values ​​from the emission spectrum observation model.

[0034] According to the three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging of the present invention, the calculation method of the absorption coefficient of soot particles and the gas absorption coefficient is as follows:

[0035] In the band without gas absorption peaks:

[0036] ,

[0037] In the formula is the monochromatic absorption coefficient of the flame. The absorption coefficient of black smoke particles;

[0038] ,

[0039] In the formula The first radiation constant, For wavelength, The second radiation constant, Let be the temperature of the nth spatial grid in the spatial grid model. Let be the radiation intensity of the measured signal received by the nth spatial grid in the spatial grid model;

[0040] Within the band containing gas absorption peaks:

[0041] ,

[0042] In the formula The gas absorption coefficient;

[0043] .

[0044] The beneficial effects of this invention are as follows: The method of this invention is a reconstruction method based on adaptive threshold segmentation and iterative regularization. By defining a temperature threshold, a two-step reconstruction process is performed: first, the temperature distribution of the high-temperature region is obtained using a two-dimensional infrared image; then, the influence of the high-temperature region on the measurement signal is separated, and a three-dimensional image acquired using a hyperspectral light field is used for reconstruction, thereby improving the sensitivity of the low-temperature region. This method effectively improves the accuracy of wide-temperature-range combustion flame temperature reconstruction, providing more reliable technical support for combustion flow field diagnosis. It also provides support for further solving for particle concentration and gas concentration. Attached Figure Description

[0045] Figure 1 This is a flowchart of the three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging as described in this invention;

[0046] Figure 2 This is a schematic diagram of the structure of the first type of image acquisition device in the method of the present invention;

[0047] Figure 3 This is a schematic diagram of the structure of the second type of image acquisition device in the method of the present invention;

[0048] Figure 4 This is a schematic diagram of the structure of the third type of image acquisition device in the method of the present invention. Detailed Implementation

[0049] 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.

[0050] Specific Implementation Method 1: Combination Figures 1 to 4 As shown, this invention provides a three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging, including:

[0051] A two-dimensional infrared camera was used to capture the combustion flow field of the target flame region, resulting in a two-dimensional infrared temperature distribution image of the flame; simultaneously, a hyperspectral light field camera was used to capture the combustion flow field of the target flame region, resulting in a hyperspectral light field image of the flame.

[0052] Blackbody radiation calibration was performed on two-dimensional infrared temperature distribution images and hyperspectral light field images of flames using the blackbody radiation law, resulting in two-dimensional and three-dimensional blackbody radiation intensity curves. Two-dimensional spectral gray values ​​were extracted from the two-dimensional infrared temperature distribution images of flames and combined with the two-dimensional blackbody radiation intensity curves to obtain the two-dimensional infrared radiation intensity. Three-dimensional spectral gray values ​​were extracted from the hyperspectral light field images of flames and combined with the three-dimensional blackbody radiation intensity curves to obtain the three-dimensional visible radiation intensity and three-dimensional infrared radiation intensity.

[0053] The temperature threshold of the high-temperature region is determined based on the two-dimensional infrared radiation intensity, and the two-dimensional infrared temperature distribution image is segmented to obtain the two-dimensional high-temperature region.

[0054] The hyperspectral light field image of the flame is discretized to establish a spatial grid model. Then, the multispectral radiation source term matrix equation is constructed by combining the three-dimensional visible radiation intensity and the three-dimensional infrared radiation intensity. The three-dimensional temperature field is reconstructed by combining the topological constraint prior information provided by the two-dimensional high-temperature region, and the initial solution of the three-dimensional temperature field is calculated. The temperature field of the high-temperature region is reconstructed by combining the temperature threshold of the high-temperature region and using adaptive threshold segmentation iterative regularization to extract the temperature field of the high-temperature region from the initial solution of the three-dimensional temperature field.

[0055] Based on the reconstructed temperature field of the high-temperature region, the radiation contribution generated by the high-temperature region is separated from the hyperspectral light field image of the flame to obtain the separated hyperspectral light field image of the flame; then, based on the separated hyperspectral light field image of the flame, the multispectral radiation source term matrix equation of the low-temperature region is established, and the temperature field of the low-temperature region is reconstructed; the temperature field of the low-temperature region is optimized by an iterative optimization algorithm to obtain the reconstructed temperature field of the low-temperature region; combined with the reconstructed temperature field of the high-temperature region, the entire temperature field is obtained.

[0056] After extracting and reconstructing the temperature field in the high-temperature region, the reconstructed temperature field in the high-temperature region is kept unchanged. Separating the influence of the high-temperature region on the measurement signal is equivalent to filtering out the results generated by the high temperature. Then, using the spectral and spatial features in the hyperspectral light field image, combined with the multispectral radiation source term matrix equation and mathematical methods such as compression and noise reduction, LSQR, the temperature distribution in the low-temperature region is reconstructed.

[0057] When using a 2D infrared camera to photograph the combustion flow field, it is necessary to ensure that the camera covers the entire flame area and adjust the focal length and exposure time to obtain the best image quality. A synchronous control module can be used to control a hyperspectral light field camera and a 2D infrared camera to simultaneously photograph the same flame area, recording the three-dimensional spatial and directional information of the flame. During the shooting process, it is ensured that the camera's spectral range covers the main radiation bands of the flame.

[0058] The blackbody radiation law is used to calibrate the visible and infrared light field measurement signals, obtaining the blackbody radiation intensity curve of the flame. The spectral grayscale values ​​of the flame field image are then extracted to obtain the flame's spectral radiation intensity. Subsequently, a spatial grid model is established by discretizing the three-dimensional space of the combustion field. A multispectral radiation source term matrix equation is constructed by combining the spectral radiation intensity data, achieving effective fusion of spectral and spatial information to obtain the visible radiation intensity at different wavelengths. and infrared radiation intensity .

[0059] The range of the high-temperature region is determined based on the temperature distribution in the two-dimensional infrared image, and a temperature threshold for the high-temperature region is defined. An adaptive thresholding segmentation algorithm is used to segment the two-dimensional infrared image, extracting pixels from the high-temperature region and mapping them to the corresponding regions in the hyperspectral light field image. In other words, using the temperature threshold determined by the two-dimensional infrared image, thresholding segmentation is performed on the high-temperature and low-temperature regions of the flame hyperspectral light field image, separating the observation model of the low-temperature region from the light field observation model.

[0060] An iterative regularization algorithm is used to process the matrix equation of the multispectral radiation source terms established using hyperspectral light field images, and the temperature distribution of the hyperspectral light field images is reconstructed in three dimensions to obtain an initial solution over a wide temperature range. Then, the high-temperature region is fixed, the influence of the high-temperature region on the measurement signal is separated, the sensitivity of the low-temperature region is improved, and the matrix equation of the multispectral radiation source terms in the low-temperature region is further solved to obtain a secondary update of the solution in the low-temperature region.

[0061] Temperature distribution reconstruction in the low-temperature region: The influence of the high-temperature region on the measurement signal is separated, and the reconstruction results from the high-temperature region are used to correct the hyperspectral light field image, removing the radiation contribution from the high-temperature region. Using the spectral and spatial features of the hyperspectral light field image, combined with physical models and mathematical methods, the temperature distribution in the low-temperature region is reconstructed. An iterative optimization algorithm is then employed to further optimize the temperature distribution in the low-temperature region, improving reconstruction accuracy and reliability, and combining this with the high-temperature region to form a three-dimensional temperature field.

[0062] Furthermore, this embodiment also includes:

[0063] The absorption coefficients of soot particles and the gas absorption coefficients are solved by using the multispectral radiation source matrix equation obtained from the hyperspectral light field image of the flame and the characteristic absorption peak wavelength of the gas.

[0064] Based on Mie scattering theory, the concentration of soot particles is inverted according to the absorption coefficient of soot particles; and the temperature field is used for correction to improve accuracy.

[0065] Using a statistical narrowband model, the gas concentration is derived from the gas absorption coefficient.

[0066] Combination Figure 2 As shown, the image acquisition device includes a hyperspectral light field camera 1, a two-dimensional infrared camera 2, a synchronization controller 3, and a data processing system 4;

[0067] The hyperspectral light field camera 1 is one of the core data acquisition devices of the system, used to capture three-dimensional visible and infrared radiation information of the combustion field. Its placement must ensure clear coverage of the entire flame area, with its height roughly level with the center of the combustion field, avoiding obstruction or interference from the surrounding environment. Appropriate visible and infrared filters must be selected according to measurement requirements to obtain spectral information in specific wavelength bands. Simultaneously, the exposure time and gain are adjusted based on the brightness of the combustion field and the camera's sensitivity to ensure image quality that is neither too dark nor too overexposed, clearly capturing combustion field details while minimizing noise. The main function of this camera is to record the three-dimensional spatial and directional information of the flame, as well as the radiation intensity at different wavelengths, providing rich spectral data support for subsequent reconstruction.

[0068] The two-dimensional infrared camera 2 is used to directly acquire temperature distribution information on the surface of the combustion field, especially temperature data in high-temperature areas. Its placement must be coordinated with the hyperspectral light field camera 1 to avoid mutual interference; typically, it can be placed at a 90° angle to avoid thermal interference from the surrounding environment. Before use, temperature calibration with a standard blackbody is required to ensure measurement accuracy. Based on the infrared radiation characteristics of the combustion field, camera parameters such as exposure time and gain are adjusted to ensure image quality. The main function of the two-dimensional infrared camera is to provide direct measurement data on temperature distribution in high-temperature areas and, through calibration, ensure the reliability of the measurement results, providing an important reference for subsequent reconstruction.

[0069] Synchronization Controller 3 is a key device ensuring the coordinated operation of all system components. It is responsible for coordinating the data acquisition process of the visible and infrared hyperspectral light field camera and the two-dimensional infrared camera. Its placement should be close to the cameras, avoiding proximity to sources of strong electromagnetic interference. Depending on the camera's operating mode and measurement requirements, either hardware or software synchronization should be selected, and the synchronization accuracy adjusted to keep the acquisition time error between the two cameras within acceptable limits. The main function of the synchronization controller is to ensure the synchronicity and consistency of data acquisition, avoiding data inconsistencies caused by time deviations, and providing high-quality raw data for subsequent data processing.

[0070] Data processing system 4 is the core analysis module of the system, responsible for receiving and processing light field information from visible and infrared hyperspectral light field camera 1 and two-dimensional infrared camera 2. By combining physical models such as the radiative transfer equation and gas absorption peak wavelength, it solves for the temperature, absorption coefficient, and gas absorption coefficient of the soot particles, thereby reconstructing the three-dimensional temperature field of the flame. Using Mie scattering theory and the statistical narrowband method, it inverts the soot particle concentration and gas concentration, and then fuses the three-dimensional temperature field, particle concentration field, and gas concentration field to generate a three-dimensional multi-parameter distribution map of the combustion field. Finally, visualization technology is used to display the spatial distribution of these parameters, providing comprehensive data support for the analysis and optimization of the combustion process. The main function of the data processing system is to comprehensively analyze and process multi-source data, achieving accurate inversion and visualization of multi-parameter combustion field data, providing strong technical support for combustion diagnosis and control.

[0071] As an example, combined Figure 2 As shown, the hyperspectral light field camera includes an integrated visible and infrared light field camera and a wavelength control element. The wavelength control element is disposed in front of the lens of the integrated visible and infrared light field camera and is used to control the light wavelength of the flame.

[0072] The wavelength modulation element is a liquid crystal wavelength modulator or a grating.

[0073] Figure 2This is a single-camera integrated hyperspectral light field camera structure. It employs an integrated hyperspectral light field camera 1, integrating visible and infrared hyperspectral imaging capabilities into a single camera. Multi-band data acquisition is achieved through beam splitter or filter switching. It includes the integrated visible and infrared light field cameras 1-1 and wavelength control elements 1-2. A two-dimensional infrared camera 2 is independently installed to supplement temperature distribution information in high-temperature regions. A synchronization controller connects to the integrated hyperspectral light field camera and the two-dimensional infrared camera to ensure synchronized data acquisition. The data processing system is integrated into a computer for receiving and processing data. This structure simplifies system layout and reduces hardware costs through integrated design, while enabling rapid acquisition of multi-band data through beam splitter or filter switching. It is suitable for laboratory or small-scale combustion experiments, especially in space-constrained and cost-sensitive scenarios.

[0074] As an example, combined Figure 3 As shown, the hyperspectral light field camera includes a visible hyperspectral light field camera and an infrared hyperspectral light field camera that are set independently. A liquid crystal wavelength modulator or a grating is respectively set in front of the lens of the visible hyperspectral light field camera and the infrared hyperspectral light field camera.

[0075] Figure 3 This is a dual-camera, separate hyperspectral light field camera structure. The visible hyperspectral light field camera 1 (1-100) and the infrared hyperspectral light field camera 1-200 are independently installed, specifically for capturing visible and infrared information of the combustion field. A two-dimensional infrared camera 2 is independently installed to supplement temperature distribution information in high-temperature areas. A synchronization controller 3 connects to the three cameras to ensure synchronized data acquisition. A data processing system 4 is integrated into a computer for receiving and processing data. This structure improves the flexibility and accuracy of data acquisition through functional separation, while its modular design facilitates maintenance and upgrades. It is suitable for industrial combustion monitoring or complex combustion experiments, especially for scenarios requiring high data accuracy and functional flexibility. In the diagram, 1-101 is the visible hyperspectral light field camera body, and 1-102 is a liquid crystal wavelength modulator or grating; 1-201 is the infrared hyperspectral light field camera body, and 1-202 is a liquid crystal wavelength modulator or grating.

[0076] As an example, combined Figure 4 As shown, the visible hyperspectral light field camera is a visible hyperspectral light field camera array formed by multiple cameras, and the infrared hyperspectral light field camera is an infrared hyperspectral light field camera array formed by multiple cameras; they are used to acquire corresponding flame images from different angles respectively.

[0077] The two-dimensional infrared camera is a two-dimensional infrared camera array formed by multiple cameras, used to acquire two-dimensional images of flames from different angles.

[0078] Figure 4The system employs a multi-camera array structure. The visible hyperspectral light field camera array comprises two arrays: one consisting of visible hyperspectral light field camera array 1-100 and the other of infrared hyperspectral light field camera array 1-200, each consisting of multiple cameras that capture visible and infrared information of the combustion field from different angles. A two-dimensional infrared camera 2, composed of multiple two-dimensional infrared cameras, supplements the temperature distribution information of the high-temperature region from different angles. A synchronization controller 3 connects to all cameras to ensure synchronized data acquisition. A data processing system 4 is integrated into the computer for receiving and processing data. This structure improves data coverage and reconstruction accuracy through multi-angle acquisition, while the collaborative work of multiple cameras allows for the acquisition of higher-resolution three-dimensional multi-parameter distribution maps of the combustion field. It is suitable for diagnostics of large combustion equipment or aero-engine combustion chambers, and is particularly well-suited for scenarios with extremely high requirements for data coverage and accuracy. In the diagram, 1-101 represents the visible hyperspectral light field camera body, and 1-102 represents a liquid crystal wavelength modulator or grating; 1-201 represents the infrared hyperspectral light field camera body, and 1-202 represents a liquid crystal wavelength modulator or grating.

[0079] The hyperspectral light field camera 1 uses methods such as liquid crystal wavelength modulators to adjust the wavelength.

[0080] By utilizing multi-wavelength images acquired through hyperspectral optical field technology and employing methods such as nonnegative least squares, the three-dimensional absorption coefficient field, particle concentration field, and temperature field of particles can be reconstructed simultaneously. To further improve analytical accuracy and efficiency, this implementation proposes an instantaneous collaborative reconstruction model and algorithm based on two-dimensional infrared and three-dimensional hyperspectral optical fields. The projection matrix superimposed from two-dimensional infrared acquisitions, combined with a high-low temperature segmentation iterative algorithm, provides prior information for the three-dimensional hyperspectral optical field. This model collaboratively reconstructs and inverts the flame physical property parameter field, carbon black particle concentration field, and temperature field. In the collaborative reconstruction model, the inversion of the physical property parameter field, carbon black particle concentration field, and temperature field are interrelated. By combining hyperspectral optical field technology and two-dimensional infrared data, multi-dimensional information of the flame can be acquired simultaneously, and the collaborative reconstruction algorithm can be used to achieve accurate inversion of these parameter fields. For example, the reconstruction result of the carbon black particle concentration field can provide important constraints for the inversion of the temperature field, while the distribution of the temperature field, in turn, affects the calculation of the physical property parameter field. This collaborative reconstruction method not only improves the accuracy of the reconstruction results but also significantly enhances the model's adaptability to different combustion scenarios.

[0081] Using two-dimensional surface intensity correction light field camera data from an infrared camera for three-dimensional surface reconstruction provides global constraints, improves reconstruction accuracy, and compensates for the limitations of light field cameras. This method fully leverages the advantages of both types of cameras, providing more comprehensive and reliable information for flame detection and analysis.

[0082] Infrared cameras detect the total radiation intensity of a flame, which is the superposition of light rays from all directions on the sensor. This total intensity information reflects the global characteristics of the flame, such as the overall temperature distribution and intensity variations. Infrared cameras can provide the two-dimensional surface intensity distribution of the flame, making them suitable for analyzing its surface properties. Light field cameras record the directional intensity of light rays, enabling the reconstruction of the flame's three-dimensional structure. However, due to the complexity of the light field camera's detection mechanism, the reconstruction results may be affected by noise, calibration errors, or algorithm limitations, leading to insufficient reconstruction accuracy.

[0083] The total intensity information from the infrared camera can serve as a global constraint, helping to correct the reconstruction results of the light field camera. For example, if the intensity of a certain area reconstructed by the light field camera is inconsistent with the observation results of the infrared camera, correction can improve the reconstruction accuracy. Infrared camera data is a direct observation result and has high reliability. Introducing infrared camera data into the reconstruction process of the light field camera can improve the credibility of the reconstruction results. When reconstructing the three-dimensional structure of a flame, the light field camera may not accurately reflect the intensity distribution of certain areas. The two-dimensional surface intensity from the infrared camera can help fill these gaps, providing more complete flame information. By acquiring two-dimensional infrared images and hyperspectral light field images, the total intensity of the hyperspectral light field and the infrared camera is aligned and normalized to ensure that the data are compared at the same scale. A reconstruction method based on adaptive threshold segmentation and iterative regularization is employed. This method defines a temperature threshold and performs a two-step reconstruction process: first, the temperature distribution of the high-temperature region is obtained using the two-dimensional infrared image; then, the influence of the high-temperature region on the measurement signal is separated, and the three-dimensional image acquired by the hyperspectral light field is used for reconstruction, thereby improving the sensitivity of the low-temperature region. The corrected light field camera data is used for three-dimensional reconstruction to generate a more accurate flame model.

[0084] Combination Figures 2 to 4 As shown, the following are three specific implementations of a three-dimensional combustion multi-parameter field inversion system based on two-dimensional infrared imaging and hyperspectral light field imaging. Each implementation is adjusted in terms of structural design to adapt to different hardware configurations and functional requirements. The three implementations are respectively adjusted in the structural design of a single-camera integrated hyperspectral light field camera, a dual-camera separate hyperspectral light field camera, and a multi-camera array, to adapt to different hardware configurations and functional requirements. Figure 2 The single-camera integrated structure shown simplifies system design and is suitable for small-scale scenarios. Figure 3 The dual-camera separate structure shown improves the flexibility and accuracy of data acquisition, making it suitable for complex scenarios; Figure 4 The multi-camera array structure shown achieves high resolution and wide coverage through multi-angle acquisition, making it suitable for large equipment or scenarios requiring high precision. These three structural variations provide flexible and efficient solutions for different application scenarios.

[0085] In this embodiment, the two-dimensional infrared image provides direct temperature information of the high-temperature zone of the flame, while the hyperspectral light field image provides rich spatial, directional and spectral information, providing basic data for subsequent reconstruction of the temperature distribution of the low-temperature zone.

[0086] Hyperspectral light field camera acquires visible light field measurement signals Infrared light field measurement signal It can be used to solve for temperature fields, particle concentration fields, and gas concentration fields. In common alkane combustion, it can fuse visible light field measurement signals at multiple wavelengths. ... The temperature field and particle concentration field were obtained. Infrared light field measurement signal. The absorption of both gas and particles can be combined to separate and solve the gas absorption coefficient field, and the gas concentration field can be obtained.

[0087] The formula for determining the monochromatic radiation intensity of a flame using the blackbody radiation law is:

[0088] ,

[0089] In the formula, h is Planck's constant, c is the speed of light, λ is the wavelength, k is Boltzmann's constant, and T is the absolute temperature of the blackbody furnace. Extracting the visible light field measurement signal to obtain the blackbody radiation intensity. and infrared light field measurement signal The grayscale value is used to obtain the monochromatic radiation intensity of the flame. .

[0090] Calculate the directional radiation intensity emitted from the boundary of the radiation system as acquired by a two-dimensional infrared camera, and calculate the spectral radiation intensity of the blackbody according to Planck's law. If a multi-camera imaging system (composed of multiple CCD cameras) records the directional radiation intensity measurements of M probe rays in the radiation image... The blackbody radiation intensity distribution within the entire radiation system can be obtained by solving the linear equations, and the temperature distribution of the medium can then be calculated using Planck's law. A well-established LSQR algorithm has been proven to be an effective method for solving large linear sparse coefficient matrices, enabling the solution of the inverse problem of two-dimensional infrared camera arrays.

[0091] The combustion flow field is divided into: Each grid, combined with the radiation transfer mechanism of the combustion flow field and the light field convolution imaging model, shows the radiation intensity received on a single pixel. Represented as:

[0092] ,

[0093] In the formula , and These represent the radiation source term, sampling region, and point spread function on the detection line, respectively; This represents the optical thickness along the detection path. The matrix form of the light field convolution imaging model can be expressed as:

[0094] ,

[0095] In the formula This represents the vectorized form of measurement information. For the projection matrix, For the vectorization of the radiation source term, This refers to random noise added to the forward model.

[0096] Before image calibration, data preprocessing and feature extraction are required. Preprocessing of the acquired two-dimensional infrared images includes denoising, correction, and normalization to improve image quality and consistency. Preprocessing of the hyperspectral light field images includes spectral correction, geometric correction, and background subtraction to eliminate noise and interference. High-temperature region features, such as temperature thresholds, edge contours, and texture information, are extracted from the two-dimensional infrared images. Spectral and spatial features, such as radiant intensity, spectral curves, and three-dimensional coordinates, are extracted from the hyperspectral light field images.

[0097] When performing three-dimensional reconstruction of the temperature distribution in the high-temperature region of the hyperspectral light field image, the temperature of the carbon black particles is assumed to be the combustion field temperature. The wavelength corresponding to the matrix equation of this part of the multispectral radiation source term is the visible light region, and it is assumed that there is no gas absorption.

[0098] In the reconstruction method based on adaptive threshold segmentation and iterative regularization, a two-step reconstruction process is performed by defining a temperature threshold. The first step is to obtain an accurate temperature distribution in the high-temperature region, and the second step is to separate the influence of the high-temperature region on the measurement signal and improve the sensitivity in the low-temperature region. The detailed steps are as follows: For temperature tomography, the entire three-dimensional temperature field T is divided into high-temperature regions T0 and T1. H and low temperature zone Two different regions, high-temperature region and low temperature regions ,in This indicates the temperature threshold. Determining the temperature threshold... Previously, we first used the minimization objective function To estimate the radiation source term F and initial temperature distribution of a three-dimensional flame.

[0099] Furthermore, the method for calculating the initial solution of the three-dimensional temperature field is as follows:

[0100] By minimizing the objective function Calculate the radiation source term vector of a three-dimensional flame. and the initial solution of the three-dimensional temperature field :

[0101] ,

[0102] In the formula The source term distribution in the low-temperature region of the hyperspectral light field image. For the objective function The regularization parameter, is the projection matrix of the hyperspectral light field image of the flame. For emission spectrum observation models, It is an L1 norm;

[0103] By solving The initial temperature vector can be estimated. Threshold segmentation is achieved by using a forward transform. To remap The transformation sorts the data from largest to smallest. By truncation The temperature values ​​within the range decompose the matrix Γ into two parts: Related to high-temperature areas, Related to low temperature regions, and using threshold values This indicates the proportion of voxels in the high-temperature region relative to the total number of voxels:

[0104] .

[0105] The emission spectrum observation model of the confined space combustion flow field based on the threshold segmentation method can be rewritten as follows:

[0106] ;

[0107] In the formula, These are the projection matrices for the high-temperature region and the low-temperature region, respectively, derived from the original projection matrix through...

[0108] These are the projection matrices for the high-temperature region and the low-temperature region, respectively, derived from the original projection matrix through... Mapped to obtain; The source term distributions corresponding to the two temperature regions can be based on Calculation. If the first p eigenvectors and radiation source terms corresponding to the high-temperature region are truncated, the emission spectrum observation model of the low-temperature region of the combustion flow field can be expressed as:

[0109] ;

[0110] In the formula, And there are By fixing the initially estimated high-temperature region, an inverse problem optimization method is used to optimize the objective function. Minimize in order to update the distribution of radiation source terms in the low-temperature region of the wide-temperature combustion flow field again.

[0111] The method for obtaining the high-temperature region in the hyperspectral light field image is as follows:

[0112] The two-dimensional infrared temperature distribution image is segmented and the pixels of the two-dimensional high-temperature region are extracted based on the temperature threshold of the high-temperature region. These pixels are then mapped onto the hyperspectral light field image of the flame to obtain the high-temperature region of the hyperspectral light field image and the low-temperature region of the hyperspectral light field image.

[0113] The radiation source term of the three-dimensional flame was solved. Estimating the initial solution of the three-dimensional temperature field .

[0114] By minimizing the objective function Update the source term distribution in the low-temperature region of the hyperspectral light field image. The method is as follows:

[0115] ,

[0116] In the formula For the objective function The regularization parameter, For hyperparameters, This is the projection matrix of the low-temperature region in the hyperspectral light field image. This is the projection matrix of the high-temperature region in the hyperspectral light field image. The source term distribution in the high-temperature region of the hyperspectral light field image. This is the regularization matrix.

[0117] In the formula It is A regularization matrix, which is an identity matrix, is used to overcome the ill-conditioned nature of the inverse problem. After obtaining the updated radiation source term distribution in the low-temperature region, the radiation source terms and temperature field of the entire three-dimensional combustion flame field over the wide temperature range can be obtained through inverse transformation. The ATSIR method is used for thresholding and minimizing the objective function. and During the process, it is necessary to determine , and The value of .

[0118] The study minimizes the objective function associated with the actual measured signal and the estimated signal using a nonlinear optimization method. The method to obtain regularization parameters , and .

[0119] By minimizing the objective function get , and Value:

[0120] ,

[0121] In the formula These are measurements from the emission spectrum observation model. These are the observation values ​​from the emission spectrum observation model.

[0122] The calculation methods for the absorption coefficient of black smoke particles and the gas absorption coefficient are as follows:

[0123] Within the band without gas absorption peaks, the monochromatic absorption coefficient of the flame is equal to the absorption coefficient of the soot particles:

[0124] ,

[0125] In the formula is the monochromatic absorption coefficient of the flame. The absorption coefficient of black smoke particles;

[0126] ,

[0127] In the formula The first radiation constant, For wavelength, The second radiation constant, Let be the temperature of the nth spatial grid in the spatial grid model. Let be the radiation intensity of the measured signal received by the nth spatial grid in the spatial grid model;

[0128] Within the wavelength range containing gas absorption peaks, the monochromatic absorption coefficient of a flame is composed of both the absorption coefficient of soot particles and the absorption coefficient of the gas:

[0129] ,

[0130] In the formula The gas absorption coefficient;

[0131] .

[0132] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging, characterized in that... include: A two-dimensional infrared camera was used to capture the combustion flow field of the target flame region, resulting in a two-dimensional infrared temperature distribution image of the flame; simultaneously, a hyperspectral light field camera was used to capture the combustion flow field of the target flame region, resulting in a hyperspectral light field image of the flame. Blackbody radiation calibration was performed on two-dimensional infrared temperature distribution images and hyperspectral light field images of flames using the blackbody radiation law, resulting in two-dimensional and three-dimensional blackbody radiation intensity curves. Two-dimensional spectral gray values ​​were extracted from the two-dimensional infrared temperature distribution images of flames and combined with the two-dimensional blackbody radiation intensity curves to obtain the two-dimensional infrared radiation intensity. Three-dimensional spectral gray values ​​were extracted from the hyperspectral light field images of flames and combined with the three-dimensional blackbody radiation intensity curves to obtain the three-dimensional visible radiation intensity and three-dimensional infrared radiation intensity. The temperature threshold of the high-temperature region is determined based on the two-dimensional infrared radiation intensity, and the two-dimensional infrared temperature distribution image is segmented to obtain the two-dimensional high-temperature region. The hyperspectral light field image of the flame is discretized to establish a spatial grid model. Then, the multispectral radiation source term matrix equation is constructed by combining the three-dimensional visible radiation intensity and the three-dimensional infrared radiation intensity. Combined with the topological constraint prior information provided by the two-dimensional high-temperature region, the three-dimensional temperature field is reconstructed, and the initial solution of the three-dimensional temperature field is calculated. By combining the temperature threshold of the high-temperature region, the temperature field of the high-temperature region is reconstructed from the initial solution of the three-dimensional temperature field using adaptive threshold segmentation iterative regularization. Based on the temperature field reconstructed in the high-temperature region, the radiation contribution generated in the high-temperature region is separated from the hyperspectral light field image of the flame to obtain the separated hyperspectral light field image of the flame; then, based on the separated hyperspectral light field image of the flame, the multispectral radiation source term matrix equation of the low-temperature region is established, and the temperature field of the low-temperature region is reconstructed. An iterative optimization algorithm is used to optimize the temperature field in the low-temperature region to obtain the reconstructed temperature field in the low-temperature region; combined with the reconstructed temperature field in the high-temperature region, the entire temperature field is obtained.

2. The three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging according to claim 1, characterized in that... Also includes: The absorption coefficients of soot particles and the gas absorption coefficients are solved by using the multispectral radiation source matrix equation obtained from the hyperspectral light field image of the flame and the characteristic absorption peak wavelength of the gas. Based on Mie scattering theory, the concentration of black smoke particles is inverted according to the absorption coefficient of black smoke particles. Using a statistical narrowband model, the gas concentration is derived from the gas absorption coefficient.

3. The three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging according to claim 1, characterized in that, The hyperspectral light field camera includes an integrated visible and infrared light field camera and a wavelength control element. The wavelength control element is positioned in front of the lens of the integrated visible and infrared light field camera and is used to control the light wavelength of the flame.

4. The three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging according to claim 3, characterized in that, The wavelength modulation element is a liquid crystal wavelength modulator or a grating.

5. The three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging according to claim 1, characterized in that, The hyperspectral light field camera includes a visible hyperspectral light field camera and an infrared hyperspectral light field camera that are set independently. A liquid crystal wavelength modulator or a grating is respectively set in front of the lens of the visible hyperspectral light field camera and the infrared hyperspectral light field camera.

6. The three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging according to claim 5, characterized in that, The visible hyperspectral light field camera is a visible hyperspectral light field camera array formed by multiple cameras, and the infrared hyperspectral light field camera is an infrared hyperspectral light field camera array formed by multiple cameras; they are used to acquire corresponding flame images from different angles. The two-dimensional infrared camera is a two-dimensional infrared camera array formed by multiple cameras, used to acquire two-dimensional images of flames from different angles.

7. The three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging according to claim 2, characterized in that, The method for calculating the initial solution of the three-dimensional temperature field is as follows: By minimizing the objective function Calculate the radiation source term of a three-dimensional flame and the initial solution of the three-dimensional temperature field : ， In the formula The source term distribution in the low-temperature region of the hyperspectral light field image. For the objective function The regularization parameter, is the projection matrix of the hyperspectral light field image of the flame. For emission spectrum observation models, It is an L1 norm; The method for obtaining the high-temperature region in the hyperspectral light field image is as follows: The two-dimensional infrared temperature distribution image is segmented and the pixels of the two-dimensional high-temperature region are extracted based on the temperature threshold of the high-temperature region. These pixels are then mapped onto the hyperspectral light field image of the flame to obtain the high-temperature region of the hyperspectral light field image and the low-temperature region of the hyperspectral light field image. The radiation source term of the three-dimensional flame was solved. Estimating the initial solution of the three-dimensional temperature field .

8. The three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging according to claim 7, characterized in that, By minimizing the objective function Update the source term distribution in the low-temperature region of the hyperspectral light field image. The method is as follows: ， In the formula For the objective function The regularization parameter, For hyperparameters, This is the projection matrix of the low-temperature region in the hyperspectral light field image. This is the projection matrix of the high-temperature region in the hyperspectral light field image. The source term distribution in the high-temperature region of the hyperspectral light field image. This is the regularization matrix.

9. The three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging according to claim 8, characterized in that, By minimizing the objective function get , and Value: ， In the formula These are measurements from the emission spectrum observation model. These are the observation values ​​from the emission spectrum observation model.

10. The three-dimensional combustion multi-parameter field inversion method based on two-dimensional infrared imaging and hyperspectral light field imaging according to claim 9, characterized in that, The calculation methods for the absorption coefficient of black smoke particles and the gas absorption coefficient are as follows: In the band without gas absorption peaks: ， In the formula is the monochromatic absorption coefficient of the flame. The absorption coefficient of black smoke particles; ， In the formula The first radiation constant, For wavelength, The second radiation constant, Let be the temperature of the nth spatial grid in the spatial grid model. Let be the radiation intensity of the measured signal received by the nth spatial grid in the spatial grid model; Within the band containing gas absorption peaks: ， In the formula The gas absorption coefficient; 。

Images