Infrared thermal imaging-based monitoring system for reactant gas temperature and concentration distribution

Abstract

This invention relates to the field of infrared imaging detection and industrial process monitoring, specifically a reactive gas temperature and concentration distribution monitoring system based on infrared thermal imaging analysis. It includes modules for image data acquisition, dynamic flow field analysis, radiation coupling deconstruction, dual-field independent reconstruction, and adaptive feedback optimization. The system extracts dynamic gas regions based on optical flow vector fields, performs nonlinear radiation separation, and inverts the temperature and concentration fields. Its core is to utilize the mutual information characteristics of temperature and concentration to generate feedback correction coefficients, iteratively updating calculation parameters until the mutual information converges to a minimum. This invention achieves feature decoupling at the mathematical and statistical level, overcoming the bottleneck of a single detector's inability to distinguish between high-temperature and high-concentration radiation characteristics, and significantly improving inversion accuracy.

Description

Technical Field

[0001] This invention relates to the field of infrared imaging detection and industrial process monitoring, specifically a system for monitoring the temperature and concentration distribution of reaction gases based on infrared thermal imaging analysis. Background Technology

[0002] In industrial reaction monitoring environments, the gas flow field inside reaction vessels exhibits complex dynamic changes, and the temperature and concentration distributions have highly nonlinear coupling characteristics. Existing monitoring schemes, when using infrared thermal imaging technology for non-contact detection, typically treat the observed pixel intensity as a reflection of a single physical quantity, or employ fixed linear compensation algorithms.

[0003] Because the radiation energy observed in infrared images includes the gas's own thermal radiation, the residual radiation of background radiation after being affected by the medium, and environmental stray radiation, this ill-conditioned equation characteristic of single-source observation and multiple unknown sources makes it difficult for the system to accurately separate the absorption and emission characteristics of the gas in complex backgrounds. Although simple image processing methods have been introduced, the lack of in-depth exploration of the spatiotemporal correlation of gas flow and the absence of a feedback correction mechanism based on physical constraints result in severe signal aliasing between the retrieved temperature field and concentration field. The processing flow is sensitive to background noise, has low measurement accuracy and a slow response, making it difficult to support accurate identification and rapid early warning of abnormal reaction states.

[0004] Therefore, improving the real-time performance, accuracy, and independence of physical field reconstruction of the deconstruction of the temperature and concentration characteristics of the reacting gas has become an urgent technical problem to be solved. Summary of the Invention

[0005] When the reactant gas is in the diffusion-dominated stage or reaches a local thermal equilibrium, although there is microscopic coupling, the spatial distribution of the temperature field and concentration field can be regarded as being dominated by different physical mechanisms in terms of macroscopic statistical characteristics. They are controlled by thermal conductivity and mass diffusion coefficient, respectively, and the two show a weak correlation statistically. Therefore, their joint probability distribution should be close to the product of the marginal distributions, and their mutual information value should approach zero. Especially in the region dominated by the mass transport mechanism, or as a regularization constraint to suppress the divergence of the solution space in the strong reaction coupling region, this statistical independence is based on the difference between mass transport and energy transport in the microscopic mechanism, thus showing a decoupling feature in the macroscopic distribution texture. This system uses this physical statistical characteristic as an information theory constraint for feedback optimization. The purpose of this invention is to provide a reactive gas temperature and concentration distribution monitoring system based on infrared thermal imaging analysis to solve the problems mentioned in the background art. Specifically, the technical solution of this invention includes: The image data acquisition module is used to acquire an infrared thermal image sequence containing the flow state of the reactive gas and simultaneously call the preset background radiation reference data. The flow field dynamic analysis module is used to perform spatiotemporal correlation analysis based on the infrared thermal image sequence, calculate the optical flow vector field of the reactant gas, and extract the dynamic gas region based on the optical flow vector field. The radiation coupling deconstruction module is used to combine the background radiation reference data with the pixel intensity of the dynamic gas region to perform nonlinear radiation signal separation processing to generate the thermal radiation dominant component matrix and the medium attenuation dominant component matrix. The dual-field independent reconstruction module is used to generate a gas temperature distribution field based on the dominant component matrix of thermal radiation and to generate a gas concentration distribution field based on the dominant component matrix of medium attenuation. An adaptive feedback optimization module is used to generate feedback correction coefficients based on the mutual information characteristics of the gas temperature distribution field and the gas concentration distribution field, and to iteratively update the calculation parameters of the radiation coupling deconstruction module using the feedback correction coefficients.

[0006] Preferably, the flow field dynamic analysis module performs the following operations when calculating the optical flow vector field of the reacting gas: Call the current frame image and the previous frame image in the infrared thermal imaging sequence; Based on the brightness conservation assumption and spatial smoothness constraint, the instantaneous velocity vector of the pixel in the current frame image is calculated; Construct the optical flow vector field describing the direction and velocity of gas flow; The optical flow vector field is used to characterize the transport path and diffusion trend of the reactant gas.

[0007] Preferably, the radiation coupling deconstruction module performs the following operations when performing nonlinear radiation signal separation processing: The background radiation reference data is introduced as a known boundary condition; Construct a set of radiative transfer equations that include the gas's own emissivity and gas transmissivity; By utilizing the spatial non-uniformity of the background radiation reference data, differential calculations are performed on the radiation differences of the same gas mass in different background regions within the dynamic gas region. Solve the radiative transfer equations to separate the dominant thermal radiation component matrix, which characterizes the thermal emission intensity of the gas itself, and the dominant medium attenuation component matrix, which characterizes the degree of absorption of background radiation by the gas.

[0008] Preferably, the radiation coupling deconstruction module is also configured with temperature and concentration feature decoupling logic, used for: Calculate the mutual information value between the dominant thermal radiation component matrix and the dominant medium attenuation component matrix; Determine whether the mutual information value is less than a preset independence threshold; If the mutual information value is greater than or equal to the independence threshold, the weighting coefficients in the radiative transfer equations are adjusted until the mutual information value between the two matrices converges to the minimum. The minimization of the mutual information value represents the maximum statistical separation between the temperature feature and the concentration feature.

[0009] Preferably, the dual-field independent reconstruction module performs the following operations when inverting and generating the gas temperature distribution field and the gas concentration distribution field: Call the preset temperature-radiation mapping model and concentration-transmittance mapping model; The dominant component matrix of thermal radiation is input into the temperature-radiation mapping model to generate the gas temperature distribution field. The dominant component matrix of medium attenuation is input into the concentration-transmittance mapping model to generate the gas concentration distribution field. Wherein, both the gas temperature distribution field and the gas concentration distribution field are two-dimensional scalar fields with the same resolution as the infrared thermal image sequence.

[0010] Preferably, the adaptive feedback optimization module is also used to perform flow field consistency verification: Based on the temporal changes of the gas temperature distribution field or the gas concentration distribution field, the reconstructed flow field vector is derived; Calculate the matching error between the reconstructed flow field vector and the optical flow vector field; If the matching error exceeds a preset physical constraint threshold, a negative feedback instruction is generated to correct the background radiation reference data of the radiation coupling deconstruction module. The matching error is used to evaluate whether the reconstruction result conforms to the principle of fluid dynamics continuity.

[0011] Preferably, when acquiring background radiation reference data, the image data acquisition module performs the following operations: Under the initial state of being filled with unreacted gas, thermal radiation images of the background environment are acquired; Alternatively, identify static regions in the infrared thermal image sequence whose optical flow vector magnitude is less than the stationary noise threshold, and extract the radiation value of the static region as a local background reference; if the optical flow magnitude of the entire field of view is greater than or equal to the stationary noise threshold, i.e., no static region can be found, then call the background reference data stored at the previous moment or the initially calibrated background data as a reference. The collected or extracted data is subjected to smooth interpolation to generate the background radiation reference data covering the entire field of view.

[0012] Preferably, this system also includes: The status assessment and early warning module is used for: Receive the gas temperature distribution field and the gas concentration distribution field; Identify local overheated regions in the gas temperature distribution field and abnormal concentration gradient regions in the gas concentration distribution field; When the temperature value of the local overheated area or the gradient value of the abnormal concentration gradient area exceeds the safety threshold, an abnormal reaction warning signal is output.

[0013] Compared with the prior art, the present invention has the following improvements and advantages: 1. This invention introduces statistical mutual information features as constraints in the radiation coupling deconstruction module, and uses an adaptive feedback optimization module to dynamically adjust the background difference confidence weights in the radiative transfer equations. Through iterative updates until the mutual information values ​​of the dominant thermal radiation component and the dominant medium attenuation component converge to a minimum, the system successfully separates the temperature and concentration characteristics of the gas to the greatest extent at the mathematical and statistical level. This overcomes the technical bottleneck in traditional infrared monitoring where the data dimension of a single detector is insufficient to distinguish the radiation characteristics of high-temperature gases and high-concentration gases, and significantly improves the independence and accuracy of physical field inversion in complex environments.

[0014] 2. The system of this invention calculates the optical flow vector field through a flow field dynamic analysis module. It can not only accurately capture the gas transport path and diffusion trend using brightness conservation and spatial smoothness constraints, but also intelligently identify static regions in the field of view to extract local background references. Combined with bilinear interpolation or Kriging interpolation algorithms, the system can dynamically piece together complete background reference data covering the entire field of view using gaps in gas flow during online operation, even without equipment shutdown or emptied reaction vessels. This dual-mode background acquisition strategy greatly enhances the system's adaptability to changes in the on-site environment, avoiding production losses caused by downtime calibration.

[0015] 3. The adaptive feedback optimization module of this invention innovatively establishes a verification logic based on flow field consistency. By comparing the matching error between the reconstructed flow field vector derived from the temperature field or humidity field and the actual observed optical flow vector field, it evaluates whether the inversion result conforms to the principle of fluid dynamics continuity. When the error exceeds the physical constraint threshold, the system generates a negative feedback command to correct the background radiation reference data. This physical-data dual-driven closed-loop control mechanism ensures that the final output temperature and concentration distribution fields are not only numerically convergent, but also physically and logically rigorously unified, effectively eliminating false data distortions caused by noise or algorithm deviations.

[0016] 4. This invention benefits from the dual-field independent reconstruction module, which transforms abstract radiation components into specific two-dimensional scalar fields. The system can independently assess local overheated regions in the gas temperature distribution field and abnormal concentration gradient regions in the gas concentration distribution field. This mechanism enables the system to accurately distinguish between simple high-temperature interference and dangerous leaks accompanied by high-concentration accumulation, or high-concentration diffusion risks in low-temperature environments. Compared to traditional single-threshold alarm methods, this multi-dimensional early warning mechanism has higher specificity and can provide more accurate and timely feedback on abnormal states for industrial safety production. Attached Figure Description

[0017] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a structural diagram of the system of the present invention; Detailed Implementation To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0018] Example 1: Please see Figure 1 A reactive gas temperature and concentration distribution monitoring system based on infrared thermal imaging analysis includes: The image data acquisition module is used to acquire an infrared thermal image sequence containing the flow state of the reactive gas and simultaneously call the preset background radiation reference data. The flow field dynamic analysis module is used to perform spatiotemporal correlation analysis based on infrared thermographic sequences, calculate the optical flow vector field of the reacting gas, and extract the dynamic gas region based on the optical flow vector field. The radiation coupling deconstruction module is used to combine background radiation reference data with pixel intensity of dynamic gas region to perform nonlinear radiation signal separation processing to generate thermal radiation dominant component matrix and medium attenuation dominant component matrix. The dual-field independent reconstruction module is used to generate the gas temperature distribution field based on the dominant component matrix of thermal radiation and to generate the gas concentration distribution field based on the dominant component matrix of medium attenuation. The adaptive feedback optimization module is used to generate feedback correction coefficients based on the mutual information characteristics of the gas temperature distribution field and the gas concentration distribution field, and to iteratively update the calculation parameters of the radiation coupling deconstruction module using the feedback correction coefficients.

[0019] This embodiment details the core architecture logic of the aforementioned system; the image data acquisition module is equipped with a high-sensitivity infrared focal plane array detector, which uses a sampling frequency determined by hardware settings. The unit is Hz, which refers to the sampling frequency, which has a physical meaning of time resolution and is measured in Hz, used to acquire infrared thermal image sequences of the interior of the reaction vessel. The flow field dynamic analysis module receives this sequence, uses the gas flow characteristics to locate the analysis object, and sets a static noise threshold. Dynamic gas regions are extracted based on the criterion that the optical flow modulus value is greater than the threshold. ; Here The specific setting logic is as follows: data is collected under conditions of no gas injection. Calculate the noise magnitude distribution of the optical flow field from the background image frame, and take the distribution... upper limit value as ,in The standard deviation is used to eliminate random thermal noise interference from static backgrounds; based on this, the radiative coupling deconstruction module, based on the principle of radiative transfer, calculates the observed pixel intensity. Considered as the dominant component of thermal radiation and the dominant component of medium attenuation Nonlinear combinations are used to solve ill-conditioned equation problems with single-source observations and dual-source unknowns; The dual-field independent reconstruction module uses a pre-calibrated physical mapping model to transform abstract radiation components into specific physical fields; the adaptive feedback optimization module is based on the physical assumption that temperature and humidity distributions should be statistically independent in a free diffusion flow field. If the mutual information is too high, a feedback gradient command is generated to iteratively update the solution parameters in the radiation coupling deconstruction module until the mutual information converges. right and All pixel values ​​in the matrix are discretized and binned. The number of pixels falling into each joint interval is counted, and the joint probability distribution is obtained after normalization. The estimate; and They are respectively and The marginal probability distribution is obtained by summing the joint distribution by rows or columns; where the mutual information eigenvalues ​​are... The calculation uses the histogram statistical method, based on the Shannon entropy definition: Update the solution parameters using gradient descent. That is, the background difference confidence weight, and the update formula follows: In the formula, For the number of iterations, For learning rate step size with adaptive dimensional compensation, The mutual information eigenvalues ​​are used to dynamically adjust the constraint strength during the deconstruction process.

[0020] Example 2: When calculating the optical flow vector field of the reacting gas, the flow field dynamic analysis module performs the following operations: Call the current frame image and the previous frame image in the infrared thermal imaging sequence; Based on the brightness conservation assumption and spatial smoothness constraint, the instantaneous velocity vector of the pixel in the current frame image is calculated; Construct an optical flow vector field that describes the direction and velocity of gas flow.

[0021] The optical flow vector field is used to characterize the transport path and diffusion trend of the reactant gas. This embodiment specifies the calculation logic of the flow field dynamic analysis module. To accurately capture the flow trend of the reactant gas, the system introduces a brightness conservation constraint equation, the physical basis of which is the assumption that the same gas mass will diffuse in a very short time. The internal radiation characteristics remain unchanged, and the calculation logic follows the formula: in, The image grayscale value originates from an infrared thermal imaging sequence. for The directional velocity component, in physical terms, represents the instantaneous velocity of a pixel; For time; to address the aperture problem in the above equations, this embodiment introduces a global smoothing constraint and constructs an energy functional. Minimize the solution: In the formula, The image grayscale values ​​are respectively in Spatial and time partial derivatives of the direction; The smoothing weighting coefficient is derived from a preset value and its physical meaning is the strength of the flow field continuity constraint. For gradient operators; The flow field dynamic analysis module obtains the instantaneous velocity vector of each pixel by iteratively solving for the minimum value of the above functional. Thus, an optical flow vector field is constructed; In this embodiment, under complex flow field scenarios with low infrared image contrast and blurred edges, the variational optical flow method is used to robustly calculate the motion vector of the gas. This optical flow vector field is not only used to extract dynamic regions, but also provides a key motion trajectory tracking basis for subsequent differential calculations based on the characteristic of the same gas mass moving between different backgrounds.

[0022] Example 3: When performing nonlinear radiation signal separation processing, the radiation coupling deconstruction module performs the following operations: Background radiation reference data are introduced as known boundary conditions; Construct a set of radiative transfer equations that include the gas's own emissivity and gas transmissivity; By utilizing the spatial non-uniformity of background radiation reference data, differential calculations are performed to address the radiation differences of the same gas mass within a dynamic gas region in different background regions. Solve the radiative transfer equations to separate the dominant thermal radiation component matrix, which characterizes the intensity of the gas's own thermal emission, and the dominant medium attenuation component matrix, which characterizes the degree of absorption of background radiation by the gas.

[0023] This embodiment elaborates on the core mechanism of the radiation coupling deconstruction module; in order to separate two variables from a single observation, the module first constructs a set of radiative transfer equations: in, This is the gas's own thermal radiation term, which physically represents the source term. For gas transmittance, its physical meaning is the medium attenuation coefficient; This is the background radiation value, sourced from baseline data; Using optical flow vector fields to track the same air mass over time and The moments are located in the background. and The position above; based on this, assuming in an extremely short time Inside, the air mass itself and Keeping the system constant, the system constructs a system of difference equations: and To address the potential numerical instability during the solution of the aforementioned equations, this embodiment introduces smoothing term coefficients that are dynamically updated by an adaptive feedback optimization module. , i.e., variable regularization parameter, physically means background difference confidence weight; its initial value Set as The unit is consistent with the square of the radiation intensity, that is... Specifically, this depends on the signal-to-noise ratio level of the image acquisition device. To prevent the singularity problem of zero denominator, a regularized least squares solution formula is constructed: Among them, the dominant component matrix of dielectric attenuation Each element is the transmissivity value of the gas at that pixel location, which is dimensionless. The regularization coefficient has the same dimensions as the regularization coefficient. Same, for That is, the square of the radiance, initial value Usually set to That is, one-thousandth of the background radiation variance, to avoid numerical instability caused by an excessively small denominator; This formula not only avoids numerical singularities, but also... The introduction of this establishes a control connection with the feedback loop; the solution is... Then, substitute the original equation to find the solution. This allows for a mathematically rigorous and robust separation of the shielding properties and spontaneous emission properties of gases.

[0024] Example 4: The radiation coupling deconstruction module is also equipped with temperature and concentration feature decoupling logic, used for: Calculate the mutual information between the dominant thermal radiation component matrix and the dominant medium attenuation component matrix; Determine whether the mutual information value is less than a preset independence threshold; If the mutual information value is greater than or equal to the independence threshold, the weighting coefficients in the radiative transfer equations are adjusted until the mutual information value between the two matrices converges to the minimum. Minimizing the mutual information value represents the maximum statistical separation between temperature and concentration characteristics.

[0025] This embodiment is equipped with temperature and concentration feature decoupling logic to solve... and The problem of signal aliasing; in order to establish optimization variables Regarding the effective gradient dependency of the solution results, this embodiment first modifies the deterministic difference solution formula in Embodiment 3, and constructs a formula containing weighted parameters. Dynamic solution model: Here The regularization coefficient is a variable whose physical meaning is the background difference confidence weight, and is used to calculate... Under this dependency, the system generates and Perform Z-score standardization to eliminate dimensional differences: in, The original data, The mean, Standard deviation; Calculate mutual information value ;like The system performs parameter updates; among which the independence threshold... The setting is based on the statistical benchmark value of mutual information between two independent random variables with a finite sample size, which is usually taken as a value. to ; Targeting mutual information To address the issue of non-differentiability caused by histogram statistics, this embodiment employs the finite difference perturbation method to estimate the gradient: in, The perturbation step size, in physical terms, represents the numerical differentiation accuracy; based on this estimated gradient, the system updates using the gradient descent method. The updated formula is as follows along with Iterative adjustments and The generation process is dynamically reshaped until the mutual information converges to a minimum, thereby achieving the maximum decoupling of temperature and concentration characteristics at the statistical level.

[0026] Example 5: When the dual-field independent reconstruction module generates the gas temperature distribution field and the gas concentration distribution field, it performs the following operations: Call the preset temperature-radiation mapping model and concentration-transmittance mapping model; Input the dominant component matrix of thermal radiation into the temperature-radiation mapping model to generate the gas temperature distribution field; Input the dominant component matrix of medium attenuation into the concentration-transmittance mapping model to generate a gas concentration distribution field; Among them, the gas temperature distribution field and the gas concentration distribution field are both two-dimensional scalar fields with the same resolution as the infrared thermal image sequence.

[0027] The system has a pre-stored lookup table of absorption spectral parameters for the target gas within typical temperature and pressure ranges. This table is generated by offline access to the HITRAN database and integral averaging for the detection band. During inversion, the temperature of the current iteration is used as the reference. and pressure divider The spectral line intensity is obtained from the lookup table using bilinear interpolation. and pressure widening half width ; This embodiment describes the specific inversion process of the dual-field independent reconstruction module; this module has a pre-stored temperature-radiation mapping model. and concentration-transmittance mapping model For temperature field inversion, a deformed model using Planck's formula is employed. Processing thermal radiation components The specific model is as follows: in, The first radiation constant is based on the definition of radiance, i.e. ; It is the second radiation constant; The detector's equivalent wavelength, in units of: ; The gas emissivity is dimensionless. The dominant component matrix of thermal radiation, after narrowband spectral equivalent processing, characterizes monochromatic radiance. Its physical dimensions after calibration are: ; within the logarithmic terms in the formula It is a dimensionless number; For concentration field inversion, the gas absorption coefficient is considered. Depends on the partial pressure of the gas to be determined This embodiment solves the problem of logical infinite loops using the fixed-point iteration method; the specific steps are as follows: Initialize the gas partial pressure field Environmental reference pressure; Enter the iteration loop According to the current and temperature Query the HITRAN molecular spectral database to obtain the target gas at its center wavelength. spectral line intensity at And Lorenz half-width And based on the Lorentz linear formula, the current single-molecule absorption cross section was calculated. : in For wave number, The center wavenumber is used; note that this calculation is for the single-molecule absorption cross section and does not include the molecular number density term. Calculate the intermediate variable transmittance To prevent noise from causing This leads to an error in the logarithmic domain, requiring the application of a nonnegativity truncation constraint: in, For effective transmittance, For numerically stable minimum values, such as This minimum value The setting is based on the machine precision of computer floating-point operations, aiming to avoid negative infinity errors in logarithmic operations; Will Substituting into the modified model of Beer-Lambert's law In the formula, For the corresponding pixel coordinates The effective optical path length at the point is determined based on the geometric model of the reaction vessel; Let Avogadro's constant be denoted by , and its value be . It is used to convert molecular number density into molar concentration; and to invert and generate a new gas molar concentration distribution field. ; Using the ideal gas law to determine the molar concentration field Converted into a new partial pressure field The conversion formula is: in, It is the ideal gas constant; Calculate residuals ,like That is, less than the convergence threshold The algorithm outputs the final gas concentration distribution field, effectively breaking the deadlock of parameter acquisition dependence and ensuring the stability of numerical calculation.

[0028] Example 6: The adaptive feedback optimization module is also used to perform flow field consistency verification: Based on the temporal changes of the gas temperature distribution field or gas concentration distribution field, the reconstructed flow field vector is derived; Calculate the matching error between the reconstructed flow field vector and the optical flow vector field; If the matching error exceeds the preset physical constraint threshold, a negative feedback instruction is generated to correct the background radiation reference data of the radiation coupling deconstruction module. The matching error is used to evaluate whether the reconstruction result conforms to the principle of fluid dynamics continuity.

[0029] In this embodiment, the adaptive feedback optimization module performs flow field consistency verification; the system first performs flow field verification based on the inverted field. or Solve the reverse convection-diffusion problem to reconstruct the flow field; specifically, construct the variational energy generalization function: In the formula, As a scalar field, the gas temperature distribution field obtained by inversion is used. As a scalar field diffusion coefficient Thermal diffusivity of a gas, unit: Preset values ​​can be obtained from the property table based on the type of gas and average temperature; To ensure smoothness and regularity, the value of the regularization factor is adaptively determined using the L-curve method. To explore the velocity field, by minimizing Obtain the reconstructed flow field ; Calculate its relationship with the observed optical flow field Matching error To address the fundamental limitation of the original feedback logic, which can only reduce the background reference in one direction, this embodiment introduces an error direction discrimination factor. Defined as: in, Describing the vector norm, For sign functions; when At that time, the corrected feedback formula is a pixel-level update strategy: in, This is the feedback regulation rate, which has the reciprocal unit of velocity. , used for dimensional normalization; To reconstruct the flow field With observation of optical flow field In pixel coordinates The local matching error magnitude at that point, i.e. At this point, the matching error... Has velocity dimension , with feedback regulation rate unit The multiplication results in a dimensionless correction term; this logic gives the system the ability to adjust the background radiation reference value for local areas, ensuring that the closed-loop control can truly converge to the optimal solution according to physical constraints.

[0030] Example 7: When acquiring background radiation reference data, the image data acquisition module performs the following operations: Under the initial state of being filled with unreacted gas, thermal radiation images of the background environment are acquired; Alternatively, identify static regions in the infrared thermal image sequence whose optical flow vector magnitude is less than the stationary noise threshold, and extract the radiation value of the static region as a local background reference; if the optical flow magnitude of the entire field of view is greater than or equal to the stationary noise threshold, i.e., no static region can be found, then call the background reference data stored at the previous moment or the initially calibrated background data as a reference. The collected or extracted data is smoothed by interpolation to generate background radiation reference data covering the entire field of view.

[0031] This embodiment provides background radiation reference data. The acquisition strategy ensures the accuracy of the decoupled benchmark; in offline mode, the system starts the heating device to the operating temperature in the initial state where the reaction vessel is not filled with reactant gas, and directly acquires the background thermal radiation image; in online mode, the system first identifies the optical flow vector magnitude in the infrared thermal image sequence. The area was identified as a static background area that was not blocked by gas. Then, the radiation value of this area was extracted as a sparse sample. Using bilinear interpolation or kriging interpolation algorithms, the background radiation value of the area blocked by gas was estimated, and finally background radiation reference data covering the entire field of view was generated. This embodiment solves the problem of difficult background radiation data acquisition through a dual-mode strategy. In particular, it uses the gaps in gas flow to dynamically piece together a complete background reference, allowing the system to acquire effective reference data even when it is impossible to shut down and empty the container, thus significantly improving the system's field adaptability.

[0032] Example 8: The system also includes a status assessment and early warning module, used for: Receive the gas temperature distribution field and the gas concentration distribution field; Identify localized overheated regions in the gas temperature distribution field and regions with abnormal concentration gradients in the gas concentration distribution field; When the temperature value of a local overheated area or the gradient value of an abnormal concentration gradient area exceeds the safety threshold, an abnormal reaction warning signal is output.

[0033] This embodiment adds a status assessment and early warning module to achieve intelligent security monitoring; this module first receives... and and scan Field, identify temperature value The connected components; simultaneously, calculate Spatial gradient of the field Identify gradient values The area; where the safety threshold is... and All are pre-calibrated according to the explosion limits and critical reaction temperatures specified in the "Safety Operating Procedures for Chemical Processes" or "Material Safety Data Sheets" for the target reactive gases; If any of the above indicators exceeds the safety threshold, the system determines that local overheating or abnormal concentration accumulation has occurred, and then outputs an abnormal reaction warning signal. This embodiment, based on decoupled independent field data, can distinguish between simple high temperature and high temperature accompanied by high concentration accumulation, or high concentration leakage at low temperature. This multi-dimensional early warning mechanism has higher specificity and sensitivity than a single temperature threshold alarm, effectively reducing false alarm and false negative rates. To verify the convergence and accuracy of the adaptive feedback optimization algorithm in this embodiment, an ammonia leakage simulation scenario was constructed. The ambient background temperature was set to... The true temperature of the leaked gas is The central concentration is Monitoring is performed using this system, and the adaptive feedback optimization module then... After the next iteration, the mutual information values ​​between the dominant thermal radiation component and the dominant medium attenuation component are... From the initial convergence to ; Meanwhile, the relative error of the gas concentration distribution field output by the dual-field independent reconstruction module is determined by the initial linear compensation algorithm. Descending to The root mean square error of the gas temperature distribution field was reduced to This demonstrates the effectiveness of our system in deconstructing complex coupled fields at the data level.

[0034] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A system for monitoring the temperature and concentration distribution of reactive gases based on infrared thermal imaging analysis, characterized in that, include: The image data acquisition module is used to acquire an infrared thermal image sequence containing the flow state of the reactive gas and simultaneously call the preset background radiation reference data. The flow field dynamic analysis module is used to perform spatiotemporal correlation analysis based on the infrared thermal image sequence, calculate the optical flow vector field of the reactant gas, and extract the dynamic gas region based on the optical flow vector field. The radiation coupling deconstruction module is used to combine the background radiation reference data with the pixel intensity of the dynamic gas region to perform nonlinear radiation signal separation processing to generate the thermal radiation dominant component matrix and the medium attenuation dominant component matrix. The dual-field independent reconstruction module is used to generate a gas temperature distribution field based on the dominant component matrix of thermal radiation and to generate a gas concentration distribution field based on the dominant component matrix of medium attenuation. An adaptive feedback optimization module is used to generate feedback correction coefficients based on the mutual information characteristics of the gas temperature distribution field and the gas concentration distribution field, and to iteratively update the calculation parameters of the radiation coupling deconstruction module using the feedback correction coefficients; When performing nonlinear radiation signal separation processing, the radiation coupling deconstruction module performs the following operations: The background radiation reference data is introduced as a known boundary condition; Construct a set of radiative transfer equations that include the gas's own emissivity and gas transmissivity; By utilizing the spatial non-uniformity of the background radiation reference data, differential calculations are performed on the radiation differences of the same gas mass in different background regions within the dynamic gas region. Solve the radiative transfer equations to separate the dominant thermal radiation component matrix, which characterizes the thermal emission intensity of the gas itself, and the dominant medium attenuation component matrix, which characterizes the degree of absorption of background radiation by the gas.

2. The reaction gas temperature and concentration distribution monitoring system based on infrared thermal imaging analysis according to claim 1, characterized in that, The flow field dynamic analysis module performs the following operations when calculating the optical flow vector field of the reacting gas: Call the current frame image and the previous frame image in the infrared thermal imaging sequence; Based on the brightness conservation assumption and spatial smoothness constraint, the instantaneous velocity vector of the pixel in the current frame image is calculated; Construct the optical flow vector field describing the direction and velocity of gas flow; The optical flow vector field is used to characterize the transport path and diffusion trend of the reactant gas.

3. The reaction gas temperature and concentration distribution monitoring system based on infrared thermal imaging analysis according to claim 1, characterized in that, The radiation coupling deconstruction module is also configured with temperature and concentration feature decoupling logic, used for: Calculate the mutual information value between the dominant thermal radiation component matrix and the dominant medium attenuation component matrix; Determine whether the mutual information value is less than a preset independence threshold; If the mutual information value is greater than or equal to the independence threshold, the weighting coefficients in the radiative transfer equations are adjusted until the mutual information value between the two matrices converges to the minimum. The minimization of the mutual information value represents the maximum statistical separation between the temperature feature and the concentration feature.

4. The reactive gas temperature and concentration distribution monitoring system based on infrared thermal imaging analysis according to claim 1, characterized in that, The dual-field independent reconstruction module performs the following operations when inverting and generating the gas temperature distribution field and the gas concentration distribution field: Call the preset temperature-radiation mapping model and concentration-transmittance mapping model; The dominant component matrix of thermal radiation is input into the temperature-radiation mapping model to generate the gas temperature distribution field. The dominant component matrix of medium attenuation is input into the concentration-transmittance mapping model to generate the gas concentration distribution field. Wherein, both the gas temperature distribution field and the gas concentration distribution field are two-dimensional scalar fields with the same resolution as the infrared thermal image sequence.

5. The reaction gas temperature and concentration distribution monitoring system based on infrared thermal imaging analysis according to claim 1, characterized in that, The adaptive feedback optimization module is also used to perform flow field consistency verification: Based on the temporal changes of the gas temperature distribution field or the gas concentration distribution field, the reconstructed flow field vector is derived; Calculate the matching error between the reconstructed flow field vector and the optical flow vector field; If the matching error exceeds a preset physical constraint threshold, a negative feedback instruction is generated to correct the background radiation reference data of the radiation coupling deconstruction module. The matching error is used to evaluate whether the reconstruction result conforms to the principle of fluid dynamics continuity.

6. The reactive gas temperature and concentration distribution monitoring system based on infrared thermal imaging analysis according to claim 1, characterized in that, When acquiring background radiation reference data, the image data acquisition module performs the following operations: Under the initial state of being filled with unreacted gas, thermal radiation images of the background environment are acquired; Alternatively, identify static regions in the infrared thermal image sequence whose optical flow vector magnitude is less than the stationary noise threshold, and extract the radiation value of the static region as a local background reference; if the optical flow magnitude of the entire field of view is greater than or equal to the stationary noise threshold, i.e., no static region can be found, then call the background reference data stored at the previous moment or the initially calibrated background data as a reference. The collected or extracted data is subjected to smooth interpolation to generate the background radiation reference data covering the entire field of view.

7. The reaction gas temperature and concentration distribution monitoring system based on infrared thermal imaging analysis according to claim 1, characterized in that, Also includes: The status assessment and early warning module is used for: Receive the gas temperature distribution field and the gas concentration distribution field; Identify local overheated regions in the gas temperature distribution field and abnormal concentration gradient regions in the gas concentration distribution field; When the temperature value of the local overheated area or the gradient value of the abnormal concentration gradient area exceeds the safety threshold, an abnormal reaction warning signal is output.

Images