Methods for identifying nitrogen dioxide plume signals in imaging spectral remote sensing

By preprocessing spaceborne imaging spectral data and solving the design matrix, the error problem of nitrogen dioxide plume identification in imaging spectral remote sensing was solved, and facility-scale NO2 plume identification was realized under complex surface and strong scattering conditions, providing a reliable basis for emission source location and quantification.

CN123111986APending Publication Date: 2026-11-17WUHAN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611100938.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-23
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing imaging spectral remote sensing technologies struggle to accurately identify nitrogen dioxide plumes under complex surface conditions and strong visible light scattering, easily producing false positive results and failing to meet facility-scale emission monitoring requirements.

Method used

By preprocessing the spaceborne imaging spectral data, selecting the NO2 visible light inversion window, establishing logarithmic radiance background statistics using detector columns as units, constructing a design matrix and performing weighted least squares solution, distinguishing the NO2 absorption structure from low-frequency changes such as surface reflectivity and Rayleigh scattering, and identifying the NO2 plume enhancement amount.

Benefits of technology

It achieves stable identification of facility-scale NO2 plumes under complex surface and strong scattering conditions, reduces the impact of detector response differences, lowers the false detection rate, and provides reliable data for emission source localization and quantification.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123111986A_ABST
    Figure CN123111986A_ABST
Patent Text Reader

Abstract

This invention discloses a method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing, relating to the field of atmospheric environment monitoring technology based on imaging spectral remote sensing. The identification method provided by this invention includes: acquiring and preprocessing spaceborne imaging spectral data; selecting inversion bands; constructing background samples on a per-detector-column basis, calculating the column background mean and covariance after taking the natural logarithm of the background radiance; constructing a unit absorption spectrum based on the nitrogen dioxide absorption cross section, the sensor center wavelength, and the full width at half maximum (FWHM); designing a matrix by combining the unit absorption spectrum with low-frequency spectral basis functions, and solving for the nitrogen dioxide oblique column concentration enhancement using weighted least squares; converting the nitrogen dioxide oblique column concentration enhancement into a nitrogen dioxide vertical column concentration enhancement based on the calculated air quality factor; and checking for the presence of a nitrogen dioxide plume signal in the results. This invention can reduce false detections caused by complex surface reflectance and visible light scattering variations, and is suitable for facility-scale nitrogen dioxide plume identification.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of atmospheric environment monitoring technology based on imaging spectral remote sensing, and particularly to a method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing. Background Technology

[0002] Nitrogen dioxide (NO2) is one of the most significant atmospheric pollutants, primarily originating from thermal power plants, steel mills, petrochemical plants, chemical plants, transportation, and other high-temperature combustion processes. NO2 not only irritates the respiratory system and harms human health but also participates in the formation of ozone and secondary particulate matter, significantly impacting regional air quality and the ecological environment. Therefore, conducting refined remote sensing monitoring of NO2 emission sources is of great importance for pollution prevention and control, emission regulation, environmental law enforcement, and the analysis of the causes of air pollution. In recent years, both globally and domestically, there has been increasing attention paid to NO... x Higher requirements have been placed on the emission of atmospheric pollutants such as industrial facilities. Against this backdrop, accurately identifying NO2 plumes released from key emission sources such as industrial facilities, and further achieving emission attribution and quantitative assessment, has become a crucial technical requirement in the field of atmospheric environmental monitoring. Compared to regional concentration changes, facility-scale NO2 plumes typically exhibit characteristics such as relatively fixed source locations, significant local enhancement, and spatial morphology controlled by wind fields. Therefore, they provide a foundation for conducting emission source identification, plume tracking, and emission attribution analysis.

[0003] Current NO2 satellite remote sensing observations primarily rely on dedicated atmospheric composition detection payloads, such as the Troposphere Monitor (TROPOMI) aboard the Sentinel-5P satellite. While these payloads play a crucial role in regional and global NO2 monitoring, their spatial resolution is typically at the kilometer level, which is insufficient to meet the high spatial resolution requirements of facility-scale emission monitoring. When multiple emission sources are close together or the surface environment is complex, kilometer-level pixels often mix signals from multiple sources, making accurate source attribution difficult and hindering facility-scale emission quantification.

[0004] With the development of spaceborne imaging spectrometers, remote sensing data with high spatial resolution has provided new possibilities for facility-scale NO2 plume detection. These sensors typically have spatial resolution on the order of tens of meters, enabling clearer depiction of the spatial morphology of plumes around industrial facilities, providing a more detailed observational basis for emission source identification, plume tracking, and subsequent quantitative analysis. However, NO2 primarily exhibits absorption characteristics in the visible light band, which is significantly affected by Rayleigh scattering, aerosol scattering, and variations in surface reflectivity. There is a strong coupling relationship between the nitrogen dioxide absorption signal and the surface background and atmospheric scattering. Furthermore, the complex surface types and significant differences in reflectivity around emission facilities can easily interfere with NO2 plume identification.

[0005] Existing plume identification methods, such as matched filtering methods, are mostly based on assumptions such as weak absorption perturbations and linear background superposition. They can achieve good gas detection results under conditions of relatively homogeneous surfaces and weak scattering effects, but they still have insufficient adaptability for NO2 detection under complex surface backgrounds. Under complex surface and strong scattering conditions, these methods are prone to misinterpreting changes in surface spectral structure or scattering changes as enhanced nitrogen dioxide signals, thus producing false positives and interfering with plume identification.

[0006] Therefore, there is an urgent need to develop a nitrogen dioxide plume signal identification method for imaging spectral remote sensing, so as to achieve stable identification of facility-scale NO2 plumes under complex surface conditions and non-negligible scattering, and provide technical support for refined pollution emission supervision and coordinated pollution reduction and carbon reduction. Summary of the Invention

[0007] To address the shortcomings of existing technologies in identifying nitrogen dioxide plumes under complex surface conditions and strong visible light scattering, which often result in false positives, this invention provides a method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing. This method involves preprocessing spaceborne imaging spectral data, selecting a visible light inversion window for NO2, establishing logarithmic radiance background statistics on a per-detector-column basis, constructing a design matrix using the NO2 unit absorption spectrum and low-frequency spectral basis functions, and solving for the NO2 plume enhancement using weighted least squares. This allows for the differentiation of NO2 absorption structures from low-frequency variations such as surface reflectivity and Rayleigh scattering. Specifically, this is achieved through the following techniques.

[0008] The method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing provided by the present invention includes the following steps:

[0009] Acquire spaceborne imaging spectral data of the target area, and preprocess the spaceborne imaging spectral data to obtain radiance data to be processed;

[0010] Based on the absorption characteristics of NO2 in the visible light band, inversion bands containing NO2 absorption information are selected from the radiance data to be processed to form inversion window data;

[0011] Using detector columns as units, effective pixels located in the same column as the pixel to be inverted are taken as background pixels; the natural logarithm of the radiance spectrum of the background pixels is taken, and the column background mean vector and column background covariance matrix are calculated.

[0012] Based on the NO2 absorption cross section, and the center wavelength and full width at half maximum (FWHM) of each band of the spaceborne imaging spectrometer, a unit absorption spectrum of NO2 adapted to the current sensor spectral response is constructed.

[0013] The difference between the logarithmic radiance of the pixel to be inverted and the mean vector of the column background is used as the observation vector; a design matrix composed of the NO2 unit absorption spectrum and the low-frequency spectral basis function is constructed; the nitrogen dioxide oblique column concentration enhancement is obtained by weighted least squares solution based on the column background covariance matrix;

[0014] Calculate the air quality factor corresponding to the satellite observation scene, convert the nitrogen dioxide oblique column concentration enhancement into nitrogen dioxide vertical column concentration enhancement, and check whether it contains a nitrogen dioxide plume.

[0015] Furthermore, the spaceborne imaging spectral data includes the radiance spectrum, center wavelength, full width at half maximum (FWHM), imaging geometry, geographic coordinates, solar geometry, and surface elevation for each pixel.

[0016] Furthermore, the preprocessing includes radiometric correction, geometric correction, outlier removal, bad band removal, cloud pixel masking, and water body pixel labeling.

[0017] The cloud pixel shielding method is as follows: For cloud pixels, a cloud mask is generated based on the apparent reflectance relationship of blue, green, red, near-infrared, short-wave infrared and the band near 1.38μm.

[0018] The method for water body pixel identification is as follows: for water body pixels, a water body mask is generated based on the reflectance relationship between green light, near-infrared and short-wave infrared.

[0019] Furthermore, the inversion band is 400 nm-500 nm.

[0020] Furthermore, the method for calculating the column background mean vector and column background covariance matrix by taking the natural logarithm of the radiance spectrum of the background pixels is as follows:

[0021] The radiance spectrum L of the background pixel b If we consider it as a multivariate log-normal distribution, then ln(L b It follows the following multivariate Gaussian distribution; ;

[0022] The column background mean vector μ b The background covariance matrix Σ b The statistics by column are as follows: ;

[0023] ;

[0024] Among them, L b (n) represents the radiance spectrum of the nth background pixel, N represents the number of effective background pixels in the same detector array, and T represents the transpose of the matrix.

[0025] Furthermore, based on the NO2 absorption cross section, and the center wavelength and full width at half maximum (FWHM) of each band, the method for constructing a NO2 unit absorption spectrum adapted to the current sensor's spectral response is as follows:

[0026] Read high-resolution NO2 absorption cross section;

[0027] Based on the center wavelength and full width at half maximum (FWHM) of each band, construct the spectral response function;

[0028] The NO2 absorption cross section and the spectral response function are convolved and resampled to obtain the NO2 unit absorption spectrum k.

[0029] Furthermore, based on the center wavelength λ of the i-th band of the imaging spectrometer... c,i and spectral width parameter f i The constructed Gaussian spectral response function SRF corresponding to the i-th band. i (λ) is shown in the following formula; ;

[0030] Where λ represents wavelength, and C is the width conversion coefficient of the Gaussian spectral response function, expressed as: .

[0031] Specifically, the NO2 absorption cross section σ(λ) and the spectral response function SRF are compared. i The method for obtaining the unit absorption spectrum k of NO2 by performing convolution and resampling (λ) is as follows:

[0032] The NO2 absorption cross section σ(λ) and the spectral response function SRF are compared. i (λ) Resample to the same wavelength grid;

[0033] Based on the spectral response function SRF i (λ) Taking a weighted integral over the NO2 absorption cross section σ(λ), we obtain the NO2 unit absorption spectrum k corresponding to the i-th band. i ; ;

[0034] Where, k i The unit absorption spectrum of NO2 corresponding to the i-th band;

[0035] Repeat the above steps for each band within the target wavelength range to obtain the NO2 unit absorption spectrum k that matches each band of the spaceborne imaging spectrometer.

[0036] Furthermore, the difference between the logarithmic radiance of the pixel to be inverted and the mean vector of the column background is used as the observation vector; a design matrix composed of the NO2 unit absorption spectrum and low-frequency spectral basis functions is constructed; the method for obtaining the NO2 plume enhancement amount is to perform a weighted least squares solution on the nitrogen dioxide oblique column concentration enhancement amount based on the column background covariance matrix.

[0037] When there is NO2 plume enhancement in the pixel to be inverted, according to the Lambert-Beer law, the radiance spectrum L of the pixel to be inverted... t Represented as:

[0038] ;

[0039] Among them, L b is the radiance spectrum corresponding to the background density pixel; r is the equivalent reflectance vector, used to describe the surface reflectance and low-frequency scattering variation of the pixel to be inverted relative to the column background pixels;

[0040] L t It is also considered to conform to a log-normal distribution, that is:

[0041] ;

[0042] ;

[0043] ;

[0044] Where k is the unit absorption spectrum of NO2, α is the concentration enhancement of nitrogen dioxide on the oblique column, and μ t Let μ be the mean vector corresponding to the pixel to be inverted. b Σ is the column background mean vector. t Let Σ be the covariance matrix corresponding to the pixel to be inverted. b The background covariance matrix is ​​shown below.

[0045] Using Legendre polynomial basis functions as the basis functions for the low-frequency spectrum, the wavelengths within the inversion window are normalized to [-1, 1]. The residual model V is:

[0046] ;

[0047] Among them, P d (λ) is the d-th order Legendre polynomial, d=0,1,2,...,D; a d The corresponding coefficients are D; D is the highest order of the Legendre polynomial.

[0048] The residual model V can be written in matrix form as follows:

[0049] ;

[0050] ;

[0051] ;

[0052] ;

[0053] Where, P0(λ) to P D (λ) represents the Legendre polynomials of order 0 to D;

[0054] Based on weighted least squares, state vector The optimal estimate is:

[0055] ;

[0056] State vector The first component is the increase in NO2 concentration along the effective optical path by the α-column, and the remaining components are the coefficients of the low-frequency spectral terms.

[0057] Furthermore, the method for converting the nitrogen dioxide oblique column concentration enhancement amount into the nitrogen dioxide vertical column concentration enhancement amount is as follows:

[0058] The air quality factor (AMF) is calculated based on the imaging geometry, geographic coordinates, solar geometry, and surface elevation. The nitrogen dioxide oblique column concentration enhancement (DSCD) is converted into the nitrogen dioxide vertical column concentration enhancement (DVCD) according to the following formula.

[0059] .

[0060] Compared with the prior art, the advantages of the present invention are:

[0061] 1. By utilizing the spatial resolution of imaging spectrometers on the order of tens of meters, facility-scale identification can be performed on NO2 plumes near point sources such as industrial facilities.

[0062] 2. Establishing background statistics on a detector array basis can reduce the impact of cross-track detector response differences in imaging spectrometers on the inversion results.

[0063] 3. By jointly inverting the NO2 absorption term and the low-frequency polynomial term in the logarithmic radiance space, the NO2 absorption structure can be distinguished from low-frequency spectral changes such as surface reflectance and Rayleigh scattering, thus reducing false detections under complex surfaces.

[0064] 4. By iteratively removing background, constraining standard error, and filtering connected components, a more continuous and reliable NO2 plume region can be obtained, providing basic data for emission source location and emission quantification. Attached Figure Description

[0065] Figure 1 Example of RGB image of the target area from spaceborne imaging spectral data.

[0066] Figure 2 Here is a schematic diagram of the radiance sensitivity of the imaging spectrometer to the absorption windows of NO2, CO2, and CH4.

[0067] Figure 3 Here is a schematic diagram of the nitrogen dioxide absorption cross section and the unit absorption spectrum after convolution into the sensor's spectral response.

[0068] Figure 4 Example of a nitrogen dioxide differential oblique column concentration enhancement plot obtained by the method of the present invention.

[0069] Figure 5 Example of a nitrogen dioxide differential oblique column concentration enhancement plot in latitude and longitude coordinates. Detailed Implementation

[0070] The technical solution of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] To illustrate the method of this invention, the general form of column-level statistical inversion is first given. The specific steps are as follows:

[0072] 1. In the traditional column-level matching concept, the background spectrum is represented by the following multivariate Gaussian distribution.

[0073] .

[0074] Column background mean vector μ b The background covariance matrix Σ b The statistics are presented in the following formulas, categorized by column.

[0075] ;

[0076] ;

[0077] Among them, L b L represents the radiance spectrum of the background pixels. b (n) represents the radiance spectrum of the nth background pixel; N represents the total number of effective background pixels in the same detector array; T represents the transpose of the matrix.

[0078] 2. When gas enhancement is present, according to the Beer-Lambert law, the radiance spectrum L of the pixel to be inverted... t Written as:

[0079] ;

[0080] Where D(v) represents the vector v (v is e) -kα The diagonal matrix is ​​constructed, where k is the NO2 unit absorption spectrum and α is the concentration enhancement of the nitrogen dioxide oblique column.

[0081] The element in the i-th row and j-th column of D(v) is represented as:

[0082] ;

[0083] Where, δ ij Let Kronecker function be used.

[0084] Under the above model, the target spectrum satisfies:

[0085] ;

[0086] in, , .

[0087] If the first-order Taylor approximation is used, the residual is expressed as:

[0088] ;

[0089] The estimated increase α in nitrogen dioxide concentration along the oblique column was obtained using weighted least squares. :

[0090] ;

[0091] The above approach is suitable for scenarios with weak absorption, relatively uniform surface, and weak scattering effects.

[0092] Example

[0093] For the visible light NO2 plume processed by this invention, significant low-frequency spectral changes occur due to surface reflectivity and Rayleigh scattering. Therefore, directly using the above method can easily lead to false detections. To solve this problem, the method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing provided in this embodiment establishes background statistics in the logarithmic radiance space and introduces low-frequency spectral basis functions. The specific steps are as follows.

[0094] Step 1: Acquire the spaceborne imaging spectral data of the target area, preprocess the spaceborne imaging spectral data to obtain the radiance data to be processed.

[0095] 1. Acquire spaceborne imaging spectral data of the target area.

[0096] like Figure 1As shown, this embodiment first reads the spaceborne imaging spectral data, such as... Figure 1 As shown. These spaceborne imaging spectroscopic data can come from PRISMA, EnMAP, EMIT, GF-5, ZY-1, or other spaceborne imaging spectrometers with a continuous band from visible light to shortwave infrared.

[0097] The spectral data read from the spaceborne imaging system includes the following for each pixel: radiance spectrum, center wavelength, full width at half maximum (FWHM), imaging geometry (i.e., observation zenith angle, relative azimuth angle, and sensor height), geographic coordinates, solar geometry (i.e., solar zenith angle), and surface elevation.

[0098] 2. Preprocess the spaceborne imaging spectral data to obtain the radiance data to be processed.

[0099] The preprocessing process includes radiometric correction, geometric correction, outlier removal, bad band removal, cloud pixel masking, and water body pixel labeling.

[0100] (1) Radiation correction.

[0101] The DN value recorded by the sensor is converted into radiance, apparent reflectance, or other physical quantities at the sensor's entrance pupil.

[0102] (2) Geometric correction.

[0103] Based on the RPC parameters provided by the satellite, the latitude and longitude of each pixel of the satellite are accurately calculated.

[0104] (3) Outlier removal and bad band removal.

[0105] Invalid values ​​(or bad bands), non-positive radiance, bad bands, and obviously abnormal pixels in the spaceborne imaging spectral data are removed or set to null values.

[0106] (4) Cloud pixel shielding.

[0107] For cloud pixels, a cloud mask is generated based on the apparent reflectance relationships of blue, green, red, near-infrared, short-wave infrared, and the band near 1.38 μm.

[0108] (5) Water body pixel identifier.

[0109] For water pixels, a water mask is generated based on the reflectance relationship between green light, near-infrared, and short-wave infrared.

[0110] Step 2: Based on the absorption characteristics of NO2 in the visible light band, select the inversion band containing NO2 absorption information from the radiance data cube to be processed to form inversion window data.

[0111] like Figure 2As shown, NO2 can cause detectable radiance changes in the range of 400 nm to 500 nm. Therefore, this embodiment uses the above range as the inversion window; preferably, several continuous bands of 400-500 nm.

[0112] Compared to the shortwave infrared gas inversion window, the Rayleigh scattering and surface reflectance changes in the visible light window for NO2 in this embodiment cannot be ignored, and the low-frequency radiation component needs to be explicitly processed in the inversion model.

[0113] Step 3: Using detector columns as units, select effective pixels located in the same column as the pixel to be inverted as background pixels, and analyze the radiance spectrum L of the background pixels. b Take the natural logarithm and calculate the column background mean vector μ. b The background covariance matrix Σ b .

[0114] This embodiment uses the radiance spectrum L of the background pixel. b It is considered as a multivariate log-normal distribution, i.e., ln(L b It follows a multivariate Gaussian distribution.

[0115] ;

[0116] Then the background mean vector μ b The background covariance matrix Σ b The statistics are presented in the following formulas, categorized by column.

[0117] ;

[0118] .

[0119] Step 4: Based on the NO2 absorption cross section and the center wavelength and full width at half maximum (FWHM) of each band of the spaceborne imaging spectrometer (i.e., all bands within the 400-500 nm wavelength range), construct the NO2 unit absorption spectrum k that is adapted to the current sensor spectral response.

[0120] In this step, the NO2 unit absorption spectrum k is obtained by convolving the NO2 absorption cross section with the spectral response function of the spaceborne imaging spectrometer. The spectral response function is determined by the center wavelength and the full width at half maximum (FWHM).

[0121] Specifically, the methods for constructing the NO2 unit absorption spectrum k include:

[0122] 1. Read the high-resolution NO2 absorption cross section. The NO2 absorption cross section at 220 K is preferred.

[0123] 2. Construct spectral response functions based on the center wavelength and full width at half maximum (FWHM) of each band of the spaceborne imaging spectrometer.

[0124] Specifically, based on the center wavelength λ of the i-th band of the imaging spectrometer c,i and spectral width parameter f i The constructed Gaussian spectral response function SRF corresponding to the i-th band. i (λ) is shown in the following formula:

[0125] ;

[0126] Where λ represents wavelength; C is the width conversion coefficient of the Gaussian spectral response function, expressed as: .

[0127] 3. The NO2 absorption cross section σ(λ) and the spectral response function SRF are compared. i The method for obtaining the unit absorption spectrum k of NO2 by performing convolution and resampling (λ) is as follows:

[0128] The NO2 absorption cross section σ(λ) and the spectral response function SRF are compared. i (λ) Resample to the same wavelength grid;

[0129] Based on the spectral response function SRF i (λ) Taking a weighted integral over the NO2 absorption cross section σ(λ), we obtain the NO2 unit absorption spectrum k corresponding to the i-th band. i :

[0130] ;

[0131] Where, k i The unit absorption spectrum of NO2 corresponding to the i-th band;

[0132] Repeat the above steps for each band within the target wavelength range to obtain the NO2 unit absorption spectrum k that matches each band of the spaceborne imaging spectrometer.

[0133] Taking the ENMAP satellite as an example, the unit absorption spectrum k after convolving the NO2 absorption cross section with the spectral response function is as follows: Figure 3 As shown.

[0134] Step 5: Construct a residual model and use weighted least squares to solve for the increase in nitrogen dioxide oblique column concentration to obtain the increase in nitrogen dioxide oblique column concentration.

[0135] 1. Construct a design matrix A consisting of the NO2 unit absorption spectrum k and low-frequency spectral basis functions.

[0136] Low-frequency spectral basis functions are used to characterize smooth spectral components caused by variations in surface reflectance, Rayleigh scattering, and aerosol scattering.

[0137] When NO2 plume enhancement exists in the pixel to be inverted, the Lambert-Beer law is introduced, and the radiance spectrum L of the pixel to be inverted is... t Represented as:

[0138] ;

[0139] Among them, L b is the radiance spectrum corresponding to the background density pixel; r is the equivalent reflectance vector, used to describe the surface reflectance and low-frequency scattering variation of the pixel to be inverted relative to the column background pixels;

[0140] L t It is also considered to conform to a log-normal distribution, that is:

[0141] ;

[0142] ;

[0143] ;

[0144] Where k is the unit absorption spectrum of NO2, α is the concentration enhancement of nitrogen dioxide on the oblique column, and μ t Let μ be the mean vector corresponding to the pixel to be inverted. b Σ is the column background mean vector. t Let Σ be the covariance matrix corresponding to the pixel to be inverted. b The background covariance matrix is ​​shown below.

[0145] Since the surface reflectance difference, Rayleigh scattering and aerosol scattering represented by r usually change with low frequency with wavelength, this embodiment models ln(r) as a polynomial expansion.

[0146] Legendre polynomial basis functions are used as low-frequency spectral basis functions, and the wavelengths within the inversion window are normalized to [-1, 1] before constructing the Legendre polynomial basis functions. The residual model V is then:

[0147] ;

[0148] Among them, P d (λ) is the d-th order Legendre polynomial, d=0,1,2,...,D; a d For P d (λ) represents the coefficients; D is the highest order of the Legendre polynomial, and D can preferably be 3; in cases of complex terrain or where the number of bands allows, D can also be adaptively adjusted according to the inversion window width and the fitting residual.

[0149] Rewriting the above equation in matrix form, we get:

[0150] ;

[0151] ;

[0152] ;

[0153] ;

[0154] Right now, ;

[0155] .

[0156] Where, P0(λ) to P D (λ) represents the Legendre polynomials of order 0 to D; n b This represents the number of effective bands within the inversion window.

[0157] 2. Based on the column background covariance matrix Σ b The NO2 plume enhancement was obtained by weighted least squares solution of the nitrogen dioxide oblique column concentration enhancement.

[0158] Based on weighted least squares, state vector Optimal estimate for:

[0159] ;

[0160] State vector The first component is the increase in nitrogen dioxide concentration in the inclined column.

[0161] By simultaneously solving the α and Legendre polynomials, this invention can separate the NO2 absorption structure from the smooth spectral variations caused by surface reflectance and scattering.

[0162] Step 6: Calculate the air quality factor corresponding to the satellite observation scene, convert the nitrogen dioxide oblique column concentration enhancement into nitrogen dioxide vertical column concentration enhancement, and check whether it contains a nitrogen dioxide plume.

[0163] 1. Quality control of NO2 plume enhancement.

[0164] After the inversion is completed, based on the inversion residuals The variance of the fitting residuals is calculated to obtain the standard error of the NO2 plume enhancement, and pixels with enhancement greater than the standard error are retained.

[0165] Unreliable pixels are eliminated using low radiance masks (pixels with radiance less than 0 are uniformly defined as unreliable pixels), cloud masks, and water body masks.

[0166] 2. Calculate the Air Quality Factor (AMF) using the target area's latitude, longitude, month, solar zenith angle, observed zenith angle, surface elevation, and assumptions about the vertical distribution of NO2 in the near-surface layer. The AMF can be obtained through a radiative transfer model or lookup table based on the target area's latitude, longitude, month, solar zenith angle, observed zenith angle, surface elevation, and assumptions about the vertical distribution of NO2 in the near-surface layer.

[0167] 3. For the NO2 oblique column concentration enhancement amount DSCD, this embodiment can use the air quality factor AMF to convert the NO2 oblique column concentration enhancement amount DSCD into the NO2 vertical column concentration enhancement amount DVCD according to the following formula.

[0168] ;

[0169] like Figure 4 and Figure 5 As shown, the present invention ultimately outputs NO2 differential oblique column concentration enhancement maps, NO2 differential vertical column concentration enhancement maps, plume binary masks, cloud masks, water body masks, low radiance masks, and raster files with geographic coordinates. These outputs can be used for facility-scale NO2 plume identification, emission source location, emission attribution, and subsequent emission flux estimation.

[0170] Example 2

[0171] Based on Example 1, this embodiment further includes an iterative background update process in step five, specifically as follows:

[0172] 1. Perform an initial inversion to obtain the initial NO2 enhanced image.

[0173] 2. Perform smoothing and robust thresholding on the initial NO2 enhanced image to extract a coarse plume mask.

[0174] 3. Remove pixels corresponding to the coarse feather mask from the background pixels and recalculate the column background mean vector μ. b The background covariance matrix Σ b And then it happens again.

[0175] The above iterative method can avoid significant background statistics contamination from plume pixels.

[0176] The above detailed embodiments describe the implementation of the present invention; however, the present invention is not limited to the specific details described in the above embodiments. Within the scope of the claims and technical concept of the present invention, various simple modifications and changes can be made to the technical solution of the present invention, and these simple modifications all fall within the protection scope of the present invention.

Claims

1. A method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing, characterized in that, Includes the following steps: Acquire spaceborne imaging spectral data of the target area, and preprocess the spaceborne imaging spectral data to obtain radiance data to be processed; Based on the absorption characteristics of NO2 in the visible light band, continuous inversion bands containing NO2 absorption information are selected from the radiance data to be processed to form inversion window data; Using detector columns as units, effective pixels located in the same column as the pixel to be inverted are taken as background pixels; Take the natural logarithm of the radiance spectrum of the background pixels, and calculate the column background mean vector and column background covariance matrix; Based on the NO2 absorption cross section, and the center wavelength and full width at half maximum (FWHM) of each band of the spaceborne imaging spectrometer, a NO2 unit absorption spectrum adapted to the current sensor spectral response is constructed. A residual model was constructed, and a weighted least squares solution was used to obtain the increase in nitrogen dioxide oblique column concentration. Calculate the air quality factor corresponding to the satellite observation scene, convert the nitrogen dioxide oblique column concentration enhancement into nitrogen dioxide vertical column concentration enhancement, and check whether it contains a nitrogen dioxide plume.

2. The method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing according to claim 1, characterized in that, The spaceborne imaging spectral data includes the radiance spectrum, center wavelength, full width at half maximum (FWHM), imaging geometry, geographic coordinates, solar geometry, and surface elevation for each pixel.

3. The method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing according to claim 1, characterized in that, The preprocessing includes radiometric correction, geometric correction, outlier removal, bad band removal, cloud pixel masking, and water body pixel labeling.

4. The method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing according to claim 1, characterized in that, The inversion band is 400 nm-500 nm.

5. The method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing according to claim 1, characterized in that, The method for calculating the column background mean vector and column background covariance matrix by taking the natural logarithm of the radiance spectrum of the background pixels is as follows: The radiance spectrum L of the background pixel b If we consider it as a multivariate log-normal distribution, then ln(L b It follows the following multivariate Gaussian distribution; ; The column background mean vector μ b The background covariance matrix Σ b The statistics by column are as follows: ; ; Among them, L b (n) represents the radiance spectrum of the nth background pixel, N represents the number of effective background pixels in the same detector array, and T represents the transpose of the matrix.

6. The method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing according to claim 1, characterized in that, Based on the NO2 absorption cross section, and the center wavelength and full width at half maximum (FWHM) of each band, the method for constructing a unit absorption spectrum of NO2 that adapts to the current sensor's spectral response is as follows: Read high-resolution NO2 absorption cross section; Based on the center wavelength and full width at half maximum (FWHM) of each band, construct the spectral response function; The NO2 absorption cross section and the spectral response function are convolved and resampled to obtain the NO2 unit absorption spectrum k.

7. The method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing according to claim 6, characterized in that, Based on the center wavelength λ of the i-th band of the imaging spectrometer c,i And the Gaussian spectral response function SRF corresponding to the i-th band is constructed using the spectral width parameter fi. i (λ) is shown in the following formula: ; Where λ represents wavelength; C is the width conversion coefficient of the Gaussian spectral response function, expressed as: .

8. The method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing according to claim 7, characterized in that, The NO2 absorption cross section σ(λ) and the spectral response function SRF are compared. i The method for obtaining the unit absorption spectrum k of NO2 by performing convolution and resampling (λ) is as follows: The NO2 absorption cross section σ(λ) and the spectral response function SRF are compared. i (λ) Resample to the same wavelength grid; Based on the spectral response function SRF i (λ) Taking a weighted integral over the NO2 absorption cross section σ(λ), we obtain the NO2 unit absorption spectrum k corresponding to the i-th band. i : ; Where, k i The unit absorption spectrum of NO2 corresponding to the i-th band; Repeat the above steps for each band within the target wavelength range to obtain the NO2 unit absorption spectrum k that matches each band of the spaceborne imaging spectrometer.

9. The method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing according to claim 1, characterized in that, The method for constructing a residual model and solving for the increase in nitrogen dioxide oblique column concentration using weighted least squares is as follows: When an enhanced NO2 plume exists in the pixel to be inverted, according to the Lambert-Beer law, the radiance spectrum L of the pixel to be inverted... t Represented as: ; Among them, L b is the radiance spectrum corresponding to the background density pixel; r is the equivalent reflectance vector, used to describe the surface reflectance and low-frequency scattering variation of the pixel to be inverted relative to the column background pixels; L t It is also considered to conform to a log-normal distribution, that is: ; ; Where k is the unit absorption spectrum of NO2, α is the concentration enhancement of nitrogen dioxide on the oblique column, and μ t Let μ be the mean vector corresponding to the pixel to be inverted. b Σ is the column background mean vector. t Let Σ be the covariance matrix corresponding to the pixel to be inverted. b The background covariance matrix is ​​shown below. Using Legendre polynomial basis functions as the basis functions for the low-frequency spectrum, the wavelengths within the inversion window are normalized to [-1, 1]. The residual model V is: ; Among them, P d (λ) is the d-th order Legendre polynomial, d=0,1,2,...,D; a d The corresponding coefficients are D; D is the highest order of the Legendre polynomial. The residual model V can be written in matrix form as follows: ; ; ; ; Where, P0(λ) to P D (λ) represents the Legendre polynomials of order 0 to D; Based on weighted least squares, state vector Optimal estimate for: ; State vector The first component is the increase in the concentration of nitrogen dioxide along the effective optical path of the oblique column.

10. The method for identifying nitrogen dioxide plume signals in imaging spectral remote sensing according to claim 9, characterized in that, The method for converting the nitrogen dioxide oblique column concentration enhancement amount into the nitrogen dioxide vertical column concentration enhancement amount is as follows: Calculate the Air Quality Factor (AMF) and convert the nitrogen dioxide oblique column concentration enhancement amount DSCD to the nitrogen dioxide vertical column concentration enhancement amount DVCD according to the following formula; 。