Intelligent Detection Method and System for Remote Sensing of Methane Emissions Based on Sparse Spectrum Reconstruction

Abstract

The present invention provides a remote sensing intelligent detection method and system for methane emissions based on sparse spectral reconstruction, including: performing first-order Taylor expansion on the Beer-Lambert law; introducing spatial continuity constraints and physical diffusion constraint terms based on the wind field to derive an optimized objective function for inverting the atmospheric methane increment; simulating remote sensing observed hyperspectral radiance values under different methane increment concentrations and calculating the unit absorption characteristic spectrum; performing dimensionality reduction processing and clustering operations on the remote sensing hyperspectral image; performing matrix decomposition and sparse reconstruction on the spectral values according to the clustering results; equating the hyperspectral image reconstructed by matrix sparse decomposition to the background spectral image and calculating the atmospheric methane increment; generating atmospheric methane plume samples based on a large eddy simulation program; constructing a training sample set for machine learning estimation of the methane emission rate; constructing a machine learning model for estimating the atmospheric methane emission rate; and completing the estimation of the methane point source emission rate in the real scene based on the inversion result of the atmospheric methane increment.

Description

Technical Field

[0001] The present invention relates to the field of methane gas emission monitoring, and in particular to a methane emission remote sensing intelligent detection method and system based on sparse spectrum reconstruction. Background Art

[0002] Methane in the atmosphere is an important greenhouse gas with a global warming potential far greater than that of carbon dioxide. Accurate monitoring of its abnormal emissions is of great significance for environmental and climate management. Remote sensing technology has been widely used in recent years to monitor atmospheric methane concentrations due to its advantages of large range, all-weather and non-contact. Matched filter technology is a method that compares the observed spectrum with the known characteristic spectrum of methane to detect and quantify its enhancement in the atmosphere. It is widely used in satellite and aerial remote sensing for monitoring gas emission sources and concentration inversion.

[0003] The matched filter algorithm separates the methane signal from the background spectrum by matching the methane absorption characteristics with the "observation-background" difference spectrum and maximizing the signal-to-noise ratio using a Gaussian model. Existing matched filter techniques face many challenges in methane incremental inversion. First, in traditional algorithms, the background spectrum is usually approximated using the column average spectrum of the image; however, this average spectrum method is difficult to effectively handle surface covers or atmospheric interference with similar methane absorption characteristics, resulting in a large number of misjudgments in complex surface types or low-emission scenarios. Secondly, satellite remote sensing has a lower signal-to-noise ratio than airborne measurements. Due to the decrease in detection sensitivity, the inverted methane plume is prone to problems such as discontinuity and blurred boundaries in space. Finally, the computational efficiency of existing algorithms is low, especially when multiple iterations are required to update the background spectrum, which is particularly time-consuming when processing large-scale data. These existing technical problems significantly affect the accuracy of plume emission rate estimation based on incremental inversion results.

[0004] The traditional integrated mass enhancement algorithm aims to use the instantaneous atmospheric methane concentration distribution inverted by remote sensing, combined with the spatial geometry of the plume and wind speed data, to estimate the average emission rate under the assumption that the emission source intensity remains unchanged. In this process, the atmospheric methane increment inverted by the above steps provides the total mass information of methane emissions, and the geometry of the plume diffusion and the observed wind speed provide the basis for the estimation of emission time information. However, the existing methods still have certain limitations in the accuracy and robustness of emission rate estimation: first, the complex surface and atmospheric conditions make the spatial distribution mode of methane plumes random and diverse, and there is a high uncertainty in quantifying the plume diffusion time by equivalent diffusion length and equivalent wind speed; second, the errors in input data such as methane increment and simulated wind speed can directly and linearly affect the accuracy of emission rate estimation. Therefore, this method is prone to obvious emission rate underestimation under low wind speed and low emission conditions.

[0005] In view of the above problems, there is an urgent need for a technical solution that can simultaneously solve the following problems:

[0006] (1) The problems of false alarms and misestimations of incremental concentrations in complex surface areas and low-emission scenarios;

[0007] (2) The decrease in detection sensitivity caused by the signal-to-noise ratio of remote sensing hyperspectral observation data;

[0008] (3) The problem of underestimation of emission rates under low wind speed and low emission conditions;

[0009] (4) The problem of low efficiency of traditional iterative algorithms when processing large-scale data

[0010] Patent document CN116559902A (application number: 202310451358.5) discloses a method and system for rapid remote sensing identification and flux estimation of near-surface methane abnormal emissions, including: calculating the unit absorption characteristic spectra of near-surface methane and other trace gases using a radiative transfer model; screening the best channels for near-surface methane increment inversion; constructing a spectral similarity pixel set and a reference background spectrum for the pixel to be inverted; using the matched filtering method for optimized iterative inversion of near-surface methane increment; using morphological operation methods to identify methane point source emission outlets and methane point source emission plumes; and calculating the methane point source emission flux by combining the wind field information of the methane point source emission outlets. Summary of the Invention

[0011] Aiming at the defects in the prior art, the purpose of the present invention is to provide a remote sensing intelligent detection method and system for methane emissions based on sparse spectral reconstruction.

[0012] A remote sensing intelligent detection method for methane emissions based on sparse spectral reconstruction provided by the present invention includes

[0013] Step S1: Perform a first-order Taylor expansion on the Beer-Lambert law and introduce spatial continuity and wind field diffusion constraint terms to obtain the objective function for atmospheric methane increment inversion;

[0014] Step S2: Simulate remote sensing observation hyperspectral radiance data under different methane increment concentrations, and calculate the unit absorption characteristic spectra of methane increments;

[0015] Step S3: Perform dimensionality reduction processing and clustering operations on the remote sensing hyperspectral image, and perform matrix sparse decomposition and reconstruction on the spectral values according to the clustering results to obtain a reconstructed hyperspectral image;

[0016] Step S4: Equivalent the reconstructed hyperspectral image to a background spectral image, and solve the objective function for atmospheric methane increment inversion based on the background spectral image and the unit absorption characteristic spectra of methane increments to calculate the atmospheric methane increment inversion result;

[0017] Step S5: Generate an atmospheric methane increment plume sample and construct a training sample set for machine learning estimation of methane emission rate;

[0018] Step S6: Construct a machine learning model for estimating atmospheric methane emission rate, and use the training sample set to train the machine learning model for estimating atmospheric methane emission rate to obtain a trained machine learning model for estimating atmospheric methane emission rate; use the trained machine learning model for estimating atmospheric methane emission rate, and based on the atmospheric methane increment inversion result, complete the estimation of methane point source emission rate in the real scene.

[0019] Preferably, the step S1 includes:

[0020] Step S1.1: Perform a first-order Taylor expansion on the Beer-Lambert law:

[0021]

[0022] where is the remotely sensed hyperspectral sensor observed radiance data, is the background spectrum without methane increment absorption, is the equivalent optical thickness of methane gas;

[0023] Step S1.2: Define that the observed spectrum consists of a background spectrum and an increment spectrum:

[0024]

[0025] where is the increment of methane concentration relative to the background concentration, is the unit absorption characteristic spectrum;

[0026] Step S1.3: Define an optimization objective function for solving the methane increment concentration:

[0027]

[0028] where represents the remotely sensed observation sample number, is the error covariance matrix of the observed spectrum, is the spatial continuity regularization term, is the wind field physical diffusion regularization term;

[0029] where the spatial continuity regularization term adopts:

[0030]

[0031] where is the Laplace filter kernel, is the weight parameter of the spatial continuity constraint term; To initialize the methane increment; Indicates performing a spatial filtering operation on the methane increment concentration at the (x, y) spatial position;

[0032] The physical diffusion regularization term adopts:

[0033]

[0034] Among them, Indicates the wind vector field, Is the component of the wind field in the Direction, Is the component of the wind field in the Direction, Indicates the partial derivative of the initialized methane increment Of, Is the weight parameter of the wind field physical diffusion constraint term.

[0035] Preferably, the step S2 includes: based on the auxiliary observation conditions of hyperspectral remote sensing data, setting the simulation conditions of the atmospheric radiative transfer model; changing the initial background concentration of methane, and simulating the hyperspectral radiance data under different increment concentrations; calculating the radiance change amount corresponding to changing the unit methane increment for each band to obtain the unit absorption characteristic spectrum of the methane increment;

[0036] Among them, the auxiliary observation conditions based on hyperspectral remote sensing data include: the central wavelength of the satellite sensor, the full width at half maximum, the zenith and azimuth angles of the sun and the satellite, the atmospheric state, and the surface elevation.

[0037] Preferably, the step S3 includes:

[0038] Step S3.1: Perform principal component dimensionality reduction processing on the hyperspectral radiance image Including data standardization, calculating the covariance matrix, performing eigenvalue decomposition on the covariance matrix, and selecting the first Largest eigenvalues and eigenvectors to form the dimensionality reduction image ;

[0039] Step S3.2: Use the clustering algorithm to perform ground object clustering on the dimensionality-reduced image To obtain the class label to which each pixel belongs ;

[0040] Step S3.3: Perform matrix sparse decomposition on the spectral sample set composed of it :

[0041]

[0042] Among them, Is The left singular matrix, which contains spatial structure information; is a diagonal matrix that contains singular values; is the right singular matrix of , which contains spectral band information;

[0043] Step S3.4: Sparsify the original matrix by selecting the first largest singular values:

[0044]

[0045] where , and are the left singular matrix, diagonal matrix, and right singular matrix that retain the first singular values;

[0046] Step S3.5: Merge the reconstructed spectral matrices of all classes to obtain the complete reconstructed hyperspectral image .

[0047] Preferably, the said Step S4 includes:

[0048] Step S4.1: Solve the analytical solution form of the methane increment :

[0049]

[0050] where R represents the reconstructed background spectrum, represents the unit absorption characteristic spectrum of the methane increment;

[0051] Step S4.2: Calculate the error covariance matrix of the initial observed spectrum:

[0052] ;

[0053] where N represents the number of hyperspectral bands;

[0054] Step S4.3: Correct the methane unit absorption characteristic spectrum:

[0055]

[0056] where is the dot product operator;

[0057] Step S4.4: Calculate the initial methane increment :

[0058]

[0059] Step S4.5: Update the error covariance matrix :

[0060]

[0061] Step S4.6: Calculate the spatial continuity constraint term :

[0062]

[0063]

[0064] Step S4.7: Calculate the physical diffusion constraint term :

[0065]

[0066] Step S4.8: Calculate the methane increment after introducing the constraint term :

[0067]

[0068] Step S4.9: According to the objective function for retrieving the atmospheric methane increment, traverse the pixels in the hyperspectral image to obtain the retrieval result of the methane increment.

[0069] Preferably, the step S5 includes:

[0070] Step S5.1: Use large eddy simulation software, adopt the Smagorinsky SGS model, set multi-dimensional hierarchical simulation parameters, and generate atmospheric methane increment plume samples;

[0071] Among them, the multi-dimensional hierarchical simulation parameters include: emission rate, wind speed, turbulence intensity, diffusion intensity, temperature, time step, and spatial step;

[0072] Step S5.2: Based on the generated atmospheric methane increment plume samples, extract plume morphological features and construct a training sample set for machine learning estimation of methane emission rate;

[0073] Among them, the training sample set for machine learning estimation of methane emission rate includes: emission rate, wind speed, increment concentration distribution characteristics, and plume morphological features of the atmospheric methane increment plume samples.

[0074] Preferably, the step S6 includes:

[0075] Step S6.1: Construct a machine learning model for estimating atmospheric methane emission rate;

[0076]

[0077] Among them, represents the emission rate data in the atmospheric methane increment plume data, represents the wind speed magnitude in the simulated methane emissions, , , and are respectively the cumulative sum, mean, standard deviation, and peak value of the methane increment concentration, representing the plume morphological characteristics;

[0078] Step S6.2: Based on the constructed training sample set for machine learning estimation of methane emission rate, use machine learning techniques with regression fitting ability including random forest for model parameter training;

[0079] Step S6.3: Extract the true emission plume from the atmospheric methane increment inversion result, and use the trained machine learning model for estimating atmospheric methane emission rate to obtain the estimated result of the emission rate corresponding to the true emission plume.

[0080] According to a remote sensing intelligent detection system for methane emissions based on sparse spectral reconstruction provided by the present invention, it includes

[0081] Module M1: Perform a first-order Taylor expansion on the Beer-Lambert law and introduce spatial continuity and wind field diffusion constraint terms to obtain the objective function for atmospheric methane increment inversion;

[0082] Module M2: Simulate remote sensing observation hyperspectral radiance data under different methane increment concentrations, and calculate the unit absorption characteristic spectrum of methane increment;

[0083] Module M3: Perform dimensionality reduction processing and clustering operation on the remote sensing hyperspectral image, perform matrix sparse decomposition and reconstruction on the spectral values according to the clustering result to obtain the reconstructed hyperspectral image;

[0084] Module M4: Equivalent the reconstructed hyperspectral image to the background spectral image, solve the objective function for atmospheric methane increment inversion based on the background spectral image and the unit absorption characteristic spectrum of methane increment, and calculate the atmospheric methane increment inversion result;

[0085] Module M5: Generate atmospheric methane increment plume samples and construct a training sample set for machine learning estimation of methane emission rate;

[0086] Module M6: Construct a machine learning model for estimating atmospheric methane emission rate, and use the training sample set to train the machine learning model for estimating atmospheric methane emission rate to obtain the trained machine learning model for estimating atmospheric methane emission rate; use the trained machine learning model for estimating atmospheric methane emission rate, based on the atmospheric methane increment inversion result, to complete the estimation of methane point source emission rate in the real scene.

[0087] Preferably, the module M1 includes:

[0088] Module M1.1: Perform a first-order Taylor expansion on the Beer-Lambert law:

[0089]

[0090] where is the observed radiance data of the remote sensing hyperspectral sensor, is the background spectrum without methane incremental absorption, is the equivalent optical thickness of methane gas;

[0091] Module M1.2: Define that the observed spectrum consists of a background spectrum and an incremental spectrum:

[0092]

[0093] where is the increment of methane concentration relative to the background concentration, is the unit absorption characteristic spectrum;

[0094] Module M1.3: Define the optimization objective function for solving the methane incremental concentration:

[0095]

[0096] where represents the serial number of the remote sensing observation sample, is the error covariance matrix of the observed spectrum, is the spatial continuity regularization term, is the physical diffusion regularization term of the wind field;

[0097] where the spatial continuity regularization term adopts:

[0098]

[0099] where is the Laplacian filter kernel, is the weight parameter of the spatial continuity constraint term; is the initialized methane increment; represents performing a spatial filtering operation on the methane incremental concentration at the (x,y) spatial position;

[0100] The physical diffusion regularization term adopts:

[0101]

[0102] where represents the wind vector field, is the component of the wind field in the direction, is the component of the wind field in the direction, represents the partial derivative of the initialized methane increment ; is the weight parameter of the physical diffusion constraint term of the wind field;

[0103] The module M2 includes: setting the simulation conditions of the atmospheric radiation transfer model based on the auxiliary observation conditions of hyperspectral remote sensing data; changing the initial background concentration of methane and simulating the hyperspectral radiance data at different increment concentrations; calculating the radiance change amount of each band corresponding to the change of unit methane increment to obtain the unit absorption characteristic spectrum of methane increment;

[0104] Among them, the auxiliary observation conditions based on hyperspectral remote sensing data include: the central wavelength of the satellite sensor, the full width at half maximum, the zenith angle and azimuth angle of the sun and the satellite, the atmospheric state, and the surface elevation;

[0105] The module M3 includes:

[0106] Module M3.1: performing principal component dimensionality reduction processing on the hyperspectral radiance image , including data standardization, calculating the covariance matrix, performing eigenvalue decomposition on the covariance matrix, and selecting the first largest eigenvalues and eigenvectors to form a dimensionality-reduced image ;

[0107] Module M3.2: using a clustering algorithm to perform land cover clustering on the dimensionality-reduced image to obtain the class label to which each pixel belongs ;

[0108] Module M3.3: performing matrix sparse decomposition on the spectral sample set it composes :

[0109]

[0110] Among them, is the left singular matrix of, containing spatial structure information; is a diagonal matrix, containing singular values; is the right singular matrix of, containing spectral band information;

[0111] Module M3.4: approximating the original matrix by selecting the first largest singular values to achieve sparse processing:

[0112]

[0113] Among them, , and are the left singular matrix, diagonal matrix, and right singular matrix that retain the first singular values;

[0114] Module M3.5: Merge the reconstructed spectral matrices of all categories to obtain the complete reconstructed hyperspectral image .

[0115] Preferably, the module M4 includes:

[0116] Module M4.1: Solve the analytical solution form of the methane increment :

[0117]

[0118] where R represents the reconstructed background spectrum, represents the unit absorption characteristic spectrum of the methane increment;

[0119] Module M4.2: Calculate the error covariance matrix of the initial observed spectrum :

[0120] ;

[0121] where N represents the number of hyperspectral bands;

[0122] Module M4.3: Correct the methane unit absorption characteristic spectrum:

[0123]

[0124] where is the dot product operator;

[0125] Module M4.4: Calculate the initial methane increment :

[0126]

[0127] Module M4.5: Update the error covariance matrix :

[0128]

[0129] Module M4.6: Calculate the spatial continuity constraint term :

[0130]

[0131]

[0132] Module M4.7: Calculate the physical diffusion constraint term :

[0133]

[0134] Module M4.8: Calculate the methane increment after introducing the constraint term :

[0135]

[0136] Module M4.9: Traverse the pixels in the hyperspectral image according to the objective function inverted from the atmospheric methane increment, and obtain the methane increment inversion result;

[0137] The said module M5 includes:

[0138] Module M5.1: Use large eddy simulation software, adopt the Smagorinsky SGS model, set multi-dimensional hierarchical simulation parameters, and generate atmospheric methane increment plume samples;

[0139] Among them, the said multi-dimensional hierarchical simulation parameters include: emission rate, wind speed, turbulence intensity, diffusion intensity, temperature, time step and spatial step;

[0140] Module M5.2: Based on the generated atmospheric methane increment plume samples, extract the plume morphological characteristics and construct a training sample set for machine learning estimation of methane emission rate;

[0141] Among them, the training sample set for machine learning estimation of methane emission rate includes: emission rate, wind speed, increment concentration distribution characteristics and plume morphological characteristics of the atmospheric methane increment plume samples;

[0142] The said module M6 includes:

[0143] Module M6.1: Construct a machine learning model for estimating atmospheric methane emission rate;

[0144]

[0145] Among them, represents the emission rate data in the atmospheric methane increment plume data, represents the wind speed magnitude in the simulated methane emission, 、 、 and are respectively the cumulative sum, mean, standard deviation and peak value of the methane increment concentration, represents the plume morphological characteristics;

[0146] Module M6.2: Based on the training sample set for machine learning estimation of methane emission rate constructed, use machine learning techniques with regression fitting ability including random forest to train model parameters;

[0147] Module M6.3: Extract the true emission plume from the retrieved results of atmospheric methane increment. The true emission plume uses the trained machine learning model for estimating atmospheric methane emission rate to obtain the estimation result of the emission rate corresponding to the true emission plume.

[0148] Compared with the prior art, the present invention has the following beneficial effects:

[0149] 1. The present invention solves the problem of inaccurate estimation of background spectrum under complex surface by performing matrix decomposition and sparse reconstruction on the observed spectral values;

[0150] 2. The present invention overcomes the false alarm and misestimation problems of traditional methods in complex surface areas and low emission scenarios by constructing a more accurate background spectrum estimation;

[0151] 3. The present invention optimizes the spatial structure and boundary clarity of the retrieved plume and improves the signal-to-noise ratio of the retrieved results by introducing a regularization term of spatial continuity and wind field diffusion constraint into the objective function of methane increment retrieval;

[0152] 4. The present invention reduces the sensitivity of the methane increment retrieval error and the model wind speed error to the estimation result of the emission rate by constructing a machine learning-based atmospheric methane emission rate estimation model;

[0153] 5. The present invention overcomes the problem of underestimation of emission rate in low wind speed and low emission scenarios by constructing a machine learning-based atmospheric methane emission rate estimation model;

[0154] 6. The present invention significantly improves the computational efficiency of the algorithm by eliminating the iterative optimization process in the traditional matching filtering algorithm and is applicable to the rapid processing of large-scale data. Description of the Drawings

[0155] By reading the detailed description of the non-restrictive embodiments with reference to the following drawings, other features, objectives and advantages of the present invention will become more obvious:

[0156] Figure 1 It is a flow chart of the intelligent remote sensing detection method for methane emission based on sparse spectral reconstruction.

[0157] Figure 2 It is a schematic diagram of the retrieval effect of the intelligent remote sensing detection method for methane emission based on sparse spectral reconstruction.

[0158] Figure 3 It is a schematic diagram of the estimation accuracy of the emission rate of the intelligent remote sensing detection method for methane emission based on sparse spectral reconstruction.

[0159] Figure 4 This is a comparative example of the remote sensing intelligent detection method for methane emissions based on sparse spectral reconstruction. Specific implementation manner

[0160] The present invention will be described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several changes and improvements can still be made. These all belong to the protection scope of the present invention.

[0161] Embodiment 1

[0162] A remote sensing intelligent detection method for methane emissions based on sparse spectral reconstruction provided by the present invention includes:

[0163] Step S1: According to the extinction process of light in a medium described by the Beer-Lambert law, perform a first-order Taylor expansion on it, and introduce a regularization term according to the spatial continuity constraint and the physical diffusion constraint based on the wind field, and derive an optimization objective function for inverting the atmospheric methane increment;

[0164] Step S2: Use the existing atmospheric radiative transfer model to simulate the satellite remote sensing observation hyperspectral radiance data under different methane increment concentrations, calculate the change in radiance with the change in increment concentration, and obtain the unit absorption characteristic spectrum of the methane increment;

[0165] Step S3: Use principal component analysis to perform dimensionality reduction processing on the hyperspectral remote sensing image, perform ground object clustering based on the spectral dimensionality reduction result, perform matrix decomposition and sparse reconstruction on the spectral values of each type of ground object according to the clustering result, and obtain the reconstructed spectrum of the remote sensing hyperspectral image;

[0166] Step S4: According to the spatial sparsity principle of the methane increment, equivalent the hyperspectral image reconstructed by matrix sparse decomposition to the background spectral image, solve the optimization objective function in Step S1, and calculate the inversion result of the atmospheric methane increment;

[0167] Step S5: Use the existing large-eddy simulation program, by setting key parameters such as the emission rate, generate a large-scale multi-dimensional hierarchical atmospheric methane increment plume sample, and construct a training sample set for machine learning estimation of the methane emission rate based on its emission rate, increment concentration distribution, and spatial morphological characteristics;

[0168] Step S6: Construct a machine learning model for estimating the atmospheric methane emission rate, use the training sample set obtained in Step S5 to train the machine learning model parameters, and apply the trained model to the atmospheric methane increment inversion result obtained in Step S4 to complete the estimation of the methane point source emission rate in the real scene.

[0169] Specifically, step S1 adopts:

[0170] Step S1.1: Apply the Beer-Lambert law to the atmospheric radiation transfer process. Assume that the radiation absorption caused by the trace gas methane is relatively small, and perform a first-order Taylor expansion on this law:

[0171]

[0172] where is the observed radiance data of the remote sensing hyperspectral sensor, is the background spectrum without the incremental absorption of methane, is the equivalent optical thickness of methane gas;

[0173] Step S1.2: According to the first-order Taylor expansion form of the Beer-Lambert law, define that the observed spectrum can be composed of the background spectrum and the incremental spectrum:

[0174]

[0175] where is the increment of methane concentration relative to the background concentration, is the unit absorption characteristic spectrum

[0176] Step S1.3: By minimizing the difference between the observed spectrum and the simulated spectrum , solve for the methane concentration increment , and define the optimization objective function for solving the methane increment concentration:

[0177]

[0178] where represents the serial number of the remote sensing observation sample, is the error covariance matrix of the observed spectrum, is the spatial continuity regularization term, is the physical diffusion regularization term of the wind field;

[0179] The incremental spectrum mentioned above adopts: the product of the methane unit absorption characteristic spectrum and the incremental concentration is approximately represented;

[0180] The spatial continuity regularization term mentioned above adopts:

[0181]

[0182] where is the Laplacian filter kernel, is the weight parameter of the spatial continuity constraint term, is the initialized methane increment; represents performing a spatial filtering operation on the methane increment concentration at the (x,y) spatial position;

[0183] The described physical diffusion regularization term adopts:

[0184]

[0185] where represents the wind vector field, is the component of the wind field in the direction, is the component of the wind field in the direction, represents the partial derivative of the initialized methane increment ; is the weight parameter of the wind field physical diffusion constraint term.

[0186] Specifically, the step S2 adopts:

[0187] Based on the auxiliary observation conditions of hyperspectral remote sensing data, set the simulation conditions of the atmospheric radiation transfer model; change the initial background concentration of methane, and simulate the hyperspectral radiance data under different increment concentrations; based on the simulation results, calculate the radiance change amount corresponding to changing a unit of methane increment for each band, and obtain the unit absorption characteristic spectrum of the methane increment.

[0188] The described auxiliary observation conditions include but are not limited to: the central wavelength of the satellite sensor, the full width at half maximum, the zenith angle and azimuth angle of the sun and the satellite, the atmospheric state, the surface elevation, etc.

[0189] The described radiation transfer models include but are not limited to: MODTRAN, SCIATRAN, etc.

[0190] Specifically, the step S3 adopts:

[0191] Step S3.1: Perform principal component dimensionality reduction processing on the hyperspectral radiance image including data standardization, calculating the covariance matrix, performing eigenvalue decomposition on the covariance matrix, and selecting the first largest eigenvalues and eigenvectors to form the dimensionality reduction image ;

[0192] Step S3.2: Use a clustering algorithm to perform ground object clustering on the dimensionality-reduced image to obtain the class label to which each pixel belongs ;

[0193] Step S3.3: For the spectral sample set composed of it Perform matrix sparse decomposition:

[0194]

[0195] Wherein is the left singular matrix of, containing spatial structure information, is a diagonal matrix, containing singular values, is the right singular matrix of, containing spectral band information;

[0196] Step S3.4: Approximate the original matrix by selecting the first largest singular values to achieve sparsification:

[0197]

[0198] Wherein , and are the left singular matrix, diagonal matrix and right singular matrix that retain the first singular values;

[0199] Step S3.5: Merge the reconstructed spectral matrices of all categories to restore the spatial structure of the original image and obtain the complete reconstructed hyperspectral image ;

[0200] The hyperspectral image described above employs: A hyperspectral image for methane incremental inversion, including two-dimensional spatial dimension and spectral dimension, and the spectral dimension covers the visible light to shortwave infrared bands;

[0201] The clustering algorithm includes but is not limited to: k-means clustering algorithm and Gaussian mixture model clustering algorithm.

[0202] Specifically, Step S4 employs:

[0203] Step S4.1: Solve the analytical solution form of the methane increment according to the optimization objective function in Step S1 :

[0204]

[0205] Wherein, and are the original observed hyperspectral image and the reconstructed hyperspectral image respectively, and are the spatial continuity constraint term and the physical diffusion constraint term based on the wind field respectively.

[0206] Step S4.2: Based on the original observed hyperspectral image and the reconstructed hyperspectral image , calculate the error covariance matrix of the initial observed spectrum :

[0207]

[0208] Step S4.3: Correct the methane unit absorption characteristic spectrum, where is the dot product operator:

[0209]

[0210] Step S4.4: Calculate the initialized methane increment :

[0211]

[0212] Step S4.5: Update the error covariance matrix according to the initialized methane increment :

[0213]

[0214] Step S4.6: Calculate the spatial continuity constraint term according to the initialized methane increment :

[0215]

[0216]

[0217] Step S4.7: Calculate the physical diffusion constraint term of the wind field in combination with the initialized methane increment :

[0218]

[0219] Step S4.8: Calculate the methane increment after introducing the constraint terms:

[0220]

[0221] Step S4.9: Traverse the pixels in the hyperspectral image to obtain the inversion result of the methane increment.

[0222] Specifically, the step S5 adopts:

[0223] Step S5.1: Use existing large-eddy simulation software, adopt the Smagorinsky SGS model, and set multi-dimensional hierarchical simulation parameters to generate a large-scale methane increment plume sample;

[0224] The multi-dimensional hierarchical simulation parameters are as follows:

[0225] Emission rate: 0 - 30 t / h; Wind speed: 0 - 30 m / s; Turbulence intensity: 5 - 25 %; Diffusion intensity: 0 - 10 m 2 / s; Temperature: 10 - 30 °C; Time step: 1 - 5 steps; Spatial step: 30 m

[0226] Step S5.2: Based on the atmospheric methane increment plume sample generated in the above step, extract the plume morphological features and construct a training sample set for machine learning estimation of methane emission rate.

[0227] The plume morphological features include: total plume area, total plume perimeter, expansion radius (in the XY direction), plume skewness (in the XY direction), plume kurtosis (in the XY direction), aspect ratio of the expansion radius, deviation of the centroid from the peak (in the XY direction), area covered by the increment plume greater than the threshold area, plume coverage rate, average gradient (in the XY direction), average gradient, deviation of the centroid from the geometric center (in the XY direction), plume density, average centroid distance, normalized shape index, eccentricity.

[0228] The training sample set for machine learning estimation of methane emission rate is composed of: extracting the emission rate, wind speed, increment concentration distribution characteristics, and plume morphological features of the atmospheric methane increment plume sample.

[0229] Specifically, step S6 adopts:

[0230] Step S6.1: Construct a machine learning estimation model for atmospheric methane emission;

[0231] The machine learning estimation model is as follows:

[0232]

[0233] Among them, represents the emission rate data in the atmospheric methane increment plume data, represents the wind speed magnitude in the simulated methane emission, , , and are respectively the cumulative sum, mean, standard deviation, and peak value of the methane increment concentration, represents the plume morphological features mentioned in step S5.

[0234] Step S6.2: Based on the constructed training sample set of methane emission rates, use machine learning techniques with regression fitting capabilities such as random forest for model parameter training;

[0235] Step S6.3: Extract the true emission plume from the atmospheric methane increment inversion result in Step S4, and apply the trained model to the true emission plume to obtain the estimated emission rate result corresponding to the true emission plume.

[0236] The present invention also provides a remote sensing intelligent detection system for methane emissions based on sparse spectral reconstruction. The remote sensing intelligent detection system for methane emissions based on sparse spectral reconstruction can be implemented by executing the process steps of the remote sensing intelligent detection method for methane emissions based on sparse spectral reconstruction. That is, those skilled in the art can understand the remote sensing intelligent detection method for methane emissions based on sparse spectral reconstruction as a preferred implementation manner of the remote sensing intelligent detection system for methane emissions based on sparse spectral reconstruction.

[0237] Example 2

[0238] Example 2 is a preferred example of Example 1

[0239] According to a remote sensing intelligent detection method for methane emissions based on sparse spectral reconstruction provided by the present invention, as Figures 1 to 4 shown, it includes:

[0240] Step 1: For methane gas with radiative absorption characteristics in the atmosphere, according to the extinction process of light in a medium described by the Beer-Lambert law, perform a first-order Taylor expansion on it, and introduce a regularization term according to the spatial continuity constraint and the physical diffusion constraint based on the wind field, and derive an optimization objective function for inverting the atmospheric methane increment.

[0241] Step 1.1: Apply the Beer-Lambert law to the atmospheric radiation transfer process, and assume that the radiative absorption amount that can be caused by trace gas methane is small, and perform a first-order Taylor expansion on this law:

[0242]

[0243] where is the observed spectrum of a remote sensing sensor with hyperspectral imaging capabilities, is the incident spectrum without methane absorption or the background spectrum without methane increment absorption, is the optical thickness of methane gas.

[0244] Step 1.2: According to the first-order Taylor expansion form of the Beer-Lambert law, assume that the observed spectrum can be composed of the background spectrum and the methane increment spectral absorption amount :

[0245]

[0246] Among them, is the increment of atmospheric methane concentration relative to the background concentration, is the unit absorption characteristic spectrum, and the background spectrum multiplied by the optical thickness of methane gas can be approximately expressed as the product of the unit absorption characteristic spectrum of methane and the incremental concentration .

[0247] Step 1.3: By minimizing the difference between the observed spectrum and the simulated spectrum , solve for the methane concentration increment , and define the optimization objective function as:

[0248]

[0249] Where represents the serial number in the entire satellite remote sensing observation sample set, is the error covariance matrix of the observed spectrum

[0250] Step 1.4: To ensure the spatial continuity of the methane incremental concentration inversion result, further add a regularization term based on the two-dimensional spatial gradient to the optimization objective function, then the optimization objective function is updated to:

[0251]

[0252]

[0253] Where is the Laplacian filter kernel, is the weight parameter of the spatial continuity constraint term, is the initialized methane increment; represents the spatial filtering operation on the methane incremental concentration at the (x,y) spatial position.

[0254] Step 1.5: To ensure that the methane incremental concentration inversion result conforms to the actual physical environment, further add a regularization term based on wind field diffusion to the optimization objective function, then the optimization objective function is updated to:

[0255]

[0256]

[0257] Where represents the wind vector field, is the wind field at The component in the direction, is the component of the wind field in the direction, represents the partial derivative of the initialized methane increment and is the weight parameter of the physical diffusion constraint term of the wind field.

[0258] Step 2: Based on the existing open-source atmospheric radiative transfer model, simulate the satellite remote sensing observed hyperspectral radiance data at different methane increment concentrations, calculate the change in radiance with the change in increment concentration, and obtain the unit absorption characteristic spectrum of the methane increment.

[0259] Step 2.1: Based on the auxiliary observation conditions of the satellite hyperspectral remote sensing data (including but not limited to the central wavelength of the satellite sensor, full width at half maximum, zenith and azimuth angles of the sun and satellite, atmospheric state, and surface elevation, etc.), set the simulation conditions of the atmospheric radiative transfer model (including but not limited to the MODTRAN, SCIATRAN models).

[0260] Step 2.2: In the atmospheric state simulation conditions, change the initial background concentration of methane and simulate the satellite remote sensing observed hyperspectral radiance data at different increment concentrations.

[0261] Step 2.3: Based on the simulation results, calculate the change in radiance corresponding to a unit change in methane increment for each band, and obtain the unit absorption characteristic spectrum of the methane increment .

[0262] Step 3: For the remote sensing hyperspectral image of the methane increment to be retrieved, use principal component analysis to reduce the dimension of the remote sensing image from the visible light to the shortwave infrared band, perform land cover clustering based on the spectral dimension reduction results, and perform matrix decomposition and sparse reconstruction on the spectral values of each land cover category according to the clustering results to obtain the reconstructed background spectrum of each pixel to be retrieved in the remote sensing hyperspectral image.

[0263] Step 3.1: Prepare the remote sensing hyperspectral image for methane increment retrieval , where and represent the spatial dimensions of the image, represents the spectral dimension of the image, and the spectral dimension covers the spectral information from visible light to shortwave infrared.

[0264] Step 3.2: Perform principal component analysis dimension reduction processing on the hyperspectral image , which includes data standardization, calculation of the covariance matrix, eigenvalue decomposition of the covariance matrix to obtain eigenvalues and eigenvectors, and selection of the first largest eigenvalues and the corresponding eigenvectors to form the dimension-reduced image of the hyperspectral image ​ , where represents the spectral dimension after spectral dimensionality reduction, the quantity is much smaller than .

[0265] Step 3.3: Set the number of clustering categories to , and use a clustering algorithm (including but not limited to the k-means clustering algorithm and the Gaussian mixture model clustering algorithm) to perform land cover clustering on the dimensionality-reduced image to obtain the class label to which each pixel belongs .

[0266] Step 3.4: For each type of land cover , perform matrix sparse decomposition on the original spectral sample set composed of it :

[0267]

[0268] where is the left singular matrix of , which contains the spatial structure information of the spectral data, is a diagonal matrix that contains singular values and represents the energy distribution of the spectral data, is 's right singular matrix, which contains spectral band information.

[0269] Step 3.5: Approximate the original matrix by selecting the first largest singular values to achieve sparsification processing, and obtain the reconstructed spectral data of the pixels belonging to class :

[0270]

[0271] where, , and are the left singular matrix, diagonal matrix, and right singular matrix that retain the first largest singular values.

[0272] where, the said first largest singular values represent the singular values whose variance interpretation rate is greater than 1%.

[0273] Step 3.6: Merge the reconstructed spectral matrices of all classes to restore the spatial structure of the original image and obtain the complete reconstructed hyperspectral image .

[0274] Step 4: According to the principle of spatial sparsity of methane increment, the hyperspectral image reconstructed by matrix sparse decomposition is equivalent to the background spectral image, and the optimization objective function in Step 1 is solved to calculate the inversion result of atmospheric methane increment.

[0275] Step 4.1: According to the optimization objective function in Step 1 , solve for the analytical solution form of methane increment :

[0276]

[0277] where and are the original observed hyperspectral image and the reconstructed hyperspectral image respectively, and are the spatial continuity constraint term and the physical diffusion constraint term based on the wind field respectively.

[0278] Step 4.2: According to the original observed hyperspectral image and the reconstructed hyperspectral image , calculate the error covariance matrix of the initial observed spectrum:

[0279]

[0280] Step 4.3: Correct the unit absorption characteristic spectrum of methane, where is the dot product operator:

[0281]

[0282] Step 4.4: Calculate the initialized methane increment according to the above variables:

[0283]

[0284] Step 4.5: Update the error covariance matrix according to the initialized methane increment :

[0285]

[0286] Step 4.6: Calculate the spatial continuity constraint term according to the initialized methane increment :

[0287]

[0288]

[0289] Step 4.7: According to the initialized methane increment Calculating the physical diffusion constraint term related to wind field information :

[0290]

[0291] Step 4.8: Calculate the methane increment after introducing the constraint term :

[0292]

[0293] Step 4.9: Traverse each pixel in the hyperspectral image to obtain the inversion result of the atmospheric methane increment with the same size as the image space

[0294] Step 5: By setting the key parameters of large-eddy simulation such as emission rate, atmospheric conditions, wind speed, etc., generate large-scale multi-dimensional hierarchical atmospheric methane increment plume samples, extract the emission rate, increment concentration distribution, and spatial morphological characteristics of the methane plume samples, and construct a training sample set for machine learning estimation of methane emission rate

[0295] Step 5.1: Use existing open-source fluid mechanics simulation software with large-eddy simulation function, adopt the Smagorinsky SGS model, and set multi-dimensional hierarchical simulation parameters to realize the generation of large-scale methane increment plume samples

[0296] The multi-dimensional hierarchical simulation parameters include: emission rate: 0 - 30 t / h, wind speed: 0 - 30 m / s, turbulence intensity: 5 - 25%, diffusion intensity: 0 - 10 m 2 / s, temperature: 10 - 30 °C, time step: 1 - 5 steps, spatial step: 30 m;

[0297] Step 5.2: Based on the atmospheric methane increment plume generated in the above steps, extract the plume morphological characteristics and construct a training sample set for machine learning estimation of methane emission rate

[0298] The plume morphological characteristics include:

[0299] Total plume area: , where represents the actual geographical distance represented by the image pixel, represents the plume pixel coordinates

[0300] Total plume perimeter: , where is the number of pairs of vertically or horizontally adjacent pixels of the plume boundary pixels, is the number of pairs of diagonally adjacent pixels

[0301] Expansion radius (XY direction): , , , , where is the incremental concentration at the coordinates in the methane plume, represents the cumulative incremental concentration of the complete plume, represents the centroid of the methane plume.

[0302] Plume skewness (XY direction): , .

[0303] Plume kurtosis (XY direction): , .

[0304] Expansion radius aspect ratio: .

[0305] Deviation of centroid from peak (XY direction): , , where represents the coordinates of the peak point of the methane plume concentration.

[0306] Area covered by the incremental plume greater than the threshold : .

[0307] Plume coverage rate: , where represents the area of the minimum bounding rectangle.

[0308] Average gradient (XY direction): , , where represents taking the average of all pixel results in the plume.

[0309] Average gradient: .

[0310] Deviation of centroid from geometric center (XY direction): , , where represents the coordinates of the geometric center of the methane plume concentration.

[0311] Plume density: .

[0312] Average centroid distance: .

[0313] Normalized shape index: .

[0314] Eccentricity: , , , where and are the covariance matrices eigenvalues.

[0315] Step 5.3: Extract the emission rate, wind speed, incremental concentration distribution characteristics, and plume morphology characteristics of the atmospheric methane incremental plume samples to form a training sample set for machine learning estimation of methane emission rate.

[0316] Step 6: Construct a machine learning model for estimating atmospheric methane emission rate. Use the training sample set obtained in Step 5 to train the machine learning model parameters, and apply the trained model to the atmospheric methane incremental inversion results obtained in Step 4 to obtain the corresponding emission rate.

[0317] Step 6.1: Construct a machine learning estimation model for atmospheric methane emission:

[0318]

[0319] where represents the emission rate data in the atmospheric methane incremental plume data, represents the wind speed magnitude in the simulated methane emission, , , and are respectively the cumulative sum, mean, standard deviation, and peak value of the methane incremental concentration, represents the plume morphology characteristics mentioned in Step 5.

[0320] Step 6.2: Based on the constructed methane emission rate training sample set, use machine learning techniques with regression fitting capabilities such as random forest to train the model parameters.

[0321] Step 6.3: Use 200 ppm·m as the threshold to distinguish environmental noise and methane emission, extract the true emission plume in the atmospheric methane incremental inversion results of Step 4, and apply the trained model to the true emission plume to obtain the corresponding emission rate estimation result.

[0322] As Figure 4 shown, the atmospheric methane monitoring effects of the traditional method and the method of the present invention are compared:

[0323] Among them, there is more spatial noise in the results obtained by the traditional method, while the results obtained by the method of the present invention can more clearly identify the methane emission plume; the deviation between the results obtained by the traditional method and the true value is larger, while the deviation between the results obtained by the method of the present invention and the true value is closer to 0; there are more false alarm increments in the results obtained by the traditional method, while the results obtained by the method of the present invention have fewer false alarm increments.

[0324] Those skilled in the art know that, in addition to implementing the systems, devices and their respective modules provided by the present invention in the form of pure computer-readable program code, the method steps can be logically programmed to enable the systems, devices and their respective modules provided by the present invention to be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. to achieve the same program. Therefore, the systems, devices and their respective modules provided by the present invention can be considered as a kind of hardware components, and the modules included therein for implementing various programs can also be regarded as the structures within the hardware components; the modules for implementing various functions can also be regarded as either software programs for implementing methods or structures within hardware components.

[0325] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. Without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A remote sensing intelligent detection method for methane emissions based on sparse spectral reconstruction, characterized in that, including Step S1: Perform a first-order Taylor expansion on the Beer-Lambert law and introduce spatial continuity and wind field diffusion constraint terms to obtain the objective function for atmospheric methane increment inversion; Step S2: Simulate remotely sensed observed hyperspectral radiance data at different methane increment concentrations, and calculate the unit absorption characteristic spectrum of methane increment; Step S3: Perform dimensionality reduction processing and clustering operations on the remotely sensed hyperspectral image, and perform matrix sparse decomposition and reconstruction on the spectral values according to the clustering results to obtain a reconstructed hyperspectral image; Step S4: Equivalent the reconstructed hyperspectral image to the background spectral image, and solve the objective function for atmospheric methane increment inversion based on the background spectral image and the unit absorption characteristic spectrum of methane increment to calculate the atmospheric methane increment inversion result; Step S5: Generate an atmospheric methane increment plume sample and construct a training sample set for machine learning estimation of methane emission rate; Step S6: Construct a machine learning model for estimating atmospheric methane emission rate, and use the training sample set to train the machine learning model for estimating atmospheric methane emission rate to obtain a trained machine learning model for estimating atmospheric methane emission rate; Use the trained machine learning model for estimating atmospheric methane emission rate to complete the estimation of methane point source emission rate in the real scene based on the atmospheric methane increment inversion result; The said Step S1 includes: Step S1.1: Perform a first-order Taylor expansion on the Beer-Lambert law: Among them, is the observed radiance data of the remote sensing hyperspectral sensor, is the background spectrum without the incremental absorption of methane, is the equivalent optical thickness of methane gas; Step S1.2: Define that the observed spectrum consists of a background spectrum and an increment spectrum: wherein, is the increment of methane concentration relative to the background concentration; is the unit absorption characteristic spectrum; Step S1.3: Define the optimization objective function for solving the methane increment concentration: Among them, represents the serial number of the remote sensing observation sample, is the error covariance matrix of the observed spectrum, is the spatial continuity constraint term, is the physical diffusion constraint term of the wind field; Among them, the said spatial continuity constraint term adopts: Among them, is the Laplace filter kernel, is the weight parameter of the spatial continuity constraint term; is the initialized methane increment; represents performing a spatial filtering operation on the methane increment concentration at the (x, y) spatial position; The said wind field physical diffusion constraint term adopts: Among them, represents the wind vector field, is the component of the wind field in the direction, is the component of the wind field in the direction, represents the partial derivative of the initialized methane increment with respect to, is the weight parameter of the physical diffusion constraint term of the wind field.

2. The intelligent remote sensing detection method for methane emissions based on sparse spectral reconstruction according to claim 1, characterized in that The said Step S2 includes: Based on the auxiliary observation conditions of hyperspectral remote sensing data, set the simulation conditions of the atmospheric radiation transfer model; Change the initial background concentration of methane and simulate the hyperspectral radiance data at different increment concentrations; Calculate the change in radiance at each band corresponding to a unit change in methane increment to obtain the unit absorption characteristic spectrum of methane increment; Among them, the said auxiliary observation conditions based on hyperspectral remote sensing data include: the central wavelength of the satellite sensor, the full width at half maximum, the zenith angle and azimuth angle of the sun and the satellite, the atmospheric state, and the surface elevation.

3. The intelligent remote sensing detection method for methane emissions based on sparse spectral reconstruction according to claim 1, wherein, The said Step S3 includes: Step S3.1: Perform principal component dimensionality reduction processing on the hyperspectral radiance image , including data standardization, calculating the covariance matrix, performing eigenvalue decomposition on the covariance matrix, and selecting the first largest eigenvalues and eigenvectors to form a dimensionality-reduced image ; Step S3.2: Use a clustering algorithm to perform ground object clustering on the image after dimensionality reduction to obtain the class label to which each pixel belongs ; Step S3.3: Perform matrix sparse decomposition on the spectral sample set it constitutes as follows: Among them, is the left singular matrix of, containing spatial structure information; is a diagonal matrix, containing singular values; is the right singular matrix of, containing spectral band information; Step S3.4: Approximate the original matrix by selecting the top largest singular values to achieve sparsification: Among them, , and are the left singular matrix, diagonal matrix, and right singular matrix that retain the first singular values; Step S3.5: Combine the reconstructed spectral matrices of all categories to obtain a complete reconstructed hyperspectral image .

4. The intelligent remote sensing detection method for methane emissions based on sparse spectral reconstruction according to claim 1, wherein The said Step S4 includes: Step S4.1: Solve for the methane increment in the form of an analytical solution: where R represents the reconstructed background spectrum, represents the unit absorption characteristic spectrum of methane increment; Step S4.2: Calculate the error covariance matrix of the initial observation spectrum : ； Among them, N represents the number of hyperspectral bands; Step S4.3: Correct the unit absorption characteristic spectrum of methane: wherein, is a dot product operator; Step S4.4: Calculate the initial methane increment :[[]]END]] Step S4.5: Update the error covariance matrix : Step S4.6: Calculate the spatial continuity constraint term : Step S4.7: Calculate the physical diffusion constraint term of the wind field : Step S4.8: Calculate the methane increment after introducing the constraint term : Step S4.9: According to the objective function for atmospheric methane increment inversion, traverse the pixels in the hyperspectral image to obtain the atmospheric methane increment inversion result.

5. The intelligent remote sensing detection method for methane emissions based on sparse spectral reconstruction according to claim 1, characterized in that The said Step S5 includes: Step S5.1: Use large eddy simulation software, adopt the Smagorinsky SGS model, and set multi-dimensional hierarchical simulation parameters to generate an atmospheric methane increment plume sample; Among them, the said multi-dimensional hierarchical simulation parameters include: emission rate, wind speed, turbulence intensity, diffusion intensity, temperature, time step, and spatial step; Step S5.2: Based on the generated atmospheric methane increment plume sample, extract the plume morphological characteristics and construct a training sample set for machine learning estimation of methane emission rate; Among them, the training sample set for machine learning estimation of methane emission rate includes: the emission rate, wind speed, incremental concentration distribution characteristics, and plume morphology characteristics of the atmospheric methane incremental plume sample.

6. The intelligent remote sensing detection method for methane emissions based on sparse spectral reconstruction according to claim 1, characterized in that The step S6 includes: Step S6.1: Construct a machine learning model for estimating atmospheric methane emission rate; Among them, represents the emission rate data in the atmospheric methane increment plume data, represents the wind speed magnitude in the simulated methane emissions, , , and are respectively the cumulative sum, mean, standard deviation, and peak value of the methane increment concentration, represents the plume morphological characteristics; Step S6.2: Based on the constructed training sample set for machine learning estimation of methane emission rate, use machine learning techniques with regression fitting ability including random forest to train the model parameters; Step S6.3: Extract the true emission plume from the inversion result of atmospheric methane increment. The true emission plume uses the trained machine learning model for estimating atmospheric methane emission rate to obtain the emission rate estimation result corresponding to the true emission plume.

7. A remote sensing intelligent detection system for methane emissions based on sparse spectral reconstruction, characterized in that, including Module M1: Perform a first-order Taylor expansion on the Beer-Lambert law and introduce spatial continuity and wind field diffusion constraint terms to obtain the objective function for atmospheric methane increment inversion; Module M2: Simulate remote sensing observation hyperspectral radiance data under different methane increment concentrations, and calculate the unit absorption characteristic spectrum of methane increment; Module M3: Perform dimensionality reduction processing and clustering operation on the remote sensing hyperspectral image, perform matrix sparse decomposition and reconstruction on the spectral values according to the clustering result to obtain the reconstructed hyperspectral image; Module M4: Equivalent the reconstructed hyperspectral image to the background spectral image, solve the objective function for atmospheric methane increment inversion based on the background spectral image and the unit absorption characteristic spectrum of methane increment, and calculate the inversion result of atmospheric methane increment; Module M5: Generate atmospheric methane increment plume samples and construct a training sample set for machine learning estimation of methane emission rate; Module M6: Construct a machine learning model for estimating atmospheric methane emission rate, and use the training sample set to train the machine learning model for estimating atmospheric methane emission rate to obtain the trained machine learning model for estimating atmospheric methane emission rate; use the trained machine learning model for estimating atmospheric methane emission rate to complete the estimation of methane point source emission rate in the real scene based on the inversion result of atmospheric methane increment; The module M1 includes: Module M1.1: Perform a first-order Taylor expansion on the Beer-Lambert law: Among them, is the observed radiance data of the remote sensing hyperspectral sensor, is the background spectrum without the incremental absorption of methane, is the equivalent optical thickness of methane gas; Module M1.2: Define that the observed spectrum consists of the background spectrum and the incremental spectrum: wherein, is the increment of methane concentration relative to the background concentration, is the unit absorption characteristic spectrum; Module M1.3: Define the optimization objective function for solving the methane increment concentration: Among them, represents the serial number of remote sensing observation samples, is the error covariance matrix of the observed spectrum, is the spatial continuity constraint term, is the physical diffusion constraint term of the wind field; Among them, the spatial continuity constraint term adopts: Among them, is the Laplacian filter kernel, is the weight parameter of the spatial continuity constraint term; is the initialized methane increment; represents performing a spatial filtering operation on the methane increment concentration at the (x, y) spatial position; The wind field physical diffusion constraint term adopts: Among them, represents the wind vector field, is the component of the wind field in the direction, is the component of the wind field in the direction, represents the partial derivative of the initialized methane increment ; is the weight parameter of the physical diffusion constraint term of the wind field; The module M2 includes: Based on the auxiliary observation conditions of hyperspectral remote sensing data, set the simulation conditions of the atmospheric radiation transfer model; change the initial background concentration of methane and simulate the hyperspectral radiance data under different increment concentrations; calculate the change in radiance at each band corresponding to a unit change in methane increment to obtain the unit absorption characteristic spectrum of methane increment; Among them, the auxiliary observation conditions based on hyperspectral remote sensing data include: the central wavelength of the satellite sensor, the full width at half maximum, the zenith angle and azimuth angle of the sun and the satellite, the atmospheric state, and the surface elevation; The module M3 includes: Module M3.1: For the hyperspectral radiance image Perform principal component dimensionality reduction processing, including data standardization, calculating the covariance matrix, performing eigen decomposition on the covariance matrix, and selecting the first several largest eigenvalues and eigenvectors to form the dimensionality-reduced image ; Module M3.2: Use the clustering algorithm to perform ground object clustering on the dimensionality-reduced image to obtain the class label to which each pixel belongs ; Module M3.3: For the spectral sample set it composes Perform matrix sparse decomposition: Among them, is the left singular matrix of, containing spatial structure information; is a diagonal matrix, containing singular values; is the right singular matrix of, containing spectral band information; Module M3.4: Achieve sparsification by approximating the original matrix by selecting the top largest singular values: Among them, , and are the left singular matrix, diagonal matrix, and right singular matrix that retain the first singular values; Module M3.5: Merge the reconstructed spectral matrices of all categories to obtain a complete reconstructed hyperspectral image .

8. The intelligent remote sensing detection system for methane emissions based on sparse spectral reconstruction according to claim 7, characterized in that The module M4 includes: Module M4.1: Solving the methane increment in the form of an analytical solution: where R represents the reconstructed background spectrum, represents the unit absorption characteristic spectrum of methane increment; Module M4.2: Calculate the error covariance matrix of the initial observed spectrum : ； Among them, N represents the number of hyperspectral bands; Module M4.3: Calibrate the unit absorption characteristic spectrum of methane: wherein, is a dot product operator; Module M4.4: Calculate the initial methane increment : Module M4.5: Update the error covariance matrix : Module M4.6: Calculate the spatial continuity constraint term : Module M4.7: Calculate the physical diffusion constraint term of the wind field : Module M4.8: Calculate the methane increment after introducing constraint terms : Module M4.9: According to the objective function for retrieving atmospheric methane increment, traverse the pixels in the hyperspectral image to obtain the methane increment retrieval result; The said module M5 includes: Module M5.1: Use large eddy simulation software, adopt the Smagorinsky SGS model, set multi-dimensional hierarchical simulation parameters, and generate atmospheric methane increment plume samples; Among them, the multi-dimensional hierarchical simulation parameters include: emission rate, wind speed, turbulence intensity, diffusion intensity, temperature, time step, and space step; Module M5.2: Based on the generated atmospheric methane increment plume samples, extract the plume morphological features and construct a training sample set for machine learning estimation of methane emission rate; Among them, the training sample set for machine learning estimation of methane emission rate includes: the emission rate, wind speed, increment concentration distribution characteristics, and plume morphological characteristics of the atmospheric methane increment plume samples; The said module M6 includes: Module M6.1: Construct a machine learning model for estimating atmospheric methane emission rate; Among them, represents the emission rate data in the atmospheric methane increment plume data, represents the wind speed magnitude in the simulated methane emissions, , , and are respectively the cumulative sum, mean, standard deviation and peak value of the methane increment concentration, represents the plume morphological characteristics; Module M6.2: Based on the constructed training sample set for machine learning estimation of methane emission rate, use machine learning techniques with regression fitting ability including random forest for model parameter training; Module M6.3: Extract the real emission plume from the methane increment retrieval result of the atmosphere. The real emission plume uses the trained machine learning model for estimating atmospheric methane emission rate to obtain the emission rate estimation result corresponding to the real emission plume.

Images