A mid-infrared hyperspectral emission and reflection characteristic integrated inversion method

Abstract

The present application belongs to the technical field of quantitative remote sensing, and relates to a kind of middle infrared hyperspectral emission and reflection characteristic integrated retrieval method, comprising: obtaining middle infrared hyperspectral satellite remote sensing data, and optimizing processing is carried out to the sensing entrance pupil radiance data, to obtain optimized entrance pupil radiance data;Middle infrared hyperspectral satellite remote sensing data includes sensing entrance pupil radiance data;Optimized entrance pupil radiance data is carried out atmospheric correction using radiation transfer model, obtains ground surface brightness temperature and atmospheric downlink radiation, and calculates solar irradiance reaching ground surface;Middle infrared hyperspectral ground surface bidirectional reflectance factor retrieval framework is constructed, and ground surface bidirectional reflectance factor retrieval is carried out based on middle infrared hyperspectral ground surface bidirectional reflectance factor retrieval framework, to obtain ground surface bidirectional reflectance factor and ground surface equivalent radiance;Scene adaptive framework is constructed, and ground surface temperature and ground surface emissivity are retrieved based on scene adaptive framework;To complete the accurate separation of ground surface emission radiation and reflection radiation.

Description

Technical Field

[0001] This invention relates to the field of quantitative remote sensing technology, and specifically discloses an integrated inversion method for mid-infrared hyperspectral emission and reflectance characteristics. Background Technology

[0002] Mid-infrared remote sensing technology plays an irreplaceable role in Earth observation and ecological environment research. In soil and water monitoring, the mid-infrared band shows a significant response to information such as soil moisture and organic matter content, effectively supporting farmland water management and desertification control. The accuracy of surface temperature retrieval from mid-infrared data is relatively less sensitive to changes in atmospheric water vapor and emissivity errors, and the emissivity of typical land features exhibits unique spectral variations in the mid-infrared band. Overall, mid-infrared remote sensing technology not only expands the sources of spectral information for monitoring surface elements but also provides crucial technical support for fields such as natural disaster early warning and ecological environment assessment, demonstrating broad application prospects and profound scientific value.

[0003] As a prerequisite for remote sensing applications, mid-infrared data products are an important information carrier, and the accuracy of their production or inversion directly affects subsequent scientific analysis. In addition to primary products such as entrance pupil radiance produced after spatial correction and spectral radiometric calibration, mid-infrared remote sensing products also include inversion products such as land surface brightness temperature (LSBT), land surface bidirectional reflectance (LSBR), land surface temperature (LST), and land surface emissivity / emissivity (LSE). Among them, surface brightness temperature is the temperature-quantified value of surface radiance, which encompasses the surface's own emitted radiation, atmospheric downdraft reflected by the surface, and direct solar radiation reflected by the surface in two directions. In this patent, surface reflectivity specifically refers to surface two-way reflectivity, which directly participates in the reflection of direct solar radiation by the surface and is of great significance for studying surface radiation transmission, energy balance, and spectral characteristics of land features. Surface temperature is the direct driving force for long-wave radiation exchange between the Earth and the atmosphere, and can sensitively reveal the spatiotemporal dynamic characteristics of the Earth-atmosphere system balance. It also provides key information for the quantitative assessment of surface thermal inertia, soil moisture status, and energy-hydrological cycle. As a core surface parameter, it has been widely used in climate change monitoring, urban heat island effect analysis, and other fields. Surface emissivity, as an inherent radiation characteristic of land features, not only describes the surface geometry and chemical composition but is also significantly affected by factors such as waveband position, observation angle, dielectric constant, composition, surface roughness, and water content. It shows unique application advantages in describing the spatiotemporal evolution of land feature types and mineral resource exploration.

[0004] The emission and reflection characteristics of the mid-infrared spectral band (3~5 μm) mainly include two aspects: emission characteristics and reflection characteristics. The emission characteristic parameters that need to be inverted are surface temperature and surface emissivity, while the reflection characteristic parameter is surface dichroism. For a long time, the inversion of surface emission and reflection characteristics from mid-infrared hyperspectral remote sensing has been one of the challenges in the field of quantitative remote sensing technology. On the one hand, in the mid-infrared radiation signal received by satellite sensors, the outgoing radiation from the surface, the downward atmospheric radiation reflected in the hemispherical direction, the direct solar radiation reflected in the dichroism, and the atmospheric path radiation are strongly coupled together and difficult to separate. On the other hand, under ideal conditions with complete atmospheric correction, a known N-channel mid-infrared hyperspectral sensor can establish N observation equations, but it is necessary to invert 2N+1 unknown surface parameters (including N dichroisms, N emissivities, and 1 temperature), resulting in an underdetermined solution. Therefore, it is necessary to utilize the redundant information between mid-infrared hyperspectral channels, add constraints, or reduce the number of unknowns to obtain a solution to the equations, thereby avoiding ill-conditioned inversion.

[0005] The development of remote sensing surface bidirectional reflectance retrieval technology currently focuses mainly on the visible and near-infrared spectral bands, primarily including: 1) radiative transfer model inversion method; 2) multi-angle observation method; and 3) empirical statistical method. Among these, the radiative transfer model inversion method aims to utilize models to achieve high-precision calculations of atmospheric and solar parameters, and directly calculate surface bidirectional reflectance based on radiative transfer theory; the multi-angle observation method involves using the same sensor to observe the same target surface from multiple angles, establishing a bidirectional reflectance distribution function (BRDF) to obtain the bidirectional reflectance of the target object at a specific observation / solar angle; the empirical statistical method relies on ground-measured data, directly estimating the surface bidirectional reflectance by establishing a statistical relationship between observed radiance (hereinafter referred to as "radiance") and the measured surface bidirectional reflectance factor (BRF).

[0006] Currently, the main methods for remote sensing surface temperature retrieval include: 1) single-channel algorithm; 2) split-window algorithm; and 3) multi-angle algorithm. Among these, the single-channel algorithm relies on precise atmospheric parameters to perform atmospheric correction on remote sensing observations from a single mid-infrared channel, and retrieves surface temperature under known surface emissivity. This method has a clear principle and mature application, and has been widely used in processing data from sensors such as Landsat TM / ETM+ and HJ-1B. The split-window algorithm utilizes the difference in atmospheric absorption between two adjacent mid-infrared channels to effectively reduce atmospheric effects and achieves surface temperature retrieval under the premise that the emissivity of both channels is known. It is currently mainly applied to sensors such as NOAA / AVHRR and FY-3 VIRR. The multi-angle algorithm is based on the atmospheric differential absorption characteristics caused by the difference in the radiation transfer path between the surface and the satellite under different observation angles, thereby achieving surface temperature retrieval. Typical representative data sources include multi-angle observation satellite sensors such as Sentinel-3 / SLSTR.

[0007] The main methods for inverting remote sensing surface emissivity include: 1) Normalized Difference Vegetation Index (NDVI); 2) Surface Classification; 3) Gray Body Method; 4) Envelope Algorithm; and 5) Reference Channel Method. The NDVI method utilizes the high correlation between NDVI and surface emissivity to establish an empirical function relationship for emissivity estimation. The surface classification method relies on the visible and near-infrared spectral bands to distinguish land cover types and combines this with the typical emissivity ranges for different land cover types to achieve inversion. The gray body method assumes that the surface emissivity within a certain spectral band is approximately independent of wavelength, thus simplifying the radiative transfer equation and solving for the emissivity. The envelope method selects a channel with an approximately 1 surface emissivity, constructs an emissivity spectral envelope, and derives the emissivity of other channels accordingly. Under precise atmospheric correction, the reference channel method calculates the surface temperature based on the known surface emissivity of a certain channel, combined with radiative transfer theory, and uses the surface temperature as a known condition to further deduce the emissivity of other channels.

[0008] The main algorithms for simultaneously inverting remote sensing surface temperature and emissivity include: 1) the dual-temperature method; 2) the MODIS day-night algorithm; and 3) the temperature and emissivity separation algorithm (TES). The dual-temperature method is based on the assumption that surface emissivity remains constant over short timescales. It utilizes multi-temporal satellite observations (day / night) to increase constraints, thereby simultaneously inverting surface temperature and emissivity. Based on day and night observation data from seven mid- and long-wave infrared channels, the MODIS day-night algorithm solves for surface temperature and emissivity using statistical regression and least-squares fitting methods. This method requires the assumption that some parameters remain consistent throughout the day and night, such as surface emissivity, atmospheric water vapor content, and temperature profiles. Under the premise of accurate atmospheric correction, the TES algorithm relies on the observations of five channels of the ASTER sensor and uses the emissivity normalization method to introduce empirical statistical constraints (the exponential relationship between the maximum and minimum normalized emissivity) to achieve the synchronous inversion of surface temperature and emissivity. However, this method is significantly dependent on the accuracy of atmospheric parameters, and its accuracy is limited by the inversion of gray bodies such as water bodies / vegetation.

[0009] In summary, current research on the utilization of mid-infrared hyperspectral data is still less effective than that on the visible, near-infrared, and thermal infrared bands because it simultaneously couples surface / atmospheric thermal radiation and direct solar radiation disturbances, making it difficult to effectively extract the characteristic information of surface emission and reflection. Furthermore, although some methods for retrieving mid-infrared dichroism reflectance have been published, these methods are only applicable to multispectral data and have limited applicability to hyperspectral data. Therefore, there is an urgent need to develop novel inversion methods for surface emission / reflection characteristics based on mid-infrared hyperspectral remote sensing.

[0010] To overcome the limitations of the above methods, this invention proposes an integrated inversion method for mid-infrared hyperspectral emission and reflection characteristics, which enables simultaneous inversion of surface temperature, mid-infrared hyperspectral surface diaxial reflectivity, and surface emissivity, thereby achieving accurate separation of surface emitted and reflected radiation. Summary of the Invention

[0011] The purpose of this invention is to provide an integrated inversion method for mid-infrared hyperspectral emission and reflectance characteristics, addressing the problem of simultaneously inverting surface temperature, mid-infrared hyperspectral surface dichroism, and surface emissivity to achieve accurate separation of surface emitted and reflected radiation. The specific solution is as follows:

[0012] A method for integrated inversion of mid-infrared hyperspectral emission and reflectance characteristics includes: acquiring mid-infrared hyperspectral satellite remote sensing data and optimizing the sensor entrance pupil radiance data to obtain optimized entrance pupil radiance data; the mid-infrared hyperspectral satellite remote sensing data includes sensor entrance pupil radiance data; performing atmospheric correction on the optimized entrance pupil radiance data using a radiative transfer model to obtain surface brightness temperature and downward atmospheric radiation, and calculating the solar irradiance reaching the surface; constructing a mid-infrared hyperspectral surface two-dimensional reflectance inversion framework, and performing surface two-dimensional reflectance inversion based on the mid-infrared hyperspectral surface two-dimensional reflectance inversion framework to obtain surface two-dimensional reflectance and surface equivalent radiance; constructing a scene-adaptive framework, and inverting the surface temperature and surface emissivity based on the scene-adaptive framework.

[0013] Furthermore, the optimization process includes cloud masking, bad value removal, and / or recalibration.

[0014] Furthermore, the atmospheric correction of the optimized entrance pupil radiance data using the radiative transfer model to obtain the surface brightness temperature and downward atmospheric radiation, and to calculate the solar irradiance reaching the surface, includes: acquiring product data of the mid-infrared hyperspectral entrance pupil radiance product and the corresponding temperature and humidity atmospheric profile at the time of product acquisition; the product data includes the observation time, observation angle, and solar angle; inputting the product data and the temperature and humidity atmospheric profile into the radiative transfer model for atmospheric correction to obtain corrected data; the corrected data includes surface brightness temperature, downward atmospheric radiation, and solar irradiance.

[0015] Furthermore, the construction of the mid-infrared hyperspectral surface two-dimensional reflectance inversion framework, and the inversion of surface two-dimensional reflectance based on the mid-infrared hyperspectral surface two-dimensional reflectance inversion framework to obtain surface two-dimensional reflectance and surface equivalent radiance, includes: constructing a three-channel combination for each channel and establishing an inversion model of mid-infrared hyperspectral surface two-dimensional reflectance based on the three-channel combination; simulating data in the radiation database using a radiative transfer model to obtain simulated data; the radiation database includes a surface emissivity, two-dimensional reflectance spectral library, and an atmospheric profile library; the simulated data includes simulated surface brightness temperature and simulated solar irradiance; fitting the mid-infrared hyperspectral surface two-dimensional reflectance inversion model based on the simulated data to obtain the mid-infrared hyperspectral surface two-dimensional reflectance inversion framework; inputting the surface brightness temperature and solar irradiance into the mid-infrared hyperspectral surface two-dimensional reflectance inversion framework to obtain the surface equivalent radiance and surface two-dimensional reflectance of each channel.

[0016] Furthermore, the equivalent radiance of the Earth's surface is defined as:

[0017] ;

[0018] in, This is the Planck lookup table for channel i; For channel Equivalent brightness temperature of the Earth's surface; The equivalent radiance of channel i on the ground surface; For the surface brightness temperature of channel i; For channel i, the surface radiance; Let i be the surface diaxial reflectance. Let i be the solar irradiance reaching the Earth's surface through channel i;

[0019] The equivalent brightness temperature of the Earth's surface is defined as:

[0020] ;

[0021] in, Let the equivalent brightness temperature of the surface of channel j be the solution. , and These are the surface brightness temperatures for channels i, j, and k, respectively; , , , and These are the fitting coefficients for the initial, first, second, third, and fourth surface equivalent brightness temperatures, respectively.

[0022] The dichroic reflectance of the Earth's surface is defined as:

[0023] ;

[0024] in, Let the surface diaxial reflectance of channel j be the solution to be found; For channel j, the surface radiance; The equivalent radiance of the ground surface for channel j; Let be the solar irradiance reaching the Earth's surface via channel j.

[0025] Furthermore, the construction of the scene-adaptive framework and the inversion of surface temperature and surface emissivity based on the scene-adaptive framework include: determining whether the target scene has a prior spectral library; the target scene refers to a specific area containing specific land cover composition and environmental characteristics observed by mid-infrared hyperspectral remote sensing; if the target scene has a prior spectral library, then the surface temperature and surface emissivity are inverted using a surface temperature and surface emissivity separation algorithm based on dictionary sparse representation; the dictionary refers to an emissivity dictionary; if the target scene does not have a prior spectral library, then the surface temperature and surface emissivity are inverted using a piecewise linear constraint surface temperature and surface emissivity separation algorithm; the new surface emissivity output by the piecewise linear constraint surface temperature and surface emissivity separation algorithm expands the training library and capacity of the emissivity dictionary, and a new inversion result is obtained through atomic rearrangement based on the dictionary sparse representation method.

[0026] Furthermore, the method of using a dictionary-based sparse representation algorithm to separate land surface temperature and emissivity to obtain land surface temperature and emissivity includes: convolving emissivity spectra from the spectral library into each channel, and performing sparse coding and dictionary training using orthogonal matching pursuit and K-singular value decomposition to obtain an emissivity dictionary; a channel refers to the smallest unit for acquiring radiation data from a mid-infrared hyperspectral entrance pupil radiance product; given an initial value of land surface temperature, reconstructing the land surface emissivity using the emissivity dictionary, and establishing a first cost function; repeatedly establishing the first cost function until the first cost function is less than a first threshold or the number of iterations is greater than a limit, at which point the iteration stops, and the final land surface temperature and emissivity are output.

[0027] Furthermore, dictionary training includes spectral library precoding and dictionary atom optimization updates;

[0028] The spectral library precoding is defined as follows:

[0029] ;

[0030] ;

[0031] ;

[0032] Where n is the spectral number; N is the total number of mid-infrared hyperspectral spectra; Emissivity ratio spectrum; This is the original channel-based mid-infrared emissivity spectrum; mean indicates taking the average value. The emissivity empirical spectrum; L is the length of a single empirical emissivity spectrum; max represents the maximum value; min represents the minimum value; For empirical spectral libraries; J is the total number of emissivity spectra;

[0033] The dictionary atomic optimization update is defined as follows:

[0034] ;

[0035] in, The optimal sparse representation is the emissivity dictionary of the training output; For emission rate identification; for and Joint optimization minimizes the objective function; For dictionary identifiers; Identifier for sparse coefficient matrix; Represents the L2 norm; The signal to be trained; A dictionary of atoms to be updated; For the training signal sparse coefficient matrix; This indicates that the condition is being met; It is the zero norm; for The Middle A length of A sparse coefficient column vector; Sparsity;

[0036] The observation matrix is ​​defined as:

[0037] ;

[0038] ;

[0039] ;

[0040] ;

[0041] in, The observation matrix; It is an augmented matrix of the identity matrix; The empirical observation matrix; For Hadema; It is a unit vector of length L; For normalized row vectors; It is the identity matrix; It is a zero matrix; Let J be a unit vector of length J; The initial signal to be compressed for observation; Signal to be trained The pseudo-inverse matrix; It is a 2-norm; This is a transpose operation; for The Middle A column vector of length L; To be The operation of combining column vectors into a single matrix;

[0042] By reconstructing the surface emissivity using an emissivity dictionary, the definitions of surface temperature and surface emissivity are obtained:

[0043] ;

[0044] ;

[0045] In the formula, Represents Haderma; The equivalent radiance vector of the Earth's surface; It is a wavelength vector; This is the downward atmospheric radiation vector; The emissivity sparsity coefficient; The sparse coefficients are unknown; It is the Planck function; The radiance representing the Earth's surface temperature at various wavelengths; Let N be a vector consisting entirely of 1s. This means minimizing the objective function; This is the reconstructed emissivity value; The function representing the reconstruction of sparse coefficients into surface emissivity:

[0046] ;

[0047] ;

[0048] in, For the reconstructed emissivity ratio spectrum; This represents the maximum value of the emissivity spectrum. This represents the minimum value of the emissivity spectrum. Represents taking the first vector. arrive One element; To obtain the reconstructed emissivity spectrum during the solution process;

[0049] The first cost function is defined as:

[0050] ;

[0051] in, The first cost function; The square of the L2 norm; This is the surface emissivity vector after dictionary reconstruction; To reconstruct the identifier; The characteristic radiance of the initial surface temperature at various wavelengths; This represents the initial value of the surface temperature in each iteration; This is the equivalent radiance vector of the Earth's surface.

[0052] Furthermore, the method for separating land surface temperature and land surface emissivity using piecewise linear constraints to invert and obtain land surface temperature and land surface emissivity includes: given an initial value of land surface temperature, determining the fitting coefficients of the piecewise linear equation of emissivity to obtain a piecewise emissivity spectrum; calculating the land surface emissivity using the piecewise emissivity spectrum and establishing a second cost function; repeatedly establishing the second cost function until the second cost function is less than a second threshold or the number of iterations is greater than a limit, stopping the iteration, and outputting the final land surface temperature and land surface emissivity.

[0053] Furthermore, the segmented emissivity spectrum is defined as:

[0054] ;

[0055] in, For channel indexing; For channel The center wavelength; For channel Emission rate; The slope of the emissivity segment; is the emissivity segment intercept; k is the index of the emissivity segment fitted line; M is the total number of emissivity segment fitted lines; and These represent the number of channels in segment K and segment K+1, respectively. The range of channel indexes within each segment;

[0056] Surface emissivity is defined as:

[0057] ;

[0058] in, The radiance at the channel represents the equivalent brightness temperature of the Earth's surface. For the surface temperature in the channel The radiance at that location; For channel Downward atmospheric radiation;

[0059] The second cost function is defined as:

[0060] ;

[0061] in, This is the second cost function; N is the channel index; N is the total number of channels; The emissivity reconstructed from the coefficients of a system of univariate linear equations in each iteration; The initial value of the surface temperature in the channel The radiance at that location; The initial value of the land surface temperature in each iteration. The integrated inversion method for mid-infrared hyperspectral emission and reflectance characteristics provided by this invention has the following advantages and beneficial effects:

[0062] (1) This method addresses the limitations of hyperspectral remote sensing in the mid-infrared band. After optimizing the quality of satellite entrance pupil radiance products and performing atmospheric correction, it utilizes only the difference in solar irradiance between the three channels. Based on the linear assumption between the parameters of the three channels, it first solves the underdetermined problem in the simultaneous inversion of surface bidirectional reflectance and surface equivalent radiance / brightness temperature, thus achieving the initial separation of surface emitted radiation and surface reflected radiation in the mid-infrared band.

[0063] (2) This method proposes to utilize the idea of ​​scene adaptation. Based on the fact that the TES algorithm can achieve underdetermined solutions for surface temperature and surface emissivity, it determines whether the application scenario has the prior spectral library conditions. If it does, the DSRE TES algorithm is used; otherwise, the LSEC TES algorithm is used. In addition, this adaptive framework can deeply integrate the two TES algorithms, making the two algorithms complementary to each other, so as to cover more application scenarios.

[0064] (3) By solving the double underdetermined problem, this method can efficiently invert the surface two-dimensional reflectance spectrum, surface temperature and surface emissivity spectrum, thereby further realizing the surface's own emission radiation. and surface reflected radiation High-precision separation. Attached Figure Description

[0065] Figure 1 An exemplary flowchart of an integrated inversion method for mid-infrared hyperspectral emission and reflection characteristics provided by the present invention;

[0066] Figure 2 An exemplary flowchart illustrating the process of optimizing mid-infrared entrance pupil radiance quality and atmospheric correction provided by the present invention;

[0067] Figure 3 This is an exemplary flowchart of the mid-infrared hyperspectral surface two-dimensional reflectance inversion method provided by the present invention;

[0068] Figure 4 An exemplary flowchart of the scene adaptive framework that integrates multiple methods for separating surface temperature and surface emissivity provided by the present invention;

[0069] Figure 5 An exemplary flowchart for constructing an emissivity sparse dictionary to separate surface temperature and surface emissivity provided by the present invention;

[0070] Figure 6 This is an exemplary flowchart for separating surface temperature and surface emissivity based on piecewise linear constraints provided by the present invention.

[0071] Figure 7 This is an exemplary structural diagram of an integrated inversion device for mid-infrared hyperspectral emission and reflection characteristics provided by the present invention. Detailed Implementation

[0072] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0073] This invention provides an integrated inversion method and apparatus for mid-infrared hyperspectral emission and reflectance characteristics. This method and apparatus are applicable to satellite remote sensing and airborne remote sensing fields, including: quality assessment of mid-infrared hyperspectral entrance pupil radiance data; optimization of data signal-to-noise ratio using techniques such as spectral / radiative recalibration; atmospheric correction of the entrance pupil radiance based on a radiative transfer model; calculation of solar radiation parameters; calculation of surface bidirectional reflectance by removing the influence of direct solar radiation using a physical empirical model based on the differences in solar radiation and information redundancy between mid-infrared hyperspectral channels; using the surface radiance after removing direct solar radiation as input, determining whether the underlying surface type is prior, selecting an appropriate temperature / emissivity separation method, and simultaneously inverting the target surface temperature and emissivity. Specifically, as shown... Figure 1 As shown, some embodiments of the present invention provide an integrated inversion method for mid-infrared hyperspectral emission and reflection characteristics, which may include steps S110 to S140.

[0074] Step S110: Acquire mid-infrared hyperspectral satellite remote sensing data, and optimize the entrance pupil radiance data to obtain optimized entrance pupil radiance data. The mid-infrared hyperspectral satellite remote sensing data includes at least the sensor entrance pupil radiance data. Optimization processing includes cloud masking, outlier removal, and / or recalibration. Recalibration refers to optimizing data quality if the payload has not been calibrated for a long period.

[0075] Step S120: Use the radiative transfer model to perform atmospheric correction on the optimized entrance pupil radiance data to obtain the surface brightness temperature and downward atmospheric radiation, and calculate the solar irradiance reaching the surface.

[0076] Step S130: Based on empirical assumptions and physical derivations, a mid-infrared hyperspectral surface two-dimensional reflectance inversion framework is constructed. Surface two-dimensional reflectance is then inverted based on this framework to obtain surface two-dimensional reflectance and equivalent surface radiance, thus resolving the underdetermined solution problem and achieving separation between surface two-dimensional reflectance and equivalent surface radiance. The mid-infrared hyperspectral surface two-dimensional reflectance inversion framework refers to a series of methods and procedures for inverting mid-infrared hyperspectral surface two-dimensional reflectance.

[0077] Step S140: Construct a scene adaptive framework and invert the surface temperature and surface emissivity based on the scene adaptive framework. This includes selecting the sparse dictionary reconstruction method if the scene has a prior spectral library, and selecting the piecewise linear constraint method otherwise, so as to achieve simultaneous inversion of surface temperature and surface emissivity.

[0078] Through the embodiments of the present invention, the effects of cloud cover and payload performance degradation on satellite entrance pupil radiance can be effectively avoided. With the support of atmospheric correction, the accurate inversion of mid-infrared hyperspectral surface biaxial reflectance is achieved based on a physical-empirical model. Subsequently, a scene-adaptive separation framework for surface temperature and surface emissivity is established. When the scene has a priori spectral library, the sparse representation of the dictionary is preferentially selected to reconstruct the surface emissivity, and the surface temperature is inverted at the same time. When the priori conditions are not available, piecewise linear constraints are selected to achieve the simultaneous inversion of surface temperature and surface emissivity.

[0079] The following will combine Figures 2-6 ,right Figure 1 The publicly available integrated inversion method for mid-infrared hyperspectral emission and reflectance characteristics is described in detail.

[0080] In this embodiment of the invention, the quality assessment of the mid-infrared hyperspectral entrance pupil radiance data is first performed: cloud masking and outlier removal. Cloud masking refers to the detection and shielding of pixels containing clouds. During this process, cloud products released by the same satellite platform as the target payload are preferentially used, or a cloud discrimination threshold can be set to exclude pixels containing clouds. Outlier removal mainly excludes null values, low response values, oversaturated values, edge effect anomalies, and high-noise stripes. Based on the calibration status of the mid-infrared hyperspectral data, it is determined whether recalibration is needed. If recalibration is required, the quality of the entrance pupil radiance is optimized using the acquired spectral and radiometric performance parameters. For example, if the mid-infrared hyperspectral satellite payload has not undergone spectral / radiometric calibration for a long time, it is determined that site-alternative calibration or other operations need to be performed on the target payload to obtain more accurate spectral / radiometric performance parameters, ensuring that the subsequent inversion results can be controlled within effective accuracy.

[0081] like Figure 2 As shown, Figure 2 Steps S1101 to S1102 schematically illustrate the flowchart of the mid-infrared entrance pupil radiance quality optimization provided by the present invention.

[0082] Step S1101: Obtain the entrance pupil radiance product of the mid-infrared hyperspectral satellite, perform cloud masking based on the cloud product or by establishing a threshold, and remove bad values ​​using quality control documents and empirical thresholds.

[0083] Step S1102: If the onboard mid-infrared hyperspectral payload has not been calibrated for a long time, it is necessary to use site-based alternative calibration and other recalibration methods to obtain spectral and radiation performance parameters and optimize the entrance pupil radiance data quality.

[0084] In the embodiments of this invention, atmospheric correction is based on atmospheric radiative transfer models such as MODTRAN (MODerate Resolution Atmospheric Transmission). These models mainly calculate atmospheric / solar parameters such as atmospheric up / down radiance, atmospheric transmittance, and direct solar radiation. Input parameters to the radiative transfer model include observation time, observation angle, solar angle, atmospheric temperature and humidity profiles, and basic model settings. After atmospheric correction, parameters such as surface brightness temperature and solar irradiance reaching the surface can be calculated.

[0085] like Figure 2 As shown, Figure 2 Steps S1201 to S1202 schematically illustrate the flowchart of the atmospheric correction process for mid-infrared entrance pupil radiance provided by the present invention.

[0086] Step S1201: Obtain product data of the mid-infrared hyperspectral entrance pupil radiance product and the corresponding temperature and humidity atmospheric profile at the time of product acquisition, and set the basic parameters of the radiative transfer model. Product data includes observation time, observation angle, and solar angle, etc.

[0087] Step S1202 involves inputting the angle, time, atmospheric profile, and basic parameters into the radiative transfer model, performing atmospheric correction, and obtaining the corrected data. The corrected data includes parameters such as surface brightness temperature, solar irradiance, and downward atmospheric radiation.

[0088] Through embodiments of the present invention, a mid-infrared hyperspectral physical-empirical model is established based on the difference in direct solar radiation between channels to invert the equivalent brightness temperature and biaxial reflectance of the Earth's surface. Following basic concepts of remote sensing physics, the mid-infrared radiative transfer equation is denoted as:

[0089] ;

[0090] In the formula, L is the radiance of the sensor entrance pupil; and These are the observed zenith angle and the observed azimuth angle, respectively. Wavelength; Atmospheric transmittance along the path from the Earth's surface to the sensor; Surface radiance; This is due to the atmospheric thermal radiation itself. This refers to solar radiation backscattered by the atmosphere.

[0091] in, Further writing:

[0092] ;

[0093] The above formula, and These are the solar zenith angle and the solar azimuth angle, respectively. The surface reflectance; To reach the Earth's surface, To reach the solar irradiance at the top of the atmosphere, denoted as , where is the atmospheric transmittance along the path from the sun to the Earth's surface, and cos is the cosine function. and These are surface temperature and surface emissivity, respectively. This is due to the downward-moving atmospheric thermal radiation. This refers to solar radiation that is scattered forward from the atmosphere to the Earth's surface. This represents the Planck function.

[0094] When the wavelength is fixed, this function realizes the conversion between radiance and brightness temperature (hereinafter referred to as "brightness temperature"):

[0095] ;

[0096] in, To be at wavelength Temperature The corresponding radiance, For any temperature; and Both are Planck's constants, respectively and exp represents an exponential function. In some embodiments of this disclosure, unless otherwise specified, the unit of irradiance is . The unit of radiance is The units for brightness temperature and surface temperature are: .

[0097] In some embodiments of this disclosure, atmospheric correction calculations and inversion processes all use the mid-infrared hyperspectral channel as the smallest spectral unit. To obtain the average radiance of each channel, channel convolution needs to be performed on data with a resolution higher than that of the target sensor, i.e.:

[0098] ;

[0099] in, Number the channel; Spectral data for the channel; High-resolution spectral data; It is the spectral response function; and These represent the center wavelength and the full width at half maximum (FWHM), respectively. Typically, the spectral response function of a hyperspectral sensor follows a Gaussian distribution, as shown below:

[0100] ;

[0101] Where exp represents the exponential function and ln represents the logarithmic function.

[0102] Figure 3 This is an exemplary flowchart of the mid-infrared hyperspectral surface two-dimensional reflectance inversion method provided by the present invention. After channel convolution, the mid-infrared radiative transfer equation can be expressed as:

[0103] ;

[0104] ;

[0105] The above formula, For channel Entrance pupil radiance; For channel Atmospheric transmittance along the path from the Earth's surface to the satellite; For channel Atmospheric ascending radiance, which includes the atmosphere's own thermal radiation and the solar radiation backscattered by the atmosphere; and Channels Surface radiance and surface brightness temperature; For channel The corresponding Planck lookup table; and Channels Surface emissivity and surface dichroism; It refers to the surface temperature; For channel The atmospheric downward radiance includes the atmospheric downward radiation itself and the solar radiation scattered forward by the atmosphere. For channel Solar irradiance reaching the Earth's surface.

[0106] The inversion of surface two-dimensional reflectance was carried out using a mid-infrared hyperspectral physical-empirical model. This model mainly utilizes the differences in direct solar radiation between different channels. The establishment and derivation of the model are as follows.

[0107] Assume the equivalent brightness temperature of the Earth's surface Defined as the Earth's surface brightness temperature excluding direct solar radiation, i.e.:

[0108] ;

[0109] in, This is the Planck lookup table for channel i; For channel Equivalent brightness temperature of the Earth's surface; The equivalent radiance of channel i on the ground surface; For the surface brightness temperature of channel i; For channel i, the surface radiance; Let i be the surface diaxial reflectance. Let be the solar irradiance reaching the Earth's surface via channel i.

[0110] Expanding the Planck function with respect to temperature using a second-order Taylor series, we have:

[0111] ;

[0112] ;

[0113] make Combining the two equations above, we get:

[0114] ;

[0115] in, For temperature The first derivative of the Planck function with respect to temperature; This is the temperature baseline.

[0116] like Figure 3 As shown, it includes: step S1301, selecting two additional channels for any one channel to form a three-channel combination, and establishing an inversion model of mid-infrared hyperspectral surface biaxial reflectance based on the three-channel combination through physical derivation.

[0117] make For the three adjacent passages , and ,have:

[0118] ;

[0119] ;

[0120] ;

[0121] Assume that the surface dichroism and the surface equivalent brightness temperature of the three channels have the following linear relationship:

[0122] ;

[0123] ;

[0124] ;

[0125] In the last formula above, , , , Combining the above six equations, we get:

[0126] ;

[0127] make , , , The coefficients in the above formula are:

[0128] ;

[0129] ;

[0130] ;

[0131] When the quadratic term of the brightness temperature difference is used as a higher-order term... During the expansion, the equivalent brightness temperature of the Earth's surface was obtained:

[0132] ;

[0133] in, Let the equivalent brightness temperature of the surface of channel j be the solution. , and These are the surface brightness temperatures for channels i, j, and k, respectively; , , , and These are the fitting coefficients for the initial, first, second, third, and fourth equivalent surface brightness temperatures, respectively. The above formula can be generalized to other channels of the mid-infrared hyperspectral sensor. In practice, a large amount of mid-infrared hyperspectral simulation data is typically used to fit the coefficients.

[0134] Based on the above formula, the equivalent brightness temperature of the Earth's surface can be obtained. And the surface dichroism can be calculated:

[0135] ;

[0136] in, Let the surface diaxial reflectance of channel j be the solution to be found; For channel j, the surface radiance; The equivalent radiance of the ground surface for channel j; Let be the solar irradiance reaching the Earth's surface via channel j.

[0137] In summary, by utilizing the differences in solar irradiance between three adjacent channels and based on the assumption of a linear relationship between the surface diaxial reflectance and the equivalent brightness temperature of the surface in the three channels, the mid-infrared hyperspectral surface diaxial reflectance inversion model solves the underdetermined problem in the inversion: the N channel observation equations are solved to obtain 2N unknowns (N surface diaxial reflectances and N surface equivalent radiance / brightness temperature) to achieve preliminary separation of surface reflected radiation and emitted radiation.

[0138] Step S1302: The data in the radiation database is used as input to the radiative transfer model. A large amount of data is simulated using the radiative transfer model to obtain simulated data. The simulated data is then used to fit the mid-infrared hyperspectral surface two-dimensional reflectance inversion model to obtain fitted data. The fitted data is then used to fit the mid-infrared hyperspectral surface two-dimensional reflectance inversion model to obtain model coefficients, thus obtaining the mid-infrared hyperspectral surface two-dimensional reflectance inversion framework. The radiation database includes a surface emissivity and two-dimensional reflectance spectral library and an atmospheric profile library; the simulated data includes simulated surface brightness temperature and simulated solar irradiance.

[0139] The empirical coefficients of the mid-infrared hyperspectral surface biaxial reflectance inversion model are obtained as follows: Based on an atmospheric radiative transfer model (e.g., MODTRAN), and with the support of the TIGR atmospheric profile database, the surface temperature, surface emissivity, surface biaxial reflectance, solar angle, and satellite observation angle are reasonably set. Combined with the satellite sensor channel response function, the satellite entrance pupil radiance / brightness temperature, surface radiance / brightness temperature, and equivalent surface radiance / brightness temperature are simulated. The simulated data covers mid-infrared radiative transfer under different surface, atmospheric, solar, and observation conditions to ensure the universality of the surface biaxial reflectance inversion algorithm. Using massive simulated surface brightness temperature and equivalent surface brightness temperature data, the fitting coefficients are determined through multivariate least squares regression fitting.

[0140] The TIGR atmospheric profile database contains atmospheric temperature profiles, water vapor profiles, ozone profiles, and other data for polar, tropical, and mid-latitude regions; surface temperature is set with reference to the bottom temperature of the atmospheric profile; surface emissivity is selected from the ASTRE2.0 ground object emissivity spectral library, which includes emissivity spectra of various typical ground objects such as water bodies, vegetation, soil, rocks, and man-made objects.

[0141] Step S1303: The satellite-measured surface brightness temperature and solar irradiance output by atmospheric correction are used as input mid-infrared hyperspectral surface diaxial reflectance inversion framework. The equivalent radiance of the surface is calculated using the mid-infrared hyperspectral surface diaxial reflectance inversion framework, and then the surface diaxial reflectance of each channel is calculated.

[0142] Using the equivalent brightness temperature of the land surface obtained from the above inversion as the initial input, the TES algorithm can be used to simultaneously invert land surface temperature and land surface emissivity. The TES algorithm, based on assumptions or externally introduced empirical relationships / constraints, solves the underdetermined problem in the separation process of land surface temperature / emissivity: solving N+1 unknowns (N land surface emissivity and 1 land surface temperature) using N observation equations. Currently widely used TES algorithms include ASTER TES and piecewise linear TES. Considering the applicable scenarios and inversion accuracy of different TES algorithms, when there is prior emissivity spectral database data, dictionary-based sparse representation for emissivity (DSRE) TES is preferred for simultaneous inversion of land surface temperature and emissivity; if there is no prior emissivity spectral database, piecewise linear (LSEC) TES is used for inversion. The aforementioned adaptive TES approach can cover most application scenarios. Furthermore, the two TES algorithms can complement each other; the inversion results of LSEC TES can expand the training library and dictionary capacity of DSRE TES, while DSRE TES can reconstruct the emissivity with a high signal-to-noise ratio through multi-atom rearrangement, aiming to achieve better separation results. Figure 4 As shown, the process of the scene adaptive framework integrating multiple methods for separating surface temperature and surface emissivity provided by this invention includes:

[0143] Step S1401: Input the equivalent surface radiance and atmospheric downdraft radiation of each channel of the mid-infrared hyperspectral spectrum to establish a scene-adaptive mid-infrared surface temperature and surface emissivity separation framework. Determine whether the target scene has a priori spectral library; the target scene refers to a specific area containing specific land cover composition and environmental characteristics observed by mid-infrared hyperspectral remote sensing.

[0144] Step S1402: If the target scene has a priori spectral library, then the surface temperature and surface emissivity are obtained by using the dictionary sparse representation-based surface temperature and surface emissivity separation algorithm (i.e., DSRE TES).

[0145] Figure 5 An exemplary flowchart for constructing an emissivity sparse dictionary to separate surface temperature and surface emissivity provided by the present invention is shown below. Figure 5 As shown, the process of constructing an emissivity sparse dictionary to separate surface temperature and surface emissivity includes:

[0146] Step S14021 involves convolving the emissivity spectra from the spectral library into each channel, and then performing sparse coding and dictionary training using the Orthogonal Matching Pursuit (OMP) and K-Singular Value Decomposition (K-SVD) algorithms to obtain the emissivity dictionary. A channel refers to the smallest unit for acquiring radiation data from a mid-infrared hyperspectral entrance pupil radiance product. The DSRE TES implementation process consists of two parts: emissivity dictionary training and emissivity reconstruction. Based on the dictionary learning principle, emissivity dictionary training requires sparse coding and dictionary atom updates. The sparse coding uses the Orthogonal Matching Pursuit (OMP) algorithm, and the dictionary atom updates use the K-Singular Value Decomposition (K-SVD) algorithm. The emissivity dictionary training mainly involves the following steps:

[0147] Step 1: Pre-encode the emissivity spectral library, i.e.:

[0148] ;

[0149] ;

[0150] ;

[0151] In the formula, n is the spectral number; N is the total number of mid-infrared hyperspectral spectra; This is the original channel-type mid-infrared emissivity spectrum; This is the emissivity ratio spectrum; mean indicates taking the average value. The emissivity empirical spectrum; L=N+2 is the length of a single empirical emissivity spectrum; max indicates taking the maximum value; min indicates taking the minimum value; is an empirical spectral library; J is the total number of emissivity spectra.

[0152] Step 2: Dictionary Atom Optimization and Update. Iteratively update the dictionary atoms to obtain the emissivity dictionary. That is, in Under compressed observation, this dictionary is included in the dataset. The above has the best sparse representation, namely:

[0153] ;

[0154] in, The optimal sparse representation is the emissivity dictionary of the training output; For emission rate identification; for and Joint optimization minimizes the objective function; For dictionary identifiers; Identifier for sparse coefficient matrix; Represents the L2 norm; The signal to be trained; A dictionary of atoms to be updated; For the training signal sparse coefficient matrix; This indicates that the condition is being met; It is the zero norm; for The nth sparse coefficient column vector of length L; Sparsity;

[0155] Observation matrix Defined as:

[0156] ;

[0157] in:

[0158] ;

[0159] ;

[0160] ;

[0161] in, The observation matrix; It is an augmented matrix of the identity matrix; The empirical observation matrix; For Hadema; It is a unit vector of length L; For normalized row vectors; It is the identity matrix; It is a zero matrix; Let J be a unit vector of length J; The initial signal to be compressed for observation; Signal to be trained The pseudo-inverse matrix; It is a 2-norm; This is a transpose operation; for The Middle A column vector of length L; To be The operation of combining column vectors into a single matrix;

[0162] When the surface emissivity in the radiative transfer equation is compressed and expressed sparsely, the surface emissivity is reconstructed using an emissivity dictionary, and the surface temperature and surface emissivity are obtained by solving for them:

[0163] ;

[0164] ;

[0165] In the formula, Represents Haderma; The equivalent radiance vector of the Earth's surface; It is a wavelength vector; This is the downward atmospheric radiation vector; The emissivity sparsity coefficient; The sparse coefficients are unknown; It is the Planck function; The radiance representing the Earth's surface temperature at various wavelengths; Let N be a vector consisting entirely of 1s. This means minimizing the objective function; This is the reconstructed emissivity value; This represents a function that reconstructs the sparse coefficients as the surface emissivity; Let be the surface temperature in each iteration.

[0166] ;

[0167] ;

[0168] in, For the reconstructed emissivity ratio spectrum; This represents the maximum value of the emissivity spectrum. This represents the minimum value of the emissivity spectrum. Represents taking the first vector. Up to N elements; This is the reconstructed emissivity spectrum used in the solution; max represents taking the maximum value; min represents taking the minimum value.

[0169] Under normal circumstances, Therefore, the number of unknowns (sparse coefficients) The number of non-zero values ​​and the surface temperature are much smaller than the number of equations. The equations are solved, and the surface temperature and surface emissivity can be obtained by inversion.

[0170] Step S14022: Given an initial value for the surface temperature, reconstruct the surface emissivity using the emissivity dictionary, establish a first cost function, and stop iterating when the first cost function is less than a first threshold or the number of iterations is greater than a limit in a certain iteration, and output the final surface temperature and surface emissivity.

[0171] Since the solution to the unknowns cannot be obtained directly, iterative solutions are required: First, given an initial surface temperature value, calculate an initial emissivity value, typically the maximum or average of the equivalent brightness temperature of the surface. Second, reconstruct the emissivity using an emissivity dictionary to reduce errors in the emissivity. Third, establish a cost function; if the cost function is less than a threshold... Alternatively, if the iteration count exceeds 12, the iteration terminates, outputting the surface temperature and reconstructed emissivity as the inversion results. If the iteration cannot terminate, the reconstructed emissivity, equivalent surface radiance, and downward atmospheric radiance are used as inputs to perform a golden section search to find a reasonable initial value for the surface temperature, and the iteration restarts. The first cost function is:

[0172] ;

[0173] in, The first cost function; The square of the L2 norm; This is the surface emissivity vector after dictionary reconstruction; To reconstruct the identifier; The characteristic radiance of the initial surface temperature at various wavelengths; This represents the initial value of the surface temperature in each iteration; This is the equivalent radiance vector of the Earth's surface. If the iteration terminates, the output is... and This serves as the final inversion result.

[0174] Step S1403: If the target scene does not have a prior emissivity spectrum library, and if the surface temperature and surface emissivity are to be inverted simultaneously, the surface temperature and surface emissivity are obtained by using the piecewise linear constraint surface temperature and surface emissivity separation algorithm (LSEC TES).

[0175] Figure 6 This is an exemplary flowchart illustrating the separation of surface temperature and surface emissivity based on piecewise linear constraints provided by the present invention. Figure 6 As shown, the process for separating surface temperature and surface emissivity based on piecewise linear constraints includes:

[0176] LSEC TES divides the emissivity spectrum into M segments, each segment covering With M channels, the entire spectrum can be fitted by M straight lines, meaning the piecewise emissivity spectrum can be represented by a set of linear equations:

[0177] ;

[0178] in, For channel indexing; For channel The center wavelength; For channel Emission rate; The slope of the emissivity segment; is the emissivity segment intercept; k is the index of the emissivity segment fitted line; M is the total number of emissivity segment fitted lines; and These represent the number of channels in segment K and segment K+1, respectively. This represents the range of channel indices within each segment. It should generally be greater than or at least equal to 3, that is... Therefore, within a linear function, three or more emissivity unknowns can be reduced to two univariate linear function coefficients, thereby reducing the number of unknowns.

[0179] After atmospheric correction and surface two-dimensional reflectance inversion, the equivalent surface radiance and atmospheric downdraft radiance are known. Substituting the above equation into the mid-infrared radiative transfer equation, the surface emissivity can be obtained:

[0180] ;

[0181] in, The equivalent brightness temperature of the surface in the channel The radiance at that location; For the surface temperature in the channel The radiance at that location; For channel The downward atmospheric radiation; based on the above formula, the introduction of piecewise linear fitting makes the underdetermined problem have a deterministic solution, and N equations can solve for 2M+1 unknowns (the coefficients of 2M piecewise linear equations and 1 surface temperature).

[0182] However, since the unknown cannot be directly expressed using analytical formulas... , and Therefore, iterative solutions are needed to optimize and approximate the unknowns. In LSEC TES, the iterative process is similar to DSRE TES. First, the second cost function is defined:

[0183] ;

[0184] In the formula, This is the second cost function; N is the channel index; N is the total number of channels; The initial temperature value in each iteration is generally set to the maximum or average value of the equivalent brightness temperature of the Earth's surface; The emissivity reconstructed from the coefficients of a system of univariate linear equations in each iteration; The initial value of the surface temperature in the channel The radiance at that location.

[0185] Step S14031: Given an initial temperature value, estimate the fitting coefficients of the piecewise linear equation for emissivity to obtain the piecewise emissivity spectrum; use the piecewise linear equation to calculate the piecewise emissivity spectrum to obtain the surface emissivity; establish a second cost function; when the second cost function is less than the second threshold or the number of iterations is greater than the limit in a certain iteration, stop the iteration and output the final surface temperature and surface emissivity.

[0186] The iterative steps can be detailed as follows: Set an initial value for the surface temperature; directly solve for the initial value of the surface emissivity for each channel based on the initial value of the surface temperature; use the initial value of the surface emissivity for each channel to fit the coefficients of a univariate equation system about the wavelength piecewise; reconstruct the emissivity using the coefficients of the univariate equation system and calculate the cost function; if the value of the cost function is less than... If the number of iterations exceeds 20, the iteration terminates and the output is... and As the final inversion values ​​of surface temperature and surface emissivity; if the iteration does not terminate, then based on the principle of near-surface radiative transfer, the equivalent surface brightness temperature, atmospheric downdraft radiance, and reconstructed surface emissivity are used to search for a new initial value of surface temperature using the Newton-Raphson iteration method, and this initial value is used as the starting point to iterate again.

[0187] In step S1404, the new surface emissivity output by the piecewise linear constraint method can expand the training library and capacity of the emissivity dictionary. Based on the dictionary sparse representation method, more accurate inversion results can be obtained through atomic rearrangement.

[0188] Once the surface temperature, mid-infrared hyperspectral surface dichroism, and surface emissivity are obtained through inversion, surface emitted radiation and reflected radiation can be accurately separated. Surface emitted radiation is... , and These are surface temperature and surface emissivity, respectively; surface reflected radiation is... , This is downward atmospheric radiation.

[0189] Based on the disclosed method, the present invention also provides an integrated inversion device for mid-infrared hyperspectral emission and reflection characteristics, which will be described below in conjunction with... Figure 7 The device is described in detail below. Another aspect of the invention provides an integrated inversion device for mid-infrared hyperspectral emission and reflectance characteristics, comprising: a module for mid-infrared hyperspectral satellite entrance pupil radiance quality evaluation, quality optimization, and data screening; an atmospheric correction module; a module for simultaneous inversion of mid-infrared hyperspectral surface two-dimensional reflectance and surface equivalent brightness temperature; and a target scene-adaptive surface temperature / emissivity separation module.

[0190] Figure 7 This is an exemplary structural diagram of an integrated inversion device for mid-infrared hyperspectral emission and reflection characteristics provided by the present invention. Figure 7 As shown, the integrated inversion device 700 for mid-infrared hyperspectral emission and reflection characteristics provided in this application includes a satellite entrance pupil data quality evaluation and optimization module 710, a land surface atmospheric radiance and solar irradiance calculation module 720, a land surface bidirectional reflectance inversion module 730 based on a physical-empirical model, and a scene adaptive temperature emissivity simultaneous inversion module 740.

[0191] The satellite entrance pupil data quality evaluation and optimization module 710 is used to acquire mid-infrared hyperspectral satellite remote sensing data, perform cloud masking and bad value removal on the entrance pupil radiance data, and if the payload has not been calibrated for a long time, a recalibration operation is required to optimize the data quality.

[0192] The surface atmospheric radiance and solar irradiance calculation module 720 is used to perform atmospheric correction on mid-infrared hyperspectral entrance pupil radiance data using a radiative transfer model, obtain surface brightness temperature and downward atmospheric radiation, and calculate the solar irradiance reaching the surface.

[0193] The surface biaxial reflectance inversion module 730, based on a physical-empirical model, is used to construct a mid-infrared hyperspectral surface biaxial reflectance inversion method based on empirical assumptions and physical derivations, and to solve the underdetermined solution problem in order to achieve the separation of surface biaxial reflectance and surface equivalent radiance.

[0194] The scene-adaptive temperature and emissivity simultaneous inversion module 740 is used to construct a scene-adaptive framework. If the scene has a priori spectral library, the sparse dictionary reconstruction method is selected; otherwise, the piecewise linear constraint method is selected to achieve simultaneous inversion of surface temperature and surface emissivity.

[0195] It should be noted that the embodiments of the device part are similar to those of the aforementioned method part, and the technical effects achieved are also similar. For specific details, please refer to the aforementioned method embodiment part, which will not be repeated here.

[0196] According to embodiments of the present invention, any multiple of the following modules can be combined into one module: the satellite entrance pupil data quality evaluation and optimization module 710, the surface atmospheric radiance and solar irradiance calculation module 720, the surface biaxial reflectance inversion module 730 based on a physical-empirical model, and the scene adaptive temperature emissivity simultaneous inversion module 740. Alternatively, any one of these modules can be split into multiple modules. Or, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of the present invention, at least one of the satellite entrance pupil data quality evaluation and optimization module 710, the surface atmospheric radiance and solar irradiance calculation module 720, the surface biaxial reflectance inversion module 730 based on a physical-empirical model, and the scene adaptive temperature emissivity simultaneous inversion module 740 can be at least partially implemented as a hardware circuit, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a packaged system, an application-specific integrated circuit (ASIC), or any other reasonable method of integrating or packaging the circuit, or implemented in hardware or firmware, or in any one of the three implementation methods of software, hardware, and firmware, or in a suitable combination of any of them. Alternatively, at least one of the satellite entrance pupil data quality evaluation and optimization module 710, the surface atmospheric radiance and solar irradiance calculation module 720, the surface biaxial reflectance inversion module 730 based on a physical-empirical model, and the scene adaptive temperature emissivity simultaneous inversion module 740 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0197] The above are merely preferred embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for retrieving emission and reflectance properties of mid-infrared hyperspectral data, characterized in that, include: Mid-infrared hyperspectral satellite remote sensing data was acquired, and the entrance pupil radiance data was optimized to obtain optimized entrance pupil radiance data. Mid-infrared hyperspectral satellite remote sensing data includes entrance pupil radiance data; Atmospheric correction was performed on the optimized entrance pupil radiance data using a radiative transfer model to obtain the surface brightness temperature and downward atmospheric radiation, and the solar irradiance reaching the surface was calculated. A mid-infrared hyperspectral surface biaxial reflectance inversion framework was constructed, and surface biaxial reflectance inversion was performed based on the framework to obtain surface biaxial reflectance and equivalent surface radiance, including: For each channel, a three-channel combination is constructed, and an inversion model for mid-infrared hyperspectral surface biaxial reflectance based on the three-channel combination is established; Simulated data were obtained by simulating data in the radiation database using a radiative transfer model. The radiation database includes a spectral library of surface emissivity, diaxial reflectivity, and an atmospheric profile library. The simulated data includes simulated surface brightness temperature and simulated solar irradiance. The inversion model of mid-infrared hyperspectral surface biaxial reflectance was fitted based on simulation data to obtain the mid-infrared hyperspectral surface biaxial reflectance inversion framework. By inputting the surface brightness temperature and solar irradiance into the mid-infrared hyperspectral surface diaxial reflectance inversion framework, the equivalent surface radiance and surface diaxial reflectance of each channel are obtained. A scene-adaptive framework is constructed, and surface temperature and surface emissivity are obtained by inversion based on the scene-adaptive framework; including the sparse dictionary reconstruction method if the scene has a prior spectral library, and the piecewise linear constraint method otherwise, so as to achieve simultaneous inversion of surface temperature and surface emissivity.

2. The integrated retrieval method of mid-infrared hyperspectral emission and reflectance characteristics according to claim 1, characterized in that, The optimization process includes cloud masking, bad value removal, and / or recalibration.

3. The integrated retrieval method of mid-infrared hyperspectral emission and reflectance characteristics according to claim 1, characterized in that, The process of using a radiative transfer model to perform atmospheric correction on the optimized entrance pupil radiance data to obtain surface brightness temperature and downward atmospheric radiation, and calculating the solar irradiance reaching the surface, includes: Acquire product data for mid-infrared hyperspectral entrance pupil radiance products and corresponding temperature and humidity atmospheric profiles at the time of product acquisition; product data includes observation time, observation angle, and solar angle; Product data and atmospheric temperature and humidity profiles are input into the radiative transfer model for atmospheric correction to obtain corrected data; the corrected data includes surface brightness temperature, atmospheric downdraft radiation, and solar irradiance.

4. The integrated retrieval method of mid-infrared hyperspectral emission and reflectance characteristics according to claim 1, characterized in that, The equivalent radiance of the Earth's surface is defined as: ； wherein, is the Planck look-up table for channel i; is the surface albedo for channel i; is the surface equivalent black body temperature; is the surface equivalent radiance for channel i; is the surface black body temperature for channel i; is the surface radiance for channel i; is the surface bidirectional reflectance for channel i; is the solar irradiance reaching the surface for channel i; The equivalent brightness temperature of the Earth's surface is defined as: ； wherein, is the surface equivalent black body temperature of the channel j to be solved; , and are the surface black body temperatures of the channels i, j and k, respectively; , , , and are the initial, first, second, third and fourth surface equivalent black body temperature fitting coefficients, respectively. The surface dichroism is defined as: ； wherein, is the bidirectional reflectance of the surface of the channel j to be solved; is the radiance of the surface of the channel j; is the equivalent radiance of the surface of the channel j; is the solar irradiance reaching the surface of the channel j.

5. The integrated retrieval method of mid-infrared hyperspectral emission and reflectance characteristics according to claim 1, characterized in that, The construction of the scene adaptive framework, and the inversion of surface temperature and surface emissivity based on the scene adaptive framework, includes: Determine whether the target scene has a priori spectral library; the target scene refers to a specific area observed by mid-infrared hyperspectral remote sensing that contains specific land cover composition and environmental characteristics; If the target scene has a prior spectral library, then the surface temperature and surface emissivity are obtained by using a dictionary-based sparse representation algorithm for separating surface temperature and surface emissivity; the dictionary refers to the emissivity dictionary. If the target scene does not have a prior spectral library, the surface temperature and surface emissivity are obtained by inverting the algorithm for separating surface temperature and surface emissivity with piecewise linear constraints. The new surface emissivity output by the surface temperature and surface emissivity separation algorithm based on piecewise linear constraint expands the training library and capacity of the emissivity dictionary, and the new inversion result is obtained by atomic rearrangement based on the dictionary sparse representation method.

6. The integrated retrieval method of mid-infrared hyperspectral emission and reflectance characteristics according to claim 5, characterized in that, The surface temperature and surface emissivity are inverted by using the surface temperature and surface emissivity separation algorithm based on dictionary sparse representation, and the surface temperature and surface emissivity are obtained, comprising: The emissivity spectrum in the spectral library is convolved to each channel, and sparse coding and dictionary training are performed by using the orthogonal matching pursuit method and the K singular value decomposition method to obtain an emissivity dictionary; the channel refers to the minimum unit of the radiation data obtained by the mid-infrared hyperspectral entrance pupil radiance product; Given the initial value of the surface temperature, the surface emissivity is reconstructed by using the emissivity dictionary, and a first cost function is established; The first cost function is repeatedly established until the first cost function is less than a first threshold value or the iteration number is greater than a limit value, the iteration is stopped, and the final surface temperature and surface emissivity are output.

7. The integrated retrieval method of mid-infrared hyperspectral emission and reflectance characteristics according to claim 6, characterized in that, The dictionary training includes spectral library pre-encoding and dictionary atom optimization updating; The spectral library pre-encoding is defined as: ； ； ； where n is the spectral number; N is the total number of mid-infrared hyperspectral; is the emissivity ratio spectrum; is the original channel mid-infrared emissivity spectrum; mean indicates taking the mean value; is the emissivity empirical spectrum; L is the length of a single emissivity empirical spectrum; max indicates taking the maximum value; min indicates taking the minimum value; is the emissivity empirical spectrum library; J is the total number of emissivity spectra; The dictionary atom optimization updating is defined as: ； in, The optimal sparse representation is the emissivity dictionary of the training output; For emission rate identification; for and Joint optimization minimizes the objective function; For dictionary identifiers; Identifier for sparse coefficient matrix; Represents the L2 norm; The signal to be trained; A dictionary of atoms to be updated; For the training signal sparse coefficient matrix; This indicates that the condition is being met; It is the zero norm; for The Middle A length of A sparse coefficient column vector; Sparsity; The observation matrix is defined as: ； ； ； ； in, The observation matrix; It is an augmented matrix of the identity matrix; This is the empirical observation matrix; For Hadema; It is a unit vector of length L; Normalized row vectors; It is the identity matrix; It is a zero matrix; Let J be a unit vector of length J; The initial signal to be compressed for observation; Signal to be trained The pseudo-inverse matrix; It is a norm 2; This is a transpose operation; for The Middle A column vector of length L; To be The operation of combining column vectors into a single matrix; The surface temperature and surface emissivity are solved by reconstructing the surface emissivity by using the emissivity dictionary, and the surface temperature and surface emissivity are defined as: ； ； wherein represents the Hadamard product; is the surface equivalent radiance vector; is the wavelength vector; is the atmospheric downwelling radiance vector; is the emissivity sparsity coefficient; is the sparsity coefficient unknown; is the Planck function; is the surface temperature represented by a characteristic radiance at each wavelength; is an all-ones vector of length N; denotes minimizing the objective function; is the emissivity reconstruction value; represents the function of reconstructing the sparsity coefficient to the surface emissivity; is the surface temperature in each iteration; ； ； wherein is the reconstructed emissivity ratio spectrum; is the emissivity spectrum maximum; is the emissivity spectrum minimum; represents the i-th element of the orientation vector to N elements; is the reconstructed emissivity spectrum; max denotes taking the maximum value; min denotes taking the minimum value; The first cost function is defined as: ； wherein, is a first cost function; is a square of a two-norm; is a surface emissivity vector after dictionary reconstruction; is a reconstruction indicator; is a surface temperature initial value at each wavelength; is a surface temperature initial value in each iteration; is a surface equivalent brightness vector.

8. The integrated retrieval method of mid-infrared hyperspectral emission and reflectance characteristics according to claim 5, characterized in that, The surface temperature and surface emissivity are inverted by using the surface temperature and surface emissivity separation algorithm based on piecewise linear constraint, and the surface temperature and surface emissivity are obtained, comprising: Given the initial value of the surface temperature, the fitting coefficients of the piecewise linear equation of the emissivity are determined to obtain a piecewise emissivity spectrum; The surface emissivity is calculated by using the piecewise emissivity spectrum, and a second cost function is established; The second cost function is repeatedly established until the second cost function is less than a second threshold value or the iteration number is greater than a limit value, the iteration is stopped, and the final surface temperature and surface emissivity are output.

9. The integrated retrieval method of mid-infrared hyperspectral emission and reflectance characteristics according to claim 8, characterized in that, The piecewise emissivity spectrum is defined as: ； wherein, is the center wavelength of the channel; is the center wavelength of the channel is the center wavelength of the channel; is the emissivity of the channel is the emissivity of the channel; is the emissivity segment slope; is the emissivity segment intercept; k is the index of the emissivity segment fit line; M is the total number of emissivity segment fits; and is the number of channels in the Kth segment and K+1th segment, respectively; is the range of channel indices in each segment; The surface emissivity is defined as: ； wherein, is the representative radiance of the surface equivalent black body at channel ; is the representative radiance of the surface temperature at channel ; is the atmospheric downwelling radiation at channel ; The second cost function is defined as: ； wherein, is a second cost function; is a channel index; N is the total number of channels; is the emissivity reconstructed from the coefficient of the linear equation system in each iteration; is the surface temperature initial value at channel is the represented radiance at channel is the surface temperature initial value in each iteration.

Images