Hyperspectral mixed pixel decomposition method and device based on endmember spectral variability

Abstract

The present application relates to hyperspectral remote sensing data processing technical field, especially related to a kind of based on end member spectral variability Hyperspectral Mixed Pixel Decomposition Method and device. Including: based on traditional unmixing model, construct based on end member spectral variability Hyperspectral Mixed Pixel Decomposition Error Model;Based on end member spectral statistics distribution used to characterize end member spectral variability and hyperspectral mixed pixel decomposition error model, determine the statistical distribution of abundance retrieval error;Based on the statistical distribution of abundance retrieval error, derive the statistical distribution of true abundance;End member spectrum extraction is carried out to the hyperspectral image to be unmixing, to determine the confidence interval of the ground object abundance of hyperspectral image based on the statistical distribution of true abundance.

Description

Technical Field

[0001] The present invention relates to the field of hyperspectral remote sensing data processing technology, and in particular to a method and apparatus for hyperspectral mixed pixel decomposition based on endmember spectral variability. Background Technology

[0002] With the development of remote sensing science and technology, hyperspectral remote sensing has become an important technology for Earth observation. While imaging the Earth's surface, it can acquire high-spectral resolution characteristic curves reflecting the distribution attributes of ground features pixel by pixel, and is widely used in ground feature detection and surface parameter inversion. It has been successfully applied in geological mapping, precision agriculture, and ecological environment monitoring. However, due to the complexity of ground feature distribution and the limitations of the spatial resolution of hyperspectral imagers, mixed pixels are prevalent in hyperspectral remote sensing data.

[0003] Hyperspectral mixed pixel decomposition or spectral demixing aims to decompose mixed pixels into endmember spectra representing different material types and their corresponding abundance fractions. However, in most application scenarios, the assumption of fixed endmember spectra cannot be met; theoretically, fixed spectra do not exist. On the one hand, differences in atmospheric conditions, topography and lighting conditions, surrounding environment, and other factors can lead to differences in spectral amplitude or shape in endmember spectra across time and space. On the other hand, the classification of land cover types is usually macroscopic and problem-oriented, which is related to the level of detail in land cover classification. This can result in multiple subclasses within the same type of land cover, with inherent differences in physicochemical properties, which can also lead to certain amplitude variations in endmember spectra. Both of these factors ultimately lead to the phenomenon of "different spectra for the same material" and "same spectra for different materials" in endmember spectra, known as endmember variability.

[0004] Traditional hyperspectral mixed pixel decomposition methods typically use the typical pure pixel spectra of various land features or the average spectrum of multiple pure pixel spectra as fixed endmember spectra. These endmember spectra are then used to decompose mixed pixels in the scene one by one, inverting the abundance fraction of various land features in each pixel. It is evident that traditional hyperspectral mixed pixel decomposition methods do not consider the variability of endmember spectra, resulting in poor accuracy.

[0005] Therefore, there is an urgent need for a hyperspectral mixed pixel decomposition method based on endmember spectral variability. Summary of the Invention

[0006] To address the issue of poor accuracy in traditional hyperspectral mixed pixel decomposition methods, this invention provides a hyperspectral mixed pixel decomposition method and apparatus based on endmember spectral variability.

[0007] In a first aspect, embodiments of the present invention provide a hyperspectral mixed pixel decomposition method based on endmember spectral variability, comprising:

[0008] Based on the traditional unmixing model, a hyperspectral mixed pixel decomposition error model based on endmember spectral variability is constructed.

[0009] Based on the statistical distribution of endmember spectra used to characterize the variability of endmember spectra and the hyperspectral mixed pixel decomposition error model, the statistical distribution of abundance inversion error is determined.

[0010] Based on the statistical distribution of the abundance inversion error, the statistical distribution of the true abundance is derived.

[0011] Endmember spectral extraction is performed on the hyperspectral image to be unmixed to determine the confidence interval of the land cover abundance in the hyperspectral image based on the statistical distribution of the true abundance.

[0012] Secondly, embodiments of the present invention also provide a hyperspectral mixed pixel decomposition device based on endmember spectral variability, comprising:

[0013] The building unit is used to construct a hyperspectral mixed pixel decomposition error model based on endmember spectral variability, based on the traditional unmixing model.

[0014] The determination unit is used to determine the statistical distribution of abundance inversion error based on the statistical distribution of endmember spectra used to characterize endmember spectral variability and the hyperspectral mixed pixel decomposition error model.

[0015] The derivation unit is used to derive the statistical distribution of the true abundance based on the statistical distribution of the abundance inversion error.

[0016] The decomposition unit is used to perform endmember spectral extraction on the hyperspectral image to be unmixed, so as to determine the confidence interval of the land cover abundance of the hyperspectral image based on the statistical distribution of the true abundance.

[0017] This invention provides a method and apparatus for hyperspectral mixed pixel decomposition based on endmember spectral variability. By analyzing the impact of endmember spectral variability on mixed pixel decomposition, a hyperspectral mixed pixel decomposition error model based on endmember spectral variability is constructed. An endmember spectral statistical distribution is introduced to characterize the variability of endmember spectra, and the statistical distribution of abundance inversion error is derived from the hyperspectral mixed pixel decomposition error model, thus obtaining the statistical distribution of true abundance. This method can then be directly used for mixed pixel decomposition of real hyperspectral images to obtain confidence intervals for ground cover abundance. Therefore, this approach can improve the accuracy of hyperspectral mixed pixel decomposition. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a hyperspectral mixed pixel decomposition method based on endmember spectral variability provided in an embodiment of the present invention;

[0020] Figure 2 This is a hardware architecture diagram of a computing device provided in an embodiment of the present invention;

[0021] Figure 3 This is a structural diagram of a hyperspectral mixed pixel decomposition device based on endmember spectral variability provided in an embodiment of the present invention. Detailed Implementation

[0022] 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 some embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0023] Please refer to Figure 1 This invention provides a hyperspectral mixed pixel decomposition method based on endmember spectral variability, the method comprising:

[0024] Step 100: Based on the traditional unmixing model, construct a hyperspectral mixed pixel decomposition error model based on endmember spectral variability;

[0025] Step 102: Based on the statistical distribution of endmember spectra used to characterize the variability of endmember spectra and the hyperspectral mixed pixel decomposition error model, determine the statistical distribution of abundance inversion error;

[0026] Step 104: Based on the statistical distribution of abundance inversion error, derive the statistical distribution of true abundance;

[0027] Step 106: Perform endmember spectral extraction on the hyperspectral image to be unmixed, and determine the confidence interval of land cover abundance in the hyperspectral image based on the statistical distribution of the true abundance.

[0028] In this embodiment of the invention, by analyzing the impact of endmember spectral variability on hybrid pixel decomposition, a hyperspectral hybrid pixel decomposition error model based on endmember spectral variability is constructed. An endmember spectral statistical distribution is introduced to characterize the variability of endmember spectra, and combined with the hyperspectral hybrid pixel decomposition error model, the statistical distribution of abundance inversion error is derived, thus obtaining the statistical distribution of true abundance. Therefore, this model can be directly used for hybrid pixel decomposition of real hyperspectral images to obtain confidence intervals for ground cover abundance. Thus, this scheme can improve the accuracy of hyperspectral hybrid pixel decomposition.

[0029] For step 100:

[0030] In some implementations, step 100, "constructing a hyperspectral mixed pixel decomposition error model based on endmember spectral variability based on a traditional unmixing model," may include steps S1-S2:

[0031] Step S1: Based on the traditional unmixing model, construct a hyperspectral mixed pixel decomposition model that considers the spectral variability of endmembers;

[0032] Step S2: Subtract the hyperspectral mixed pixel decomposition model from the traditional unmixing model, and use the least squares method to fit and solve the problem to obtain a hyperspectral mixed pixel decomposition error model that represents the quantitative relationship between abundance inversion error and endmember spectral estimation error.

[0033] In this embodiment, starting from the traditional unmixing model with fixed endmember spectra as input, considering the difference between the true endmember spectra within the mixed pixels and the fixed endmember spectra used during unmixing, a hyperspectral mixed pixel decomposition model with endmember spectral estimation errors caused by abundance inversion error and spectral variability is derived. Then, the difference between the hyperspectral mixed pixel decomposition model and the traditional unmixing model is calculated, and the least squares method is used to fit and solve the problem, resulting in a hyperspectral mixed pixel decomposition error model that characterizes the quantitative relationship between abundance inversion error and endmember spectral estimation error, i.e., the influence of spectral variability on abundance inversion.

[0034] In step S1, the hyperspectral mixed pixel decomposition model is as follows:

[0035]

[0036] In the formula, r∈R L×1 It is a hyperspectral mixed pixel spectrum. The endmember spectral matrix used for unmixing in traditional unmixing models. Let e∈R be the abundance scores of various land cover types unmixed using a traditional unmixing model. L×1 For the error term, R is a real number between 0 and 1, L is the number of bands, p is the number of land cover species, M is the true spectrum of each land cover species, and f is the true abundance of each land cover species. For endmember spectral estimation error, This represents the abundance inversion error.

[0037] In this embodiment, since the dense mixing between ground features is linear at the macroscopic scale, the traditional unmixing model is expressed as:

[0038]

[0039] Where, r∈R L×1 It is a hyperspectral mixed pixel spectrum. The endmember spectral matrix used for unmixing in traditional unmixing models. Let e∈R be the abundance scores of various land cover types unmixed using a traditional unmixing model. L×1 For the error term, R is a real number between 0 and 1, L is the number of bands, and p is the number of land cover species.

[0040] Since abundance scores represent the area proportion of ground features within a pixel and have practical physical meaning, they typically need to satisfy non-negativity constraints and a sum-to-one constraint, i.e.:

[0041]

[0042] Traditional mixed pixel decomposition algorithms typically use the typical pure pixel spectra of various land features or the average spectrum of multiple pure pixel spectra as fixed endmember spectra, and decompose mixed pixels in the scene one by one to obtain the abundance fraction of various land features in each pixel. To improve the accuracy of hyperspectral mixed pixel decomposition, a relationship can be established between the real model and the traditional unmixing model, i.e., a hyperspectral mixed pixel decomposition model.

[0043] In step S2, the hyperspectral mixed pixel decomposition error model is:

[0044]

[0045] In the formula, Δf∈R p×1 For abundance inversion error, G SCLS The coefficient matrix obtained by the least squares method, ΔM∈R L×p For endmember spectral estimation error, Let I represent the abundance scores of various land features obtained using a traditional unmixing model, and let I be the identity matrix. The endmember spectral matrix used for unmixing in traditional unmixing models. for The transpose of C, C = [1,1,…,1] T ∈R p×1 R is a vector consisting entirely of 1s, R is a real number between 0 and 1, L is the number of bands, and p is the number of land cover species.

[0046] In this embodiment, to establish the difference between the hyperspectral mixed pixel decomposition error model caused by endmember variability and the traditional unmixing model, while considering that the abundance inversion error Δf needs to meet the following requirements... Due to the physical constraints, and because the number of bands L in hyperspectral data is generally much larger than the number of endmember categories p within a pixel, leading to overdetermined equations (i.e., the number of equations is much larger than the number of unknowns), a least squares fitting method is used to solve the problem. Considering that the actual land cover abundance f is unknown in practical applications, the land cover abundance obtained from the actual solution is used. By replacing the second-order small quantity ΔM×Δf with the abundance inversion error Δf and ignoring the second-order small quantity ΔM×Δf, we obtain the quantitative relationship between the abundance inversion error Δf and the endmember spectral estimation error ΔM, which is the hyperspectral mixed pixel decomposition error model.

[0047] Regarding step 102:

[0048] In some implementations, the statistical distribution of abundance inversion error is as follows:

[0049]

[0050] In the formula, Δf∈R p×1 The abundance inversion error is represented by N(), where N() represents the normal distribution, 0 is the zero vector, and ∑ Δf G is the covariance matrix of abundance inversion error, where A and B are intermediate variables with no real meaning. SCLS The coefficient matrix obtained by the least squares method. Abundance scores of various land features unmixed using traditional unmixing models The elements, ∑ ii Let ∑ be the autocovariance matrix of the i-th type endmember spectrum. ij Let p be the cross-covariance matrix between the endmember spectra of class i and class j, and p be the number of land cover species, i.e., the number of endmember spectral species.

[0051] In this embodiment, it is assumed that each endmember spectrum is a Gaussian multivariate random variable, following a statistical distribution m. i ~N(μ) i ,∑ i Furthermore, because the endmember spectral matrix is ​​typically used during unmixing... The average spectrum of multiple endmember spectra is obtained, i.e. The probability distribution of the endmember spectral estimation error is Δm. i ~N(0,∑ i Substituting the probability distribution of the endmember spectral estimation error into the hyperspectral mixed pixel decomposition error model constructed in step 100, we obtain the statistical distribution of the abundance inversion error Δf as shown in the above formula. It can be understood that the abundance inversion error also follows a Gaussian distribution.

[0052] Regarding step 104:

[0053] In some implementations, the statistical distribution of true abundance is as follows:

[0054]

[0055] In the formula, f represents the true abundance of various land features, and N() represents the normal distribution. Σ represents the abundance scores of various land features unmixed using a traditional unmixing model. Δf G is the covariance matrix of abundance inversion error, where A and B are intermediate variables with no real meaning. SCLS The coefficient matrix obtained by the least squares method. Abundance scores of various land features unmixed using traditional unmixing models The elements, ∑ ii Let ∑ be the autocovariance matrix of the i-th type endmember spectrum. ij Let p be the cross-covariance matrix between the endmember spectra of class i and class j, where p is the number of land cover species, i.e., the number of endmember spectral species; and I is the identity matrix. The endmember spectral matrix used for unmixing in traditional unmixing models. for The transpose of C, C = [1,1,…,1] T ∈R p×1 It is a vector consisting entirely of 1s.

[0056] In this embodiment, based on the statistical distribution of the abundance inversion error in step 102, the statistical distribution of the true abundance as shown in the above formula can be derived.

[0057] Regarding step 106:

[0058] Through steps 100-104, a hyperspectral mixed pixel decomposition method considering endmember spectral variability can be established. Then, decomposition can be performed on each hyperspectral image to be demixed.

[0059] Specifically, step 106, "extracting endmember spectra from the hyperspectral image to be unmixed, and determining the confidence interval of land cover abundance in the hyperspectral image based on the statistical distribution of true abundance," may include steps B1-B6:

[0060] Step B1: Using an automatic endmember extraction algorithm, endmember spectra are extracted from the hyperspectral image to be unmixed to obtain the endmember spectrum set of the hyperspectral image; wherein, each type of endmember spectrum in the endmember spectrum set corresponds to a type of land cover.

[0061] In this step, the entire hyperspectral image is first segmented into multiple sub-images. Then, the HySime endmember number determination method is used to determine the number of endmember spectra in each sub-image. Subsequently, an automatic endmember extraction method (such as vertex component analysis, maximum volume method, etc.) is used to extract the corresponding endmember spectra from the sub-images. Finally, the extracted endmember spectra are subjected to k-means clustering and spectral angle recognition to determine the land cover type corresponding to the endmember spectra of each cluster, thereby obtaining the endmember spectrum set that can represent each type of land cover in the hyperspectral image.

[0062] Step B2: Based on the endmember spectral set, estimate the overall mean, autocovariance matrix, and crosscovariance matrix of each type of endmember spectrum in the hyperspectral image to obtain the corresponding estimation results.

[0063] In this embodiment, the following formulas are used to estimate the overall mean, autocovariance matrix, and crosscovariance matrix of each type of endmember spectrum in the hyperspectral image:

[0064]

[0065]

[0066]

[0067] in, This represents the estimated population mean of the i-th type endmember spectrum; ∑ ii Σ represents the estimated autocovariance matrix of the i-th type endmember spectrum; ij This represents the estimation result of the cross-covariance matrix between the endmember spectra of class i and class j; r ij r represents the j-th spectral sample of the i-th type of endmember spectrum; ik r jk represents the k-th spectral sample of the i-th and j-th endmember spectra, respectively; N represents the number of samples of the i-th and j-th endmember spectra.

[0068] Step B3: Based on the estimation results of the overall mean of each type of endmember spectrum, determine the endmember spectral matrix used for unmixing in the traditional unmixing model.

[0069] Based on the estimated overall mean of each type of endmember spectrum obtained in step B2, the overall means of each type of endmember spectrum are used to construct the endmember spectrum matrix for unmixing in traditional unmixing models, i.e.

[0070] Step B4: Using the traditional unmixing model and the endmember spectral matrix used in the traditional unmixing model, solve for the abundance fractions of various land features unmixed using the traditional unmixing model.

[0071] The endmember spectral matrix used for unmixing the traditional unmixing model obtained in step B3 is... Substituting into the traditional unmixing model, where r represents the pixel value of the hyperspectral image and the error term e can be calculated, the abundance fractions of various land cover types unmixed using the traditional unmixing model can be obtained.

[0072] Step B5 involves substituting the estimated autocovariance and crosscovariance matrices of each type of endmember spectrum, the endmember spectral matrix used in the traditional unmixing model, and the abundance fractions of various land cover types obtained by unmixing using the traditional unmixing model into the statistical distribution of the true abundance to obtain the statistical distribution of the true abundance of the hyperspectral image.

[0073] The estimation results of the autocovariance matrix and cross-covariance matrix of each type of endmember spectrum (Σ) ii and Σ ij Traditional unmixing models use endmember spectral matrices for unmixing. Abundance scores of various land features obtained by unmixing using traditional unmixing models Substituting the statistical distribution of the true abundance from step 104, we can obtain the statistical distribution of the true abundance of the hyperspectral image.

[0074] Step B6: Based on the statistical distribution of the true abundance of hyperspectral images, obtain the confidence interval of the land cover abundance of hyperspectral images.

[0075] Since the statistical distribution of the true abundance of hyperspectral images follows a Gaussian distribution, the confidence interval for the abundance of ground features in hyperspectral images can be calculated using the following formula:

[0076]

[0077] Among them, f1-f p Elements representing the true abundance of various land features. Abundance scores of various land features unmixed using traditional unmixing models The elements, Δf1-Δf p For the elements of abundance inversion error Δ, The covariance matrix Σ of the abundance inversion error in the statistical distribution of the true abundance in step 104 is... Δ The square root of the diagonal.

[0078] like Figure 2 , Figure 3 As shown, this embodiment of the invention provides a hyperspectral hybrid pixel decomposition device based on endmember spectral variability. The device embodiment can be implemented through software, hardware, or a combination of both. From a hardware perspective, as... Figure 2The diagram shown is a hardware architecture diagram of a computing device containing a hyperspectral mixed pixel decomposition device based on endmember spectral variability provided in an embodiment of the present invention. (Except for...) Figure 2 In addition to the processor, memory, network interface, and non-volatile memory shown, the computing device in the embodiment may also include other hardware, such as a forwarding chip responsible for processing packets. Taking software implementation as an example, such as... Figure 3 As shown, a device in a logical sense is formed by the CPU of its computing device reading the corresponding computer program from non-volatile memory into memory and running it. This embodiment provides a hyperspectral mixed pixel decomposition device based on endmember spectral variability, comprising:

[0079] Building unit 301 is used to construct a hyperspectral mixed pixel decomposition error model based on endmember spectral variability, based on the traditional unmixing model.

[0080] The determination unit 302 is used to determine the statistical distribution of abundance inversion error based on the statistical distribution of endmember spectra used to characterize endmember spectral variability and the hyperspectral mixed pixel decomposition error model.

[0081] Derivation unit 303 is used to derive the statistical distribution of the true abundance based on the statistical distribution of the abundance inversion error;

[0082] The decomposition unit 304 is used to extract endmember spectra from the hyperspectral image to be unmixed, so as to determine the confidence interval of the land cover abundance in the hyperspectral image based on the statistical distribution of the true abundance.

[0083] It is understood that the structures illustrated in the embodiments of the present invention do not constitute a specific limitation on a hyperspectral mixed pixel decomposition device based on endmember spectral variability. In other embodiments of the present invention, a hyperspectral mixed pixel decomposition device based on endmember spectral variability may include more or fewer component units than illustrated, or combine some component units, or split some component units, or arrange different component units. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0084] The information interaction and execution process between the various units in the above-mentioned device are based on the same concept as the method embodiment of the present invention, and the specific details can be found in the description of the method embodiment of the present invention, and will not be repeated here.

[0085] This invention also provides a computing device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements a hyperspectral mixed pixel decomposition method based on endmember spectral variability according to any embodiment of this invention.

[0086] This invention also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program causes the processor to perform a hyperspectral mixed pixel decomposition method based on endmember spectral variability according to any embodiment of this invention.

[0087] Specifically, a system or apparatus equipped with a storage medium may be provided, on which software program code implementing the functions of any of the embodiments described above is stored, and the computer (or CPU or MPU) of the system or apparatus may read and execute the program code stored in the storage medium.

[0088] In this case, the program code read from the storage medium can itself implement the function of any of the above embodiments, and therefore the program code and the storage medium storing the program code constitute part of the present invention.

[0089] Examples of storage media used to provide program code include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, program code can be downloaded from a server computer via a communication network.

[0090] Furthermore, it should be clear that not only can the program code read by the computer be executed, but also the operating system or other components operating on the computer can be instructed based on the program code to perform some or all of the actual operations, thereby realizing the function of any of the embodiments described above.

[0091] Furthermore, it is understood that the program code read from the storage medium is written to the memory set in the expansion board inserted into the computer or to the memory set in the expansion module connected to the computer. Then, based on the instructions of the program code, the CPU or other components installed on the expansion board or expansion module execute some and all of the actual operations, thereby realizing the function of any of the above embodiments.

[0092] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0093] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as ROM, RAM, magnetic disk, or optical disk.

[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A hyperspectral mixed pixel decomposition method based on endmember spectral variability, characterized in that, include: Based on the traditional unmixing model, a hyperspectral mixed pixel decomposition error model based on endmember spectral variability is constructed. Based on the statistical distribution of endmember spectra used to characterize the variability of endmember spectra and the hyperspectral mixed pixel decomposition error model, the statistical distribution of abundance inversion error is determined. Based on the statistical distribution of the abundance inversion error, the statistical distribution of the true abundance is derived. Endmember spectral extraction is performed on the hyperspectral image to be unmixed to determine the confidence interval of the land cover abundance in the hyperspectral image based on the statistical distribution of the true abundance.

2. The method according to claim 1, characterized in that, The aforementioned construction of a hyperspectral mixed pixel decomposition error model based on endmember spectral variability, based on the traditional unmixing model, includes: Based on the traditional unmixing model, a hyperspectral mixed pixel decomposition model considering the endmember spectral variability is constructed. The difference between the hyperspectral mixed pixel decomposition model and the traditional unmixing model is calculated, and the least squares method is used to fit and solve the problem to obtain a hyperspectral mixed pixel decomposition error model that can be used to characterize the quantitative relationship between abundance inversion error and endmember spectral estimation error.

3. The method according to claim 2, characterized in that, The hyperspectral mixed pixel decomposition model is as follows: In the formula, r∈R L×1 It is a hyperspectral mixed pixel spectrum. The endmember spectral matrix used for unmixing in traditional unmixing models. Let e∈R be the abundance scores of various land cover types unmixed using a traditional unmixing model. L×1 For the error term, R is a real number between 0 and 1, L is the number of bands, p is the number of land cover species, M is the true spectrum of each land cover species, and f is the true abundance of each land cover species. For endmember spectral estimation error, This represents the abundance inversion error.

4. The method according to claim 2, characterized in that, The hyperspectral mixed pixel decomposition error model is as follows: In the formula, Δf∈R p×1 For abundance inversion error, G SCLS The coefficient matrix obtained by the least squares method, ΔM∈R L×p For endmember spectral estimation error, Let I represent the abundance scores of various land features obtained using a traditional unmixing model, and let I be the identity matrix. The endmember spectral matrix used for unmixing in traditional unmixing models. for The transpose of C, C = [1,1,…,1] T ∈R p×1 R is a vector consisting entirely of 1s, R is a real number between 0 and 1, L is the number of bands, and p is the number of land cover species.

5. The method according to claim 1, characterized in that, The statistical distribution of the abundance inversion error is as follows: In the formula, Δf∈R p×1 The abundance inversion error is represented by N(), where N() represents the normal distribution, 0 is the zero vector, and ∑ Δf G is the covariance matrix of abundance inversion error, where A and B are intermediate variables with no real meaning. SCLS The coefficient matrix obtained by the least squares method. Abundance scores of various land features unmixed using traditional unmixing models The elements, ∑ ii Let Σ be the autocovariance matrix of the i-th type endmember spectrum. ij Let p be the cross-covariance matrix between the endmember spectra of class i and class j, and p be the number of land cover species, i.e., the number of endmember spectral species.

6. The method according to claim 1, characterized in that, The statistical distribution of true abundance is as follows: In the formula, f represents the true abundance of various land features, and N() represents the normal distribution. Σ represents the abundance scores of various land features unmixed using a traditional unmixing model. Δf G is the covariance matrix of abundance inversion error, where A and B are intermediate variables with no real meaning. SCLS The coefficient matrix obtained by the least squares method. Abundance scores of various land features unmixed using traditional unmixing models The elements, ∑ ii Let ∑ be the autocovariance matrix of the i-th type endmember spectrum. ij Let p be the cross-covariance matrix between the endmember spectra of class i and class j, where p is the number of land cover species, i.e., the number of endmember spectral species; and I is the identity matrix. The endmember spectral matrix used for unmixing in traditional unmixing models. for The transpose of C, C = [1,1,…,1] T ∈R p×1 It is a vector consisting entirely of 1s.

7. The method according to claim 1, characterized in that, The process of performing endmember spectral extraction on the hyperspectral image to be unmixed, and determining the confidence interval of the land cover abundance in the hyperspectral image based on the statistical distribution of the true abundance, includes: An automatic endmember extraction algorithm is used to extract endmember spectra from the hyperspectral image to be demixed, resulting in an endmember spectrum set of the hyperspectral image; wherein each type of endmember spectrum in the endmember spectrum set corresponds to a type of land cover. Based on the endmember spectrum set, the overall mean, autocovariance matrix, and crosscovariance matrix of each type of endmember spectrum in the hyperspectral image are estimated to obtain the corresponding estimation results. Based on the estimation results of the overall mean of each type of endmember spectrum, the endmember spectral matrix used for unmixing in the traditional unmixing model is determined. Using the traditional unmixing model and the endmember spectral matrix used in the traditional unmixing model, the abundance fractions of various land features unmixed by the traditional unmixing model are solved. The estimated results of the autocovariance matrix and crosscovariance matrix of each type of endmember spectrum, the endmember spectrum matrix used for unmixing by the traditional unmixing model, and the abundance fractions of various land cover types obtained by unmixing by the traditional unmixing model are substituted into the statistical distribution of the true abundance to obtain the statistical distribution of the true abundance of the hyperspectral image. Based on the statistical distribution of the true abundance of the hyperspectral image, the confidence interval of the land cover abundance of the hyperspectral image is obtained.

8. A hyperspectral hybrid pixel decomposition device based on endmember spectral variability, characterized in that, include: The building unit is used to construct a hyperspectral mixed pixel decomposition error model based on endmember spectral variability, based on the traditional unmixing model. The determination unit is used to determine the statistical distribution of abundance inversion error based on the statistical distribution of endmember spectra used to characterize endmember spectral variability and the hyperspectral mixed pixel decomposition error model. The derivation unit is used to derive the statistical distribution of the true abundance based on the statistical distribution of the abundance inversion error. The decomposition unit is used to perform endmember spectral extraction on the hyperspectral image to be unmixed, so as to determine the confidence interval of the land cover abundance of the hyperspectral image based on the statistical distribution of the true abundance.

9. A computing device comprising a memory and a processor, wherein the memory stores a computer program, and the processor, when executing the computer program, implements the method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method of any one of claims 1-7.

Images