Infrared Hyperspectral Atmospheric Sounder Noise Analysis Method and Device

Abstract

The present invention provides a method and device for noise analysis of an infrared hyperspectral atmospheric sounder, which relates to the technical field of remote sensing data analysis. The method includes: acquiring an interferogram formed by the infrared hyperspectral atmospheric sounder observing the earth target; generating target spectral data corresponding to the interferogram; performing denoising processing on the target spectral data to obtain reconstructed spectral data; determining the noise equivalent radiance variance according to the target spectral data and the reconstructed spectral data, and the noise equivalent radiance variance characterizes the noise characteristics of the infrared hyperspectral atmospheric sounder. The technical solution of the present invention provides a new method for characterizing the instrument noise of an infrared hyperspectral atmospheric sounder, which can characterize the instrument noise of the infrared hyperspectral atmospheric sounder in real time.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing data analysis, and particularly to a method and device for analyzing the noise of an infrared hyperspectral atmospheric sounder. Background Art

[0002] The Hyperspectral Infrared Atmospheric Sounder (HIRAS) is a hyperspectral remote sensor in the infrared spectral band. It is a space-based hyperspectral resolution infrared observation instrument based on Fourier interferometry technology, which can detect the atmospheric temperature profile, detect the atmospheric humidity profile, and invert the vertical distribution profile of trace gases, etc., and plays an important role in meteorological monitoring, air quality monitoring, etc.

[0003] As one of the main payloads of meteorological satellites, analyzing the noise of HIRAS is an important content for evaluating its on-orbit operation performance. In the related art, by repeated observations, spectral sets of a blackbody and cold space are obtained, and the instrument noise is calculated therefrom to realize the estimation and characterization of the HIRAS instrument noise. Since this method requires target observations of a blackbody and cold space, there are certain spatio-temporal factor limitations, it is difficult to characterize the instrument noise of HIRAS in real time, and it affects the evaluation of the on-orbit operation performance of HIRAS. Summary of the Invention

[0004] The present invention provides a method and device for analyzing the noise of an infrared hyperspectral atmospheric sounder to solve the defect of insufficient real-time performance in the evaluation of HIRAS noise in the prior art.

[0005] The present invention provides a method for analyzing the noise of an infrared hyperspectral atmospheric sounder, including:

[0006] Obtaining an interferogram formed by the infrared hyperspectral atmospheric sounder observing the earth target;

[0007] Generating target spectral data corresponding to the interferogram;

[0008] Performing denoising processing on the target spectral data to obtain reconstructed spectral data;

[0009] Determining the noise equivalent radiance variance according to the target spectral data and the reconstructed spectral data, and the noise equivalent radiance variance characterizes the noise characteristics of the infrared hyperspectral atmospheric sounder.

[0010] According to the method for analyzing the noise of an infrared hyperspectral atmospheric sounder provided by the present invention, the performing denoising processing on the target spectral data to obtain reconstructed spectral data includes:

[0011] Obtaining the difference between the target spectral data and the spectral mean value to obtain a spectral difference;

[0012] Determine the covariance matrix of the target spectral data based on the spectral difference;

[0013] Determine the noise prior matrix according to the covariance matrix of the target spectral data;

[0014] Normalize the spectral difference based on the noise prior matrix to obtain a normalized spectral difference;

[0015] Based on the normalized spectral difference, use the principal component analysis method to remove the noise in the target spectral data to obtain reconstructed spectral data.

[0016] According to a noise analysis method for an infrared hyperspectral atmospheric detector provided by the present invention, the step of using the principal component analysis method to remove the noise in the target spectral data based on the normalized spectral difference to obtain reconstructed spectral data includes:

[0017] Determine the principal component components and principal component weights based on the normalized spectral difference, and the principal component components are used to characterize the non-noise data in the target spectral data;

[0018] Reconstruct the spectral difference according to the principal component components, the principal component weights, and the noise prior matrix to obtain a target spectral difference;

[0019] Determine the reconstructed spectral data based on the target spectral difference and the spectral mean.

[0020] According to a noise analysis method for an infrared hyperspectral atmospheric detector provided by the present invention, the step of determining the principal component components and principal component weights based on the normalized spectral difference includes:

[0021] Determine the covariance matrix of the normalized spectral difference to obtain a target covariance matrix;

[0022] Perform singular value decomposition on the target covariance matrix to obtain orthogonal basis eigenvectors and eigenvalues;

[0023] Determine the principal component weights according to the orthogonal basis eigenvectors and the normalized spectral difference;

[0024] Determine the principal component components based on the change curve of the eigenvalues.

[0025] According to a noise analysis method for an infrared hyperspectral atmospheric detector provided by the present invention, the step of generating the target spectral data corresponding to the interferogram includes:

[0026] Perform Fourier transform on the interferogram to obtain initial spectral data;

[0027] Determine the mean value of the initial spectral data;

[0028] Screen the initial spectral data based on the mean value and the set threshold to obtain the target spectral data corresponding to the interferogram.

[0029] According to a method for analyzing the noise of an infrared hyperspectral atmospheric detector provided by the present invention, determining the noise equivalent radiation variance according to the target spectral data and the reconstructed spectral data includes:

[0030] Obtain the spectral residuals of the target spectral data and the reconstructed spectral data;

[0031] Determine the noise covariance matrix based on the spectral residuals;

[0032] Determine the noise equivalent radiation variance according to the noise covariance matrix.

[0033] The present invention also provides a device for analyzing the noise of an infrared hyperspectral atmospheric detector, including:

[0034] An acquisition module, configured to acquire an interferogram formed by the infrared hyperspectral atmospheric detector observing the earth target;

[0035] A generation module, configured to generate the target spectral data corresponding to the interferogram;

[0036] A processing module, configured to perform denoising processing on the target spectral data to obtain reconstructed spectral data;

[0037] A determination module, configured to determine the noise equivalent radiation variance according to the target spectral data and the reconstructed spectral data, and the noise equivalent radiation variance characterizes the noise characteristics of the infrared hyperspectral atmospheric detector.

[0038] The present invention also provides an infrared hyperspectral atmospheric detector, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the method for analyzing the noise of the infrared hyperspectral atmospheric detector as described in any one of the above is implemented.

[0039] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for analyzing the noise of the infrared hyperspectral atmospheric detector as described in any one of the above is implemented.

[0040] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, the method for analyzing the noise of the infrared hyperspectral atmospheric detector as described in any one of the above is implemented.

[0041] The noise analysis method and device for an infrared hyperspectral atmospheric sounder provided by the present invention obtain an interferogram formed by the infrared hyperspectral atmospheric sounder observing the Earth's target, generate target spectral data corresponding to the interferogram, then perform denoising processing on the target spectral data to obtain reconstructed spectral data. The reconstructed spectral data can be used as a simulated value of the spectrum of the Earth's target. Then, the noise equivalent radiation variance is determined based on the target spectral data and the reconstructed spectral data, and the noise equivalent radiation variance can be used to characterize the noise characteristics of the infrared hyperspectral atmospheric sounder. In this way, it is possible to characterize the instrument noise of the infrared hyperspectral atmospheric sounder by using the target spectral data formed by observing the Earth's target with the infrared hyperspectral atmospheric sounder, avoiding the spatio-temporal factor limitations when estimating the instrument noise by relying on a blackbody and cold air, and improving the real-time performance of instrument noise characterization. Description of the Drawings

[0042] To more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0043] Figure 1 is a schematic flowchart of the noise analysis method for an infrared hyperspectral atmospheric sounder provided by an embodiment of the present invention;

[0044] Figure 2 is a schematic diagram of the HIRAS observation range provided by an embodiment of the present invention;

[0045] Figure 3 is a schematic diagram of the change curve of the eigenvalues obtained by singular value decomposition in an embodiment of the present invention;

[0046] Figure 4 is a schematic diagram of the initial spectrogram in an embodiment of the present invention;

[0047] Figure 5 is a schematic diagram of the screening result of the initial spectral data in an embodiment of the present invention;

[0048] Figure 6 is a schematic diagram of the result of the noise equivalent radiation variance in an embodiment of the present invention;

[0049] Figure 7 is a schematic diagram of the result of the noise equivalent temperature difference in an embodiment of the present invention;

[0050] Figure 8 is a schematic structural diagram of the noise analysis device for an infrared hyperspectral atmospheric sounder provided by an embodiment of the present invention;

[0051] Figure 9It is a schematic structural diagram of the infrared hyperspectral atmospheric sounder provided by the embodiments of the present invention. Specific embodiments

[0052] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without making creative efforts fall within the protection scope of the present invention.

[0053] The radiation noise of remote sensing instruments is an important influencing factor for estimating geophysical parameters. Characterizing the radiation noise of remote sensing instruments is an important step in making full use of relevant observation data to estimate geophysical parameters. HIRAS is one of the main payloads of polar-orbiting meteorological satellites (such as the FY-3 polar-orbiting meteorological satellite). It is a space-based hyperspectral resolution infrared observation instrument based on Fourier interferometry technology. It can achieve hyperspectral resolution infrared observations of the earth-atmosphere system, can accurately detect the atmospheric temperature profile and atmospheric humidity profile with relatively high vertical resolution. By assimilating this information into numerical weather prediction models, the accuracy of numerical weather prediction can be improved. At the same time, it can retrieve the vertical distribution profiles of trace gases, such as trace gases like ozone, carbon monoxide, methane, etc., and has played a huge role in air quality monitoring. Moreover, it can provide long-term continuous records of atmospheric components for later interannual change research.

[0054] HIRAS is a hyperspectral remote sensing instrument in the infrared spectral band and is a stable meteorological satellite payload. The spectral detection range can be from 650 cm -1 to 2250 cm -1 , and the spatial resolution can reach 16 km. Since the improvement of spectral resolution and the increase in the number of spectral channels are beneficial to optimizing the vertical detection resolution, but at the same time will lead to a decrease in the channel signal-to-noise ratio. Therefore, analyzing the instrument noise of HIRAS is an important part of evaluating its on-orbit operation performance. In related technologies, the instrument noise evaluation method of remote sensing instruments mainly realizes the characterization and estimation of instrument noise by relying on the observation data of two points, namely a blackbody and cold space. Since it is necessary to conduct target observations at two points, namely a blackbody and cold space, there are certain spatio-temporal factor limitations.

[0055] During its on-orbit operation, HIRAS has three working modes: the initial mode after launch into orbit, the earth scanning mode, and the calibration mode. Among them, in the initial mode after launch into orbit, on-orbit optical calibration of the instrument is carried out through operations such as unlocking the moving mirror of the interferometer; in the earth scanning mode, several pixel fields of view (Field Of View, FOV) are used to perform cross-track scanning through a 45° scanning mirror to observe 29 fields of regard (FOR); in the calibration mode, calibration is achieved by scanning the blackbody and cold space, and during each scan, the blackbody and cold space are scanned twice for radiometric calibration. The scanned data in the calibration mode is not only the calibration basis for the spectra of earth targets but also the main technology for instrument noise analysis.

[0056] In related technologies, in the calibration mode, it is possible to first collect the scanned observation data of the blackbody and cold space in the flight orbit of the meteorological satellite, that is, use the original complex spectral data obtained by Fourier transforming the voltage sampling interferogram of the blackbody and cold space as sample data; then calculate the calibration spectra of each sample data; and then calculate the Noise Equivalent Differential Radiance (NEdN) of the blackbody and cold space based on the calibration spectrum results, and use this NEdN to characterize the instrument noise. Specifically, the average value of the sample data can be calculated, and combined with Planck's formula, the ideal blackbody radiation spectrum corresponding to the blackbody temperature is deduced, and the calibration spectra of each sample data are calculated using the following formula (1), and then the standard deviation is calculated based on the calibration spectrum results, and the NEdN of the blackbody and cold space is obtained using the following formula (2).

[0057] Formula (1) is

[0058] Formula (2) is

[0059] Among them, represents the cold space calibration spectrum, represents the cold space original complex spectral sample data, represents the blackbody original complex spectral sample data, represents the ideal blackbody radiation spectrum corresponding to the blackbody temperature, <·> represents taking the spectral mean, Re{·} represents taking the real part of the complex number, and b, p, d, and i represent the band, FOV, moving mirror scanning direction, and spectral channel respectively.

[0060] For the original interferogram of HIRAS, the phases of the external radiation and the internal radiation of the instrument differ by 180 degrees. When observing a blackbody, the external radiation dominates; when observing cold space, the external radiation is zero and the internal instrument radiation dominates; when observing a ground target, it is the sum of the external radiation and the internal radiation. Based on the instrument noise evaluation method at two points of the blackbody and cold space, the spectral sets of the blackbody and cold space are obtained through repeated observations, the ensemble variance and covariance are calculated from them, and the instrument noise of HIRAS is calculated. However, this method has the following problems:

[0061] 1) During the on-orbit operation of HIRAS, the blackbody, cold space, and Earth target are observed independently in the corresponding order. When observing the Earth target scenario, these views will lack the temperature of the blackbody source at the current moment, making it difficult to dynamically evaluate the instrument noise;

[0062] 2) When HIRAS observes the Earth target scenario, when the instrument radiation passes through the atmosphere, it will be absorbed by the atmosphere in a wave number-dependent manner. The radiation at the top of the atmosphere can be considered as radiation from a continuous temperature range. Therefore, it is difficult to accurately evaluate whether there is noise in the Earth view;

[0063] 3) The analysis depth of the instrument noise depends on the observational data of the blackbody and cold space, lacking an effective comparative analysis method.

[0064] The diagonal terms of the observational covariance matrix of a remote sensing instrument can correspond to the variance of the observational values, and its square root can reflect the radiation noise, indicating the measurement accuracy.

[0065] Based on this, the embodiments of the present invention provide a method for analyzing the noise of an infrared hyperspectral atmospheric sounder. By obtaining the interferogram formed by the infrared hyperspectral atmospheric sounder observing the Earth target, generating the target spectral data corresponding to the interferogram, then performing denoising processing on the target spectral data to obtain the reconstructed spectral data, the fitting value of the spectrum of the Earth target can be represented by the reconstructed spectral data, and then the noise equivalent radiation variance used to characterize the noise characteristics of the infrared hyperspectral atmospheric sounder can be determined according to the target spectral data and the reconstructed spectral data.

[0066] Specifically, for noise-oriented analysis, spectral residuals can be calculated based on a set of actual observed spectral values and corresponding simulated spectral values. Similar to the blackbody spectrum, the variance and covariance of the spectral residuals are considered during the calculation. The simulated spectral values can be obtained in various ways, such as through orthogonal basis methods based on Fourier transform or principal component analysis (PCA), etc. In the embodiments of the present invention, the orthogonal basis method based on PCA can be used to project the data into the orthogonal basis, and by selecting appropriate principal component components, the data is inversely transformed into the physical space. At this time, the result after the inverse transformation can be regarded as the reconstruction of the actual observed spectral values, that is, the simulated spectral values. Finally, the spectral residuals are obtained based on the actual observed spectral values of HIRAS and the simulated spectral values obtained through PCA, and the covariance matrix is calculated based on the spectral residuals to characterize the instrument noise. The target spectral data in the embodiments of the present invention is the actual observed spectral values, and the reconstructed spectral data is the simulated spectral values.

[0067] The following combines Figures 1-7 to describe the noise analysis method of the infrared hyperspectral atmospheric sounder of the present invention. This noise analysis method of the infrared hyperspectral atmospheric sounder can be applied to the infrared hyperspectral atmospheric sounder or to the infrared hyperspectral atmospheric sounder noise analysis device provided in the infrared hyperspectral atmospheric sounder. This infrared hyperspectral atmospheric sounder noise analysis device can be implemented through software, hardware, or a combination of both.

[0068] Figure 1 Exemplarily shows a schematic flowchart of the noise analysis method of the infrared hyperspectral atmospheric sounder provided by the embodiments of the present invention. Referring to Figure 1 as shown, this noise analysis method of the infrared hyperspectral atmospheric sounder can include the following steps 110 to 140.

[0069] Step 110: Obtain the interferogram formed by the infrared hyperspectral atmospheric sounder observing the Earth target.

[0070] In the embodiments of the present invention, a cloudless clear sky over the same ocean area that can be detected by HIRAS in the satellite flight orbit can be selected as the Earth target, such as the Pacific region. This can reduce observation errors and errors caused by different surface types, etc. HIRAS carried on the satellite can observe this ocean area multiple times. Each time it observes, it receives the solar radiation reflected by the ocean and absorbed by the atmosphere to form an interferogram.

[0071] For example, Figure 2 Exemplarily shows a schematic diagram of the observation range of HIRAS provided by the embodiments of the present invention. Referring to Figure 2As shown, an interferogram obtained by scanning a target ocean area with one FOV in HIRAS can be, for example, area 20 in the figure. It can be understood that each FOV of HIRAS can obtain an interferogram of its corresponding scanned area, and HIRAS can observe the target ocean area at different times to obtain an observation data set.

[0072] Step 120: Generate target spectral data corresponding to the interferogram.

[0073] After obtaining the interferogram, the interferogram can be converted into a spectrogram to obtain the target spectral data, which can reflect the solar energy spectrum after being reflected by the ocean area and absorbed by the atmosphere.

[0074] In noise-oriented analysis, there are usually biases in the simulated spectral values. In PCA processing, it is assumed that the signal is distributed in the first orthogonal basis, but the noise variance is distributed in all orthogonal bases, that is, the noise requires all principal component components to characterize. For a given principal component component, if only the noise information of the remaining principal component components is considered when characterizing the instrument noise, rather than the noise information in all orthogonal bases, the obtained simulated spectral values may be biased. For a highly redundant signal, the information components are mainly concentrated in a few orthogonal bases, so more components can be used to characterize the noise information, and the bias can be ignored at this time. Based on this, for noise-oriented analysis, data set screening can be performed so that the most prominent signals are restricted within a few larger principal components as much as possible.

[0075] Based on this, in an exemplary embodiment, generating the target spectral data corresponding to the interferogram may include: performing a Fourier transform on the interferogram to obtain initial spectral data; determining the mean value of the initial spectral data; and screening the initial spectral data based on the mean value and a set threshold to obtain the target spectral data corresponding to the interferogram. In this way, target spectral data with better consistency can be obtained through data screening.

[0076] For example, the initial spectral data can be screened by the following formula (3), and formula (3) can be expressed as:

[0077]

[0078] where R0 is the initial spectral data, that is, the initial HIRAS sample data, is the mean value of the initial spectral data, T is the set threshold, and |·| represents taking the absolute value.

[0079] Step 130: Denoise the target spectral data to obtain reconstructed spectral data.

[0080] After obtaining the target spectral data, denoising processing can be performed on the target spectral data to reconstruct the target spectral data and obtain reconstructed spectral data, which is also the simulated spectral value of the Earth target observed by HIRAS.

[0081] For example, the PCA method can be used to perform denoising processing on the target spectral data to obtain reconstructed spectral data. Specifically, performing denoising processing on the target spectral data to obtain reconstructed spectral data may include: obtaining the difference between the target spectral data and the spectral mean to obtain a spectral difference; determining the covariance matrix of the target spectral data based on the spectral difference; determining the noise prior matrix according to the covariance matrix of the target spectral data; normalizing the spectral difference based on the noise prior matrix to obtain a normalized spectral difference; and removing the noise in the target spectral data by using the principal component analysis method based on the normalized spectral difference to obtain reconstructed spectral data.

[0082] Exemplarily, assume that the obtained target spectral data R after screening is a spectral vector matrix of d×N, where d is the number of spectra of HIRAS and N is the data volume of each spectrum. The number of the target spectral data R can be expressed as d×N, then the spectral mean of the target spectral data R can be expressed as the following formula (4). Formula (4) is:

[0083]

[0084] where, R i represents the i-th group of spectral vectors in the target spectral data R.

[0085] Furthermore, for the radiation vector R, assume that its noise is additive noise, that is, R = s + ε, where s is the signal and ε is the noise term, and the noise is uncorrelated with the signal. For HIRAS, its instrument noise is mainly Gaussian noise, then it can be assumed that ε is Gaussian noise and the mean of ε is zero. Then, determining the covariance matrix of the target spectral data based on the spectral difference and determining the noise prior matrix according to the covariance matrix of the target spectral data can be realized by the following formula (5), and this formula (5) can be expressed as:

[0086]

[0087] where, represents the noise prior matrix, i.e., represents the covariance matrix of the target spectral data, R i represents the i-th group of spectral vectors in the target spectral data R, N is the data volume of each group of spectra, is the spectral mean of the target spectral data R. Thus, it can be understood that the noise prior matrix can be realized by the standard deviation of the target spectral data.

[0088] Further, the spectral difference is normalized based on the noise prior matrix, and the normalized spectral difference can be achieved through formula (6), and formula (6) can be expressed as:

[0089]

[0090] where X i is the normalized spectral difference and is a normalized zero-mean vector.

[0091] Exemplarily, based on the normalized spectral difference, the principal component analysis method is used to remove the noise in the target spectral data to obtain the reconstructed spectral data, which may include: determining the principal component components and principal component weights based on the normalized spectral difference, and the principal component components are used to characterize the non-noise data in the target spectral data; reconstructing the spectral difference according to the principal component components, principal component weights and noise prior matrix to obtain the target spectral difference; determining the reconstructed spectral data based on the target spectral difference and spectral mean.

[0092] Among them, determining the principal component components and principal component weights based on the normalized spectral difference may include: determining the covariance matrix of the normalized spectral difference to obtain the target covariance matrix; performing singular value decomposition on the target covariance matrix to obtain the orthogonal basis eigenvectors and eigenvalues; determining the principal component weights according to the orthogonal basis eigenvectors and the normalized spectral difference; determining the principal component components based on the change curve of the eigenvalues.

[0093] Specifically, the vector X i carrying the instrument noise can be used for PCA transformation processing, and the spectral difference is reconstructed in combination with the PCA transformation result, and then the reconstructed spectral data is reconstructed. During the reconstruction process, the signal in R, that is, the non-noise data, can be extracted to separate the noise.

[0094] For example, let X be a d×N-dimensional matrix, and its columns are the normalized spectral differences X i . The orthogonal basis of the d-dimensional vector X i can be obtained through the singular value decomposition (SVD) of the symmetric covariance matrix. Specifically, determining the covariance matrix of the normalized spectral difference to obtain the target covariance matrix can be achieved through formula (7), and this formula (7) can be expressed as:

[0095]

[0096] where N is the data volume of each spectrum, X T is the transpose of X, S represents the covariance matrix of the normalized spectral difference, and this covariance matrix also represents the covariance matrix after the normalization processing of the target spectral data.

[0097] Further, the SVD decomposition of S can be performed through formula (8), and formula (8) can be expressed as:

[0098] S = UAU T ;

[0099] where U is a unitary matrix, U T is the transpose of U, and A is a diagonal matrix. It can be understood that U is the orthogonal basis eigenvector, and the diagonal elements of A are the eigenvalues.

[0100] Further, based on the orthogonal basis eigenvector U and the normalized spectral difference X i , the principal component weight C i can be determined using formula (9), and formula (9) can be expressed as:

[0101] C i = U T X i .

[0102] where the elements of the vector C i are uncorrelated. The matrix U is a d×d-dimensional vector.

[0103] Further, the principal component component U τ can be determined based on the change curve of the eigenvalues, where τ represents the column index of the matrix U, and its value range is 1 ≤ τ ≤ d, and τ is an integer.

[0104] For a highly redundant signal, when τ << d, the calculation deviation of the instrument noise can be ignored. However, if τ is too small, the calculation result of the instrument noise will include the contribution from the signal; if τ is too large, since the instrument noise is distributed in all the orthogonal bases, more noise information will be lost, resulting in an underestimation of the noise variance. Therefore, the correct selection of τ is crucial. When selecting the principal component component, considering that the eigenvalues represent the weights of different orthogonal bases, the principal component component can be selected by combining the eigenvalue weights.

[0105] For example, Figure 3 exemplarily shows a schematic diagram of the change curve of the eigenvalues obtained by singular value decomposition in an embodiment of the present invention. Referring to Figure 3 as shown, the horizontal axis is the column index of the matrix U, and the vertical axis is the eigenvalue. As the column index increases, the eigenvalue decreases and remains stable after decreasing to a certain extent. The column index at the inflection point where the curve becomes stable can be determined as τ, and then the first τ columns of the matrix U are retained and the remaining d - τ columns are set to zero. The obtained matrix is the principal component component U τ .

[0106] Based on this, when only considering the first τ columns of the matrix U, according to the principal component component U τ, the weight C of the main component i and the noise prior matrix Reconstruct the spectral difference according to formula (10) to obtain the target spectral difference Furthermore, based on this target spectral difference and the spectral mean Use formula (11) to determine the reconstructed spectral data.

[0107] Formula (10) is

[0108] Formula (11) is wherein is the reconstructed spectral data.

[0109] Step 140: Determine the noise equivalent radiation variance according to the target spectral data and the reconstructed spectral data, and the noise equivalent radiation variance characterizes the noise characteristics of the infrared hyperspectral atmospheric sounder.

[0110] Exemplarily, determining the noise equivalent radiation variance according to the target spectral data and the reconstructed spectral data may include: obtaining the spectral residuals of the target spectral data and the reconstructed spectral data; determining the noise covariance matrix based on the spectral residuals; determining the noise equivalent radiation variance according to the noise covariance matrix.

[0111] For example, for redundant signals, such as spectral radiation, the noise covariance matrix S can be characterized by the covariance matrix of the spectral residuals ε . Specifically, the noise covariance matrix S can be obtained through formula (12) ε , and formula (12) can be expressed as:

[0112]

[0113] wherein, R i represents the i-th group of spectral vectors in the target spectral data R, N is the data volume of each group of spectra, represents R i and 's spectral residual, is 's transpose.

[0114] Based on this, the noise equivalent radiation variance NEdN can be obtained through formula (13), and formula (13) is:

[0115]

[0116] In an exemplary embodiment, the Noise Equivalent Difference Temperature (NEdT) is one of the important indicators for measuring the performance of an infrared detector system and can characterize the instrument sensitivity. When evaluating the noise of the HIRAS instrument, the NEdN can be converted to NEdT by combining with the Planck blackbody radiation formula (14), and formula (14) can be expressed as:

[0117]

[0118] where L represents spectral radiance, h represents Planck's constant, and the value of h is 6.626×10- 34 Joule·second, c represents the speed of light, and the value of c is 2.998×108 m / s, k represents Boltzmann's constant, and the value of k is 1.3806×10 -23 Joule / Kelvin, σ is the wave number, and ΔT is the thermodynamic temperature.

[0119] Based on formula (14), substituting the noise equivalent radiation variance NEdN into L and calculating to obtain ΔT, which is the NEdT. In this way, the NEdN can be converted to NEdT.

[0120] The noise analysis method for an infrared hyperspectral atmospheric sounder provided by the embodiment of the present invention includes obtaining an interferogram formed by the infrared hyperspectral atmospheric sounder observing the Earth target, generating target spectral data corresponding to the interferogram, then performing denoising processing on the target spectral data to obtain reconstructed spectral data. The reconstructed spectral data can be used as the simulation value of the spectrum of the Earth target. Then, the noise equivalent radiation variance is determined based on the target spectral data and the reconstructed spectral data, and the noise equivalent radiation variance can be used to characterize the noise characteristics of the infrared hyperspectral atmospheric sounder. In this way, the instrument noise of the infrared hyperspectral atmospheric sounder can be characterized by using the target spectral data formed by the infrared hyperspectral atmospheric sounder observing the Earth target, avoiding the limitations of space-time factors when estimating the instrument noise by means of a blackbody and cold air, and improving the real-time performance of instrument noise characterization.

[0121] Based on the noise analysis method for an infrared hyperspectral atmospheric sounder in the above embodiment, taking the HIRAS carried by the FY-3 satellite as an example, the noise analysis method for an infrared hyperspectral atmospheric sounder provided by the embodiment of the present invention will be further illustrated by examples.

[0122] The HIRAS carried by the Fengyun-3 satellite is mainly used for retrieving the atmospheric temperature and humidity profiles, and can also retrieve the profiles of atmospheric trace gases such as ozone, carbon monoxide, and carbon dioxide. Taking the Fengyun-3 FY-3E satellite as an example, it has 9 FOVs, can form 29 FORs through cross-track scanning, and the spatial resolution is about 16 km. Among them, the spectral range includes a long-wave infrared spectrum WL_LW and mid-wave and short-wave infrared spectra WL_MW1 and WL_MW2, and can achieve full spectral coverage from 650 cm -1 to 2550 cm -1 Its L1-level data products include the observation data of the blackbody, cold space NEdN, and Earth targets corresponding to 3 spectral detectors, as well as other auxiliary data. The spectral information of HIRAS is shown in Table 1.

[0123] Table 1

[0124]

[0125] Based on this, the following combines the L1-level Earth observation data products of HIRAS carried by the FY-3E satellite to analyze the instrument noise analysis method based on PCA, and can be verified by combining analysis methods such as blackbody and cold space. In this analysis process, one FOV of HIRAS is used as an example, such as FOV5. Then the infrared hyperspectral atmospheric sounder noise analysis method provided by the embodiments of the present invention may include the following steps S11 to step S17.

[0126] Step S11: Select observation data.

[0127] It is possible to select HIRAS to observe the Earth target and screen the ocean sample data set under cloudless clear sky. For example, select the observation data under clear sky in the Pacific region, and its observation range can be referred to Figure 2 as shown.

[0128] For a single FOV, its scanning range can be divided into several pixels, and the real part of the complex spectrum corresponding to each pixel in the observation data can be extracted to obtain the initial spectral data. Figure 4 Exemplarily shows a schematic diagram of the initial spectrogram, that is, the spectrogram corresponding to the initial spectral data. Among them, different grayscales represent different pixels, and different pixels exhibit different spectra. On the basis of Figure 4 the initial spectrogram, the initial spectral data can be further screened by combining a set threshold. For example, Figure 5 exemplarily shows a schematic diagram of the screening result of the initial spectral data, as Figure 5 shown, and the target spectral data obtained through screening is more aggregated compared to Figure 4 the initial spectral data.

[0129] Step S12: Calculate the spectral mean.

[0130] Calculate the average spectrum of the above screening results to obtain the spectral mean value, which can reflect the overall spectral change of the target spectral data through this spectral mean value.

[0131] Step S13: Calculate the spectral difference.

[0132] Subtract the spectral mean value from the target spectral data obtained by screening to obtain the spectral difference.

[0133] Step S14: Reconstruct the spectral difference based on the PCA method.

[0134] In the embodiment of the present invention, it is necessary to construct a prior matrix of the instrument noise standard deviation of HIRAS, that is, the noise prior matrix of the target spectral data, and this noise prior matrix can be represented by the standard deviation of the target spectral data. Specifically, the standard deviation of the target spectral data obtained by screening the initial spectral data and the spectral mean value calculated in step S12 can be calculated. According to this standard deviation, combined with the above formula (5), the noise prior matrix can be obtained, and this noise prior matrix is the prior matrix of the noise standard deviation.

[0135] Furthermore, the spectral difference obtained in step S13 can be normalized based on the obtained noise prior matrix to obtain the normalized spectral difference.

[0136] Before using the PCA method to reconstruct the spectral difference, it is necessary to select appropriate principal component components to ensure the separation of noise and signal, without losing more noise information, and without introducing too much signal contribution in the separated noise results. Specifically, in combination with the meaning of the eigenvalues, the appropriate principal component components can be selected according to the Figure 3 changing curve of the eigenvalues shown. After determining the appropriate principal component components, the spectral difference can be reconstructed according to the PCA method. Specifically, based on the above formula (10) and combined with Figure 3 perform PCA reconstruction on the normalized spectral difference to reconstruct the target spectral difference.

[0137] Step S15: Determine the reconstructed spectral data.

[0138] According to the target spectral difference reconstructed by the PCA method and the spectral mean value in step S12, the target spectral data can be reconstructed by using the above formula (11) to obtain the reconstructed spectral data.

[0139] Step S16: Calculate the NEDN result according to the target spectral data and the reconstructed spectral data.

[0140] The HIRAS on FY-3E contains 9 FOVs. Taking a single FOV as an example, the NEdN of this FOV can be calculated based on the reconstructed spectral data obtained by the PCA method and the target spectral data observed by HIRAS. Figure 6 An exemplary schematic diagram of the results of the noise equivalent radiance variance is shown. Among them, curve ① is the noise equivalent radiance variance curve obtained by observing the Earth target, and curve ② is the noise equivalent radiance variance curve obtained by observing the cold sky. The results show that the NEdN calculated using the PCA method and the Earth target observation data has good consistency with the NEdN results obtained by observing the cold sky. Therefore, the infrared hyperspectral atmospheric sounder noise analysis method provided by the embodiments of the present invention can effectively evaluate the instrument noise of HIRAS and avoid the limitations of space-time factors during the observation using a blackbody and the cold sky.

[0141] Step S17: Calculate NEdT according to NEdN and Planck's formula.

[0142] Based on the NEdN result in step S16 above, combined with Planck's formula (14), NEdN can be converted into NEdT. For example, Figure 7 An exemplary schematic diagram of the results of the noise equivalent temperature difference is shown. Among them, curve ③ is the NEdT curve obtained by observing the Earth target, and curve ④ is the NEdT curve obtained by observing the cold sky. By comparing and analyzing the two, it can be seen that the NEdT obtained by observing the Earth target has good consistency with the NEdT obtained by observing the cold sky. Therefore, the scheme of the present application can better realize the characterization of the instrument sensitivity and avoid the limitations of space-time factors during the observation using a blackbody and the cold sky.

[0143] The infrared hyperspectral atmospheric sounder noise analysis method provided by the embodiments of the present invention gives a new method for characterizing the instrument noise of an infrared hyperspectral atmospheric sounder. When HIRAS observes the Earth target scene, it can dynamically analyze the instrument noise based on the Earth target observation data, providing a method independent of the instrument noise assessment based on blackbody and cold sky observation data. The Earth target observation data of HIRAS can be analyzed by the PCA method to analyze the instrument noise of HIRAS, supplementing the instrument noise analysis method on the basis of conventional analyses such as blackbody and cold sky observations, and being able to verify each other; moreover, the method provided by the embodiments of the present invention overcomes some space-time limitations of blackbody and cold sky observations, improves the real-time performance of instrument noise characterization, and has a wider scope of application.

[0144] Next, the infrared hyperspectral atmospheric sounder noise analysis device provided by the present invention will be described. The infrared hyperspectral atmospheric sounder noise analysis device described below can be mutually referred to corresponding to the infrared hyperspectral atmospheric sounder noise analysis method described above.

[0145] Figure 8 The structural schematic diagram of the noise analysis device of the infrared hyperspectral atmospheric sounder provided by the embodiment of the present invention is exemplarily shown. Refer to Figure 8 As shown, the noise analysis device 1400 of the infrared hyperspectral atmospheric sounder may include an acquisition module 1410, a generation module 1420, a processing module 1430, and a determination module 1440. Among them, the acquisition module 1410 is used to acquire an interferogram formed by the infrared hyperspectral atmospheric sounder observing the earth target; the generation module 1420 is used to generate target spectral data corresponding to the interferogram; the processing module 1430 is used to perform denoising processing on the target spectral data to obtain reconstructed spectral data; the determination module 1440 is used to determine the noise equivalent radiation variance according to the target spectral data and the reconstructed spectral data, and the noise equivalent radiation variance characterizes the noise characteristics of the infrared hyperspectral atmospheric sounder.

[0146] In an exemplary embodiment, the processing module 1430 may include: a first acquisition unit, configured to acquire the difference between the target spectral data and the spectral mean to obtain a spectral difference; a first determination unit, configured to determine the covariance matrix of the target spectral data based on the spectral difference, and determine the noise prior matrix according to the covariance matrix of the target spectral data; a processing unit, configured to perform normalization processing on the spectral difference based on the noise prior matrix to obtain a normalized spectral difference; a denoising unit, configured to remove the noise in the target spectral data by using the principal component analysis method based on the normalized spectral difference to obtain reconstructed spectral data.

[0147] In an exemplary embodiment, the denoising unit may include: a first determination subunit, configured to determine the principal component components and the principal component weights based on the normalized spectral difference, and the principal component components are used to characterize the non-noise data in the target spectral data; a reconstruction subunit, configured to reconstruct the spectral difference according to the principal component components, the principal component weights, and the noise prior matrix to obtain a target spectral difference; a second determination subunit, configured to determine the reconstructed spectral data based on the target spectral difference and the spectral mean.

[0148] In an exemplary embodiment, the first determination subunit may specifically be configured to: determine the covariance matrix of the normalized spectral difference to obtain a target covariance matrix; perform singular value decomposition on the target covariance matrix to obtain orthogonal basis eigenvectors and eigenvalues; determine the principal component weights according to the orthogonal basis eigenvectors and the normalized spectral difference; determine the principal component components based on the change curve of the eigenvalues.

[0149] In an exemplary embodiment, the generation module 1420 may include: a transformation unit, configured to perform Fourier transform on the interferogram to obtain initial spectral data; a second determination unit, configured to determine the mean of the initial spectral data; a screening unit, configured to screen the initial spectral data based on the mean and a set threshold to obtain the target spectral data corresponding to the interferogram.

[0150] In an exemplary embodiment, the determination module 1440 may include: a second acquisition unit configured to acquire target spectral data and a spectral residual of the reconstructed spectral data; a third determination unit configured to determine a noise covariance matrix based on the spectral residual; and a fourth determination unit configured to determine a noise equivalent irradiance variance according to the noise covariance matrix.

[0151] Figure 9 An exemplary structural diagram of an infrared hyperspectral atmospheric sounder provided by an embodiment of the present invention is shown, as Figure 9 shown, the electronic device 1500 may include: a processor 1510, a communication interface 1520, a memory 1530, and a communication bus 1540. Among them, the processor 1510, the communication interface 1520, and the memory 1530 communicate with each other through the communication bus 1540. The processor 1510 may call logic instructions in the memory 1530 to execute the infrared hyperspectral atmospheric sounder noise analysis method provided by each of the above method embodiments. The method may include, for example: acquiring an interferogram formed by the infrared hyperspectral atmospheric sounder observing a terrestrial target; generating target spectral data corresponding to the interferogram; performing denoising processing on the target spectral data to obtain reconstructed spectral data; and determining a noise equivalent irradiance variance according to the target spectral data and the reconstructed spectral data, where the noise equivalent irradiance variance characterizes the noise characteristics of the infrared hyperspectral atmospheric sounder.

[0152] In addition, when the logic instructions in the above-mentioned memory 1530 are implemented in the form of a software functional unit and sold or used as an independent product, they may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, may be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc that can store program codes.

[0153] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the infrared hyperspectral atmospheric sounder noise analysis method provided by each of the above method embodiments. The method may include, for example: obtaining an interferogram formed by the infrared hyperspectral atmospheric sounder observing the earth target; generating target spectral data corresponding to the interferogram; performing denoising processing on the target spectral data to obtain reconstructed spectral data; determining the noise equivalent radiance variance according to the target spectral data and the reconstructed spectral data, and the noise equivalent radiance variance characterizes the noise characteristics of the infrared hyperspectral atmospheric sounder.

[0154] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is implemented to execute the infrared hyperspectral atmospheric sounder noise analysis method provided by each of the above method embodiments. The method may include, for example: obtaining an interferogram formed by the infrared hyperspectral atmospheric sounder observing the earth target; generating target spectral data corresponding to the interferogram; performing denoising processing on the target spectral data to obtain reconstructed spectral data; determining the noise equivalent radiance variance according to the target spectral data and the reconstructed spectral data, and the noise equivalent radiance variance characterizes the noise characteristics of the infrared hyperspectral atmospheric sounder.

[0155] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0156] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, also by hardware. Based on such an understanding, the above technical solution, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disc, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for noise analysis of an infrared hyperspectral atmospheric sounder, characterized in that, Including: Obtaining an interferogram formed by an infrared hyperspectral atmospheric sounder observing an Earth target; Generating target spectral data corresponding to the interferogram; Performing denoising processing on the target spectral data to obtain reconstructed spectral data; Determining a noise equivalent radiance variance according to the target spectral data and the reconstructed spectral data, where the noise equivalent radiance variance characterizes the noise characteristics of the infrared hyperspectral atmospheric sounder; The performing denoising processing on the target spectral data to obtain reconstructed spectral data includes: Obtaining a difference between the target spectral data and a spectral mean value to obtain a spectral difference; Determining a covariance matrix of the target spectral data based on the spectral difference, and determining a noise prior matrix according to the covariance matrix of the target spectral data; Performing normalization processing on the spectral difference based on the noise prior matrix to obtain a normalized spectral difference; Based on the normalized spectral difference, using principal component analysis to remove noise in the target spectral data to obtain reconstructed spectral data; The based on the normalized spectral difference, using principal component analysis to remove noise in the target spectral data to obtain reconstructed spectral data includes: Determining principal component components and principal component weights based on the normalized spectral difference, where the principal component components are used to characterize non-noise data in the target spectral data; Reconstructing a spectral difference according to the principal component components, the principal component weights, and the noise prior matrix to obtain a target spectral difference; Determining reconstructed spectral data based on the target spectral difference and the spectral mean value; The determining principal component components and principal component weights based on the normalized spectral difference includes: Determining a covariance matrix of the normalized spectral difference to obtain a target covariance matrix; Performing singular value decomposition on the target covariance matrix to obtain orthogonal basis eigenvectors and eigenvalues; Determining the principal component weights according to the orthogonal basis eigenvectors and the normalized spectral difference; Determining the principal component components based on a change curve of the eigenvalues.

2. The noise analysis method of the infrared hyperspectral atmospheric sounder according to claim 1, characterized in that The generating target spectral data corresponding to the interferogram includes: Performing Fourier transform on the interferogram to obtain initial spectral data; Determining a mean value of the initial spectral data; Based on the mean value and a set threshold, screening the initial spectral data to obtain the target spectral data corresponding to the interferogram.

3. The noise analysis method of the infrared hyperspectral atmospheric detector according to claim 1, characterized in that, The determining a noise equivalent radiance variance according to the target spectral data and the reconstructed spectral data includes: Obtaining spectral residuals of the target spectral data and the reconstructed spectral data; Determining a noise covariance matrix based on the spectral residuals; Determining a noise equivalent radiance variance according to the noise covariance matrix.

4. An apparatus for analyzing the noise of an infrared hyperspectral atmospheric sounder, characterized in that, Including: An obtaining module for obtaining an interferogram formed by an infrared hyperspectral atmospheric sounder observing an Earth target; A generating module for generating target spectral data corresponding to the interferogram; A processing module for performing denoising processing on the target spectral data to obtain reconstructed spectral data; A determining module for determining a noise equivalent radiance variance according to the target spectral data and the reconstructed spectral data, where the noise equivalent radiance variance characterizes the noise characteristics of the infrared hyperspectral atmospheric sounder; The processing module performs denoising processing on the target spectral data to obtain reconstructed spectral data, including: Obtaining the difference between the target spectral data and the spectral mean value to obtain a spectral difference; Determining the covariance matrix of the target spectral data based on the spectral difference, and determining a noise prior matrix according to the covariance matrix of the target spectral data; Performing normalization processing on the spectral difference based on the noise prior matrix to obtain a normalized spectral difference; Based on the normalized spectral difference, using the principal component analysis method to remove the noise in the target spectral data to obtain reconstructed spectral data; The processing module uses the principal component analysis method to remove the noise in the target spectral data based on the normalized spectral difference to obtain reconstructed spectral data, including: Determining the principal component components and the principal component weights based on the normalized spectral difference, where the principal component components are used to characterize the non-noise data in the target spectral data; Reconstructing the spectral difference according to the principal component components, the principal component weights, and the noise prior matrix to obtain a target spectral difference; Determining the reconstructed spectral data based on the target spectral difference and the spectral mean value; The processing module determines the principal component components and the principal component weights based on the normalized spectral difference, including: Determining the covariance matrix of the normalized spectral difference to obtain a target covariance matrix; Performing singular value decomposition on the target covariance matrix to obtain orthogonal basis eigenvectors and eigenvalues; Determining the principal component weights according to the orthogonal basis eigenvectors and the normalized spectral difference; Determining the principal component components based on the change curve of the eigenvalues.

5. An infrared hyperspectral atmospheric detector, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the infrared hyperspectral atmospheric detector noise analysis method according to any one of claims 1 to 3.

6. A non-transitory computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by the processor, it implements the infrared hyperspectral atmospheric detector noise analysis method according to any one of claims 1 to 3.

7. A computer program product comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the infrared hyperspectral atmospheric detector noise analysis method according to any one of claims 1 to 3.

Images