Method for reconstructing mid-wave infrared hyperspectral radiance

Abstract

The present application relates to the technical field of computing spectroscopy, and particularly relates to a method for reconstructing medium-wave infrared hyperspectral radiance. The method comprises the following steps: S1, constructing a hyperspectral radiance reconstruction model based on Planck's law of black-body radiation and compressed sensing theory; S2, solving a hyperspectral overcomplete dictionary based on a hyperspectral radiance sample and ODL; S3, constructing a spectral transmittance matrix based on the spectral transmittance corresponding to each channel broadband filter included in a multispectral measurement device, and obtaining a multispectral measurement matrix of the multispectral measurement device based on the spectral transmittance matrix and the hyperspectral overcomplete dictionary; S4, solving a sparse coefficient matrix approximation value based on the multispectral measurement matrix of the multispectral measurement device and a set of radiance measurement values of a target to be measured, and reconstructing a hyperspectral radiance value based on the hyperspectral overcomplete dictionary and the sparse coefficient matrix approximation value. The present application enhances the optical detection and recognition capability for high dynamic targets.

Description

Technical Field

[0001] This invention relates to the technical field of computational spectroscopy, and more particularly to a method for reconstructing the radiance of mid-wave infrared hyperspectral radiation. Background Technology

[0002] Infrared targets possess different brightness temperature "fingerprints" at different wavelengths, making them effectively detectable and identifiable through infrared hyperspectral information. Traditional infrared detection methods mostly employ detection systems based on total energy or specific spectral bands, which suffer from low detection efficiency, insufficient energy utilization, and poor identification capabilities. Furthermore, traditional hyperspectral detection techniques often sacrifice spatial or temporal resolution to acquire target hyperspectral information. Summary of the Invention

[0003] In view of this, the present invention aims to provide a method for reconstructing the radiance of mid-wave infrared hyperspectral radiation, in order to solve the problem that traditional hyperspectral equipment cannot simultaneously achieve temporal and spatial resolution. The present invention proposes to reconstruct infrared hyperspectral radiation information using the principle of compressed sensing. This invention is based on infrared computational spectroscopy technology, aiming to overcome the limitations of traditional computational spectroscopy technology in the spatiotemporal dimensions and enhance the optical detection and identification capabilities of highly dynamic targets.

[0004] To achieve the above objectives, the technical solution created by this invention is implemented as follows:

[0005] A method for reconstructing mid-wave infrared hyperspectral radiance specifically includes the following steps:

[0006] S1: A hyperspectral radiance reconstruction model is constructed based on Planck's radiation law and compressed sensing theory;

[0007] S2: Solving the hyperspectral overcomplete dictionary based on hyperspectral radiance samples and ODL;

[0008] S3: Construct a spectral transmittance matrix based on the spectral transmittance of each channel broadband filter contained in the multispectral measurement device, and obtain the multispectral measurement matrix of the multispectral measurement device based on the spectral transmittance matrix and the hyperspectral overcomplete dictionary.

[0009] S4: Solve the sparse coefficient matrix approximation value based on the multispectral measurement matrix of the multispectral measurement equipment and the set of radiance measurement values ​​of the target to be measured, and input the hyperspectral overcomplete dictionary and the sparse coefficient matrix approximation value into the hyperspectral radiance reconstruction model to reconstruct the hyperspectral radiance value.

[0010] Furthermore, step S1 specifically includes:

[0011] S11: Radiance based on Planck's radiation law Represented as:

[0012] (1);

[0013] in, The first radiation constant, The second radiation constant, wavelength The surface emissivity of the object at that location, T is the temperature of the target being measured;

[0014] S12: The measured radiance of the target in the t-th spectral channel at temperature T. Represented as:

[0015] (2);

[0016] Where t is the number of spectral channels measured. In spectral channels wavelength At this point, the spectral transmittance of the multispectral measurement device is given, and Δλ is the difference between the upper and lower limits of integration.

[0017] S13: Define the hyperspectral radiance emitted by objects in nature. , The measured radiance values ​​for each wideband are Discretizing equation (2) gives:

[0018] (3);

[0019] in, This is the spectral transmittance matrix of a multispectral measurement device. For the first The broadband band corresponds to the first Spectral transmittance at each wavelength;

[0020] S14: Based on compressed sensing theory, utilizing a hyperspectral overcomplete dictionary With sparse coefficient matrix Constructing a hyperspectral radiance reconstruction model:

[0021] (4);

[0022] (5);

[0023] in, For high spectral radiance, For the first The corresponding wavelength is the th The spectral radiance characteristic value of each atom;

[0024] make Combining equations (3) and (4), we construct the expression for the set of radiance measurements and the multispectral measurement matrix:

[0025] (6);

[0026] Where y is the set of radiance measurements. This represents a multispectral measurement matrix.

[0027] Furthermore, step S2 specifically includes:

[0028] S21: Select wavelength points in the mid-wave infrared band of 3.25um~5.20um according to the set step size to obtain the first wavelength point set;

[0029] S22: Set the range of temperature parameters and set the emissivity of the object surface to 1. Based on Planck's radiation law, calculate the spectral radiance of each wavelength point in the first wavelength point set with a temperature parameter of 1k step size; and construct a hyperspectral radiance sample based on the spectral radiance of different wavelength points at different temperatures.

[0030] S23: Input the hyperspectral radiance samples into the ODL algorithm for dictionary learning to obtain the hyperspectral overcomplete dictionary D.

[0031] Furthermore, step S3 specifically includes:

[0032] S31: Based on the first wavelength point set, obtain the spectral transmittance of each broadband filter included in the multispectral measurement device, and construct a spectral transmittance matrix using the spectral transmittance of each channel.

[0033] S32: Couple the spectral transmittance matrix with the hyperspectral overcomplete dictionary to obtain the multispectral measurement matrix of the multispectral measurement device;

[0034] Based on the radiance of the target at each wavelength point in each spectral channel and the spectral transmittance of the target at each wavelength point in each spectral channel, the radiance measurement value at each wavelength point is calculated and discretized based on equation (2) to obtain the set of radiance measurement values ​​of the target.

[0035] Furthermore, step S4 specifically includes the following steps:

[0036] S41: Input the set of radiance measurements of the target and the multispectral measurement matrix of the multispectral measurement device into the Bach-OMP algorithm for calculation, and output the approximation value of the sparse coefficient matrix. ;

[0037] S42: Based on the hyperspectral radiance reconstruction model, the reconstructed hyperspectral radiance value is obtained by solving the following formula:

[0038] (7);

[0039] in, For the reconstructed hyperspectral brightness values, D is the hyperspectral overcomplete dictionary. This is the approximation value for the sparse coefficient matrix.

[0040] Compared with the prior art, the present invention can achieve the following beneficial effects:

[0041] (1) The mid-wave infrared hyperspectral radiance reconstruction method described in this invention solves the pain points of traditional infrared detection and improves detection and identification capabilities: it breaks through the limitations of traditional detection systems based on total energy or specific spectral bands, avoiding the problems of low detection efficiency and insufficient energy utilization; at the same time, it solves the defects of traditional hyperspectral detection that sacrifices spatial or temporal resolution, and can effectively obtain the "fingerprint" information of infrared target brightness temperature, thereby enhancing the optical detection and identification capabilities of highly dynamic targets.

[0042] (2) The mid-wave infrared hyperspectral radiance reconstruction method described in this invention has high reconstruction accuracy and reliable performance: By reconstructing the hyperspectral radiance and emissivity of four materials, the SAM values ​​are close to 0 rad, the shape similarity is high, and the maximum relative error is better than 2%; and it is less affected by emissivity, and can be reconstructed well even when the emissivity is unknown, providing accurate data support for subsequent target identification, etc.

[0043] (3) The mid-wave infrared hyperspectral radiance reconstruction method described in this invention uses the ODL algorithm to construct a dictionary, which improves data processing efficiency compared with the traditional K-SVD; the Batch-OMP algorithm pre-stores the Gram matrix, which improves the sparse coding efficiency; and the five-channel design reduces hardware complexity. Attached Figure Description

[0044] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0045] Figure 1 A schematic flowchart of the mid-wave infrared hyperspectral radiance reconstruction method described in the embodiments of the present invention;

[0046] Figure 2 The spectral response curves of the five channels described in the embodiments of the present invention;

[0047] Figure 3 The invention provides a graph showing the mid-wave infrared emissivity values ​​of different materials at different temperatures, as described in the embodiments of the present invention.

[0048] Figure 4 The comparison curve between the reconstructed radiance result and the original value of the present invention is shown in the embodiment of the present invention.

[0049] Figure 5 The comparison curve between the reconstructed emissivity result and the original value described in the embodiment of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not constitute a limitation thereof.

[0051] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0052] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on this invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0053] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0054] The invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0055] like Figure 1 As shown, this invention proposes a method for reconstructing mid-wave infrared hyperspectral radiance, which specifically includes the following steps:

[0056] S1: A hyperspectral radiance reconstruction model is constructed based on Planck's radiation law and compressed sensing theory;

[0057] S2: Solving the hyperspectral overcomplete dictionary based on hyperspectral radiance samples and ODL;

[0058] S3: Construct a spectral transmittance matrix based on the spectral transmittance of each channel broadband filter contained in the multispectral measurement device, and obtain the multispectral measurement matrix of the multispectral measurement device based on the spectral transmittance matrix and the hyperspectral overcomplete dictionary.

[0059] S4: Solve the sparse coefficient matrix approximation value based on the multispectral measurement matrix of the multispectral measurement equipment and the set of radiance measurement values ​​of the target to be measured, and input the hyperspectral overcomplete dictionary and the sparse coefficient matrix approximation value into the hyperspectral radiance reconstruction model to reconstruct the hyperspectral radiance value.

[0060] It should be noted that, firstly, a hyperspectral radiance reconstruction model is derived based on Planck's radiation law. Secondly, to achieve a sparse representation of the mid-wave infrared spectral radiance signal, this invention designs a mid-wave infrared hyperspectral overcomplete dictionary based on the Online Dictionary Learning (OLD) algorithm. First, the mid-wave wavelengths from 3.25µm to 5.20µm are discretized according to a broadband filter step size. Then, at a temperature T ranging from 273K to 5000K and an emissivity of 1, the spectral radiance at different wavelengths at different temperatures is calculated according to Planck's radiation law. Finally, the obtained spectral radiance at different wavelengths at different temperatures is used as sample signals and substituted into the online dictionary learning algorithm to obtain the hyperspectral overcomplete dictionary. The core idea of ​​online dictionary learning is stochastic approximation, that is, processing only one or a small batch of sample data at a time. Compared with traditional K-SVD, ODL significantly improves processing efficiency in terms of data processing. The essence of online dictionary learning is to learn a dictionary by optimizing a non-convex objective function, so that the input spectral signal can be sparsely represented by a small number of atoms in the dictionary. This invention, based on Batch-OMP, indirectly obtains the reconstructed mid-wave infrared hyperspectral radiance values ​​by solving for approximation values ​​of the sparse coefficient matrix. The core idea of ​​Batch-OMP is to pre-calculate and store the Gram matrix of the dictionary, avoiding redundant calculations of inner products between atoms. Compared with traditional OMP algorithms, it utilizes a shared Gram matrix, significantly improving the computational efficiency and accuracy of the sparse coding process. This invention can solve for the spectral radiance values ​​of 175 bands based on five-channel wide-band mid-wave infrared spectral radiance values.

[0061] In some embodiments, step S1 specifically includes:

[0062] S11: Radiance based on Planck's radiation law Represented as:

[0063] (1);

[0064] in, The first radiation constant, The second radiation constant, wavelength The surface emissivity of the object at that location, T is the temperature of the target being measured;

[0065] S12: The measured radiance of the target in the t-th spectral channel at temperature T. Represented as:

[0066] (2);

[0067] Where t is the number of spectral channels measured. In spectral channels wavelength At this point, the spectral transmittance of the spectral transmittance measuring device is Δλ, which is the difference between the upper and lower limits of integration.

[0068] Specifically, the upper and lower limits of integration can be set to 3.25 μm and 5.20 μm, respectively. S13: Define the hyperspectral radiance emitted by objects in nature. , The measured radiance values ​​for each wideband are Discretizing equation (2) gives:

[0069] (3);

[0070] in, This is the spectral transmittance matrix of the spectral transmittance measurement device. For the first The broadband band corresponds to the first Spectral transmittance at each wavelength;

[0071] S14: Based on compressed sensing theory, utilizing a hyperspectral overcomplete dictionary With sparse coefficient matrix Constructing a hyperspectral radiance reconstruction model:

[0072] (4);

[0073] (5);

[0074] in, For high spectral radiance, For the first The corresponding wavelength is the th The spectral radiance characteristic value of each atom;

[0075] It should be noted that when the actual temperature is known and the emissivity of the target is within a given range, the radiance of the target is uniquely determined and can be sparsified by an overcomplete hyperspectral dictionary. Therefore, it is assumed that a certain overcomplete hyperspectral dictionary exists. The hyperspectral radiance of the target being measured is sparse, and the hyperspectral dictionary is comprehensive. Represented as: Additionally, the sparse coefficient matrix , Indicates the first The sparsity coefficients corresponding to each atom.

[0076] make Combining equations (3) and (4), we construct the expression for the set of radiance measurements and the multispectral measurement matrix:

[0077] (6);

[0078] Where y is the set of radiance measurements. Represents the multispectral measurement matrix, derived from a hyperspectral overcomplete dictionary. Spectral response curve of infrared radiation measurement system Obtained by projection.

[0079] It should be noted that, as described above, once the hyperspectral dictionary is complete... Once determined, the spectral response curve of the spectral transmittance measurement device can be obtained. Then, according to formula (6), the multispectral radiance measurement value and... Then the sparsity coefficient can be obtained. Once the sparsity coefficients are determined, the hyperspectral radiance value can be reconstructed using equation (4).

[0080] In some embodiments, step S2 specifically includes:

[0081] S21: Select wavelength points in the mid-wave infrared band of 3.25um~5.20um according to the set step size to obtain the first wavelength point set;

[0082] S22: Set the range of temperature parameters and set the emissivity of the object surface to 1. Based on Planck's radiation law, calculate the spectral radiance of each wavelength point in the first wavelength point set with a temperature parameter of 1k step size; and construct a hyperspectral radiance sample based on the spectral radiance of different wavelength points at different temperatures.

[0083] S23: Input the hyperspectral radiance samples into the ODL algorithm for dictionary learning to obtain the hyperspectral overcomplete dictionary D.

[0084] It should be noted that the ODL algorithm is an existing algorithm. The specific process is as follows: the spectral radiance set is input into ODL for learning, and a hyperspectral overcomplete dictionary is obtained. Initialize sparse constraints and regularization parameters, and pre-allocate sparse autocorrelation cumulative matrices for ODL. Data-coefficient cross-correlation cumulative matrix Define the iteration stopping threshold.

[0085] The objective function is solved using a greedy algorithm to obtain the sparse coefficient matrix. And based on the sparse autocorrelation cumulative matrix Data-coefficient cross-correlation cumulative matrix sparse coefficient matrix Obtain the sparse autocorrelation cumulative matrix Data-coefficient cross-correlation cumulative matrix .

[0086] Greedy algorithms include LARS or Batch-OMP.

[0087] The objective function is:

[0088] (7);

[0089] Where D is the hyperspectral overcomplete dictionary, and n is the total number of samples. It is a sparse coefficient matrix. Let be the spectral radiance at the i-th iteration. Here, i is the regularization parameter, and i is the number of iterations.

[0090] Based on sparse coefficient matrix sparse autocorrelation cumulative matrix Data-coefficient cross-correlation cumulative matrix The block coordinate descent method is used to analyze the hyperspectral overcomplete dictionary. Perform column-by-column updates to obtain a hyperspectral overcomplete dictionary. .

[0091] Determine if hyperspectral overcomplete dictionary If the sparse constraint is satisfied, the final hyperspectral overcomplete dictionary is obtained; otherwise, the current sparse coefficient matrix is ​​replaced with the updated sparse coefficient matrix and the solution is repeated until the final hyperspectral overcomplete dictionary D is obtained.

[0092] In some embodiments, step S3 specifically includes:

[0093] S31: Based on the first wavelength point set, obtain the spectral transmittance of each broadband filter included in the multispectral measurement device, and construct a spectral transmittance matrix using the spectral transmittance of each channel.

[0094] For example, if the first wavelength point set is {a, b, c}, and the spectral transmittance measuring device includes 5 broadband filters, then the spectral transmittance of the first broadband filter at wavelengths a, b, and c is measured using the spectral transmittance measuring device. The above process is repeated to obtain the spectral transmittance of each broadband filter at wavelengths a, b, and c, and finally a 5×3 spectral transmittance matrix is ​​obtained.

[0095] S32: Couple the spectral transmittance matrix with the hyperspectral overcomplete dictionary (using equation (6)) to obtain the multispectral measurement matrix corresponding to the target to be measured;

[0096] Based on the radiance of the target at each wavelength point in each spectral channel and the spectral transmittance of the target at each wavelength point in each spectral channel, the radiance measurement value at each wavelength point is calculated and discretized based on equation (2) to obtain the set of radiance measurement values ​​of the target.

[0097] It should be noted that, firstly, an overcomplete dictionary for mid-wave infrared hyperspectral imaging can be obtained after ODL learning. Then, the multispectral measurement matrix can be obtained based on the spectral response function of the broadband filter. Finally, after obtaining the multispectral radiance measurement signal, the hyperspectral radiance signal can be reconstructed according to formulas (4) and (6). However, solving the sparse coefficient matrix in formula (6) is essentially a... The norm minimization problem. Therefore, the approximation value of the sparse coefficient matrix. It can be expressed as:

[0098] (8)

[0099] So, the reconstructed value of hyperspectral radiance It can be expressed as:

[0100] (9)

[0101] Since solving formula (7) is an NP-hard problem, this invention realizes mid-wave infrared hyperspectral radiance reconstruction based on Batch-OMP.

[0102] In some embodiments, step S4 specifically includes the following steps:

[0103] S41: Input the set of radiance measurements of the target and the multispectral measurement matrix of the multispectral measurement device into the Bach-OMP algorithm for calculation, and output the approximation value of the sparse coefficient matrix. ;

[0104] S42: Based on the hyperspectral radiance reconstruction model, the reconstructed hyperspectral radiance value is obtained by solving the following formula:

[0105] (9);

[0106] in, For the reconstructed hyperspectral brightness values, D is the hyperspectral overcomplete dictionary. This is the approximation value for the sparse coefficient matrix.

[0107] The Bach-OMP algorithm adopts the Bach-OMP algorithm from the 2008 paper "Efficient Implementation of the K-SVD Algorithm using Batch-OMP" by Rubinstein, R et al. The following is a brief introduction to the Bach-OMP algorithm process:

[0108] 1) Calculate the initial global autocorrelation matrix G0 and the initial correlation based on the set of radiance measurements of the target and the multispectral measurement matrix corresponding to the target. ;

[0109] 2) Initialize the sparse coefficient matrix approximation value Empty, Cholesky factor Empty, initialize index set Empty, and initialize the weighted norm. sparse constraints and target error ;

[0110] 3) Calculate the correlation between the current residual and each atom in the hyperspectral overcomplete dictionary D, and select the atom with the highest correlation. , where n is the number of iterations, and the initial value of n is 1. To determine the sparse constraints in the hyperspectral overcomplete dictionary D during the (n-1)th iteration, the following conditions must be met: Atoms;

[0111] 4) Based on the selection results in step 3), add the index of the atom with the highest relevance to the index set and update the index set. And calculate the submatrix corresponding to the atom with the highest correlation. And the Cholesky factor decomposition is updated to , The index set in the global autocorrelation matrix G0 is The submatrix corresponding to the atom with the highest correlation;

[0112] 5) Solve the linear system based on the index set updated in step 4) and the Cholesky factor. The approximation value of the sparse coefficient matrix is ​​obtained. , For the index set The initial correlation after transposition;

[0113] 6) Approximation value based on the sparse coefficient matrix obtained in step 5). Calculate the weighted norm and update the target error. ,in, , The intermediate variable for calculating the correlation is n, where n is the nth iteration.

[0114] 7) Determine if the target error and sparse constraints meet the stopping condition. If not, then n = n + 1, and repeat steps 3)-6); otherwise, stop the iteration and output the final approximation value of the sparse coefficient matrix. .

[0115] In step 3), the current residual A is A = y T ×y, where y is the set of measured radiance values ​​of the target to be measured.

[0116] In step 1), the initial global autocorrelation matrix G0 is calculated using the following formula:

[0117] (10);

[0118] Initial correlation The calculation formula is:

[0119] (11).

[0120] In radiation characteristic measurements, the temperature measurement range is typically 273K~5000K. Therefore, within this range, with a step size of 1K, the spectral radiance (emissivity 1) at different wavelengths is calculated according to formula (1). The obtained spectral radiance at different wavelengths at different temperatures is used as sample data, and ODL is used to learn from it to obtain a hyperspectral overcomplete dictionary. .

[0121] To reconstruct hyperspectral radiance using multispectral radiance, a connection needs to be established between a complete hyperspectral dictionary and a five-channel radiometric measurement system. As is known from compressed sensing theory, the design of the measurement matrix is ​​crucial to the signal reconstruction accuracy. Since a Gaussian matrix satisfies the constraint isometry with a very high probability, it has strong universality and can effectively sparse the measured signal. Therefore, to obtain better mid-wave infrared hyperspectral reconstruction accuracy, this invention designs five broadband filters in the 3.25µm~5.20µm band, with a step size of 10nm, center wavelengths of 3.75µm, 4.00µm, 4.25µm, 4.50µm, and 4.75µm, and a full width at half maximum (FWHM) of 500nm, and makes the spectral transmittance curves of each channel approximately Gaussian. The spectral transmittance curves of the five filters are shown below. Figure 2 As shown, the multispectral measurement matrix is ​​obtained from the spectral transmittance curve. .

[0122] As shown in formula (1), after determining the temperature and corresponding emissivity of the target, the radiance of the target can be obtained using Planck's formula. This invention will select spectral emissivity values ​​from different materials to verify the accuracy of the mid-wave infrared hyperspectral radiance inversion method based on compressed sensing proposed in this invention. This invention selects the spectral emissivity values ​​of VO2 at 373.15 K, titanium at 673.15 K, Nicrofer alloy at 1473.15 K, and CuO at 873.15 K, respectively. The spectral emissivity of each metal is as follows: Figure 3 As shown.

[0123] Based on the theory of compressed sensing, this invention proposes a mid-wave infrared hyperspectral reconstruction model based on computational spectroscopy. To establish the connection between the hyperspectral overcomplete dictionary and the measurement system, a solution for multispectral measurement matrix design and hyperspectral overcomplete dictionary reconstruction is proposed. For example... Figures 4-5 As shown, experimental results demonstrate that the mid-wave infrared hyperspectral radiance reconstruction method based on computational spectroscopy successfully reconstructed the hyperspectral radiance values ​​of four different materials at different temperatures from the radiance measurements of five channels, and indirectly reconstructed the spectral emissivity values ​​of the four materials. The radiance and emissivity reconstruction results were evaluated using SAM and relative error, respectively. The SAM values ​​of the reconstructed radiance and emissivity results for the four materials were all close to 0 rad, indicating a high similarity between the reconstructed results and the original signal shape characteristics. Furthermore, the maximum relative error of the reconstructed radiance and emissivity results for the four materials was better than 2%, indicating that the method proposed in this invention has high reconstruction accuracy. Table 1 shows the spectral radiance error, and Table 2 shows the emissivity error.

[0124] Table 1

[0125]

[0126] Table 2

[0127]

[0128] The reconstruction method proposed in this invention is less affected by emissivity and can reconstruct the mid-wave infrared hyperspectral radiance well even under conditions of unknown emissivity.

[0129] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0130] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for reconstructing mid-wave infrared hyperspectral radiance, characterized in that: Specifically, the steps include the following: S1: A hyperspectral radiance reconstruction model is constructed based on Planck's radiation law and compressed sensing theory; S2: Solving the hyperspectral overcomplete dictionary based on hyperspectral radiance samples and ODL; S3: Construct a spectral transmittance matrix based on the spectral transmittance of each channel broadband filter contained in the multispectral measurement device, and obtain the multispectral measurement matrix of the multispectral measurement device based on the spectral transmittance matrix and the hyperspectral overcomplete dictionary. S4: Solve the sparse coefficient matrix approximation value based on the multispectral measurement matrix of the multispectral measurement equipment and the set of radiance measurement values ​​of the target to be measured, and input the hyperspectral overcomplete dictionary and the sparse coefficient matrix approximation value into the hyperspectral radiance reconstruction model to reconstruct the hyperspectral radiance value.

2. The method for reconstructing mid-wave infrared hyperspectral radiance according to claim 1, characterized in that: Step S1 specifically includes: S11: Radiance based on Planck's radiation law Represented as: （1）； in, The first radiation constant, The second radiation constant, wavelength The surface emissivity of the object at that location, T is the temperature of the target being measured; S12: The measured radiance of the target in the t-th spectral channel at temperature T. Represented as: （2）； in, For the wavelength in spectral channel t At this point, the spectral transmittance of the multispectral measurement device is given, and Δλ is the difference between the upper and lower limits of integration. S13: Let the hyperspectral radiance emitted by objects in nature be... , The measured radiance values ​​for each wideband are Discretizing equation (2) gives: （3）； in, This is the spectral transmittance matrix of a multispectral measurement device. For the first The broadband band corresponds to the first Spectral transmittance at each wavelength; S14: Based on compressed sensing theory, utilizing a hyperspectral overcomplete dictionary With sparse coefficient matrix Constructing a hyperspectral radiance reconstruction model: （4）； （5）； in, For high spectral radiance, For the first The corresponding wavelength is the th The spectral radiance characteristic value of each atom; make Combining equations (3) and (4), we construct the expression for the set of radiance measurements and the multispectral measurement matrix: （6）； Where y is the set of radiance measurements. This represents a multispectral measurement matrix.

3. The method for reconstructing mid-wave infrared hyperspectral radiance according to claim 1, characterized in that: Step S2 specifically includes: S21: Select wavelength points in the mid-wave infrared band of 3.25um~5.20um according to the set step size to obtain the first wavelength point set; S22: Set the range of temperature parameters and set the emissivity of the object surface to 1. Based on Planck's radiation law, calculate the spectral radiance of each wavelength point in the first wavelength point set with a temperature parameter of 1k step size; and construct a hyperspectral radiance sample based on the spectral radiance of different wavelength points at different temperatures. S23: Input the hyperspectral radiance samples into the ODL algorithm for dictionary learning to obtain the hyperspectral overcomplete dictionary D.

4. The method for reconstructing mid-wave infrared hyperspectral radiance according to claim 2, characterized in that: Step S3 specifically includes: S31: Based on the first wavelength point set, obtain the spectral transmittance of each broadband filter included in the multispectral measurement device, and construct a spectral transmittance matrix using the spectral transmittance of each channel. S32: Couple the spectral transmittance matrix with the hyperspectral overcomplete dictionary to obtain the multispectral measurement matrix of the multispectral measurement device; Based on the radiance of the target at each wavelength point in each spectral channel and the spectral transmittance of the target at each wavelength point in each spectral channel, the radiance measurement value at each wavelength point is calculated and discretized based on equation (2) to obtain the set of radiance measurement values ​​of the target.

5. The method for reconstructing mid-wave infrared hyperspectral radiance according to claim 2, characterized in that: Step S4 specifically includes the following steps: S41: Input the set of radiance measurements of the target and the multispectral measurement matrix of the multispectral measurement device into the Bach-OMP algorithm for calculation, and output the approximation value of the sparse coefficient matrix. ; S42: Based on the hyperspectral radiance reconstruction model, the reconstructed hyperspectral radiance value is obtained by solving the following formula: （7）； in, For the reconstructed hyperspectral brightness values, D is the hyperspectral overcomplete dictionary. This is the approximation value for the sparse coefficient matrix.

Images