Spectral unmixing method for multi-modal remote sensing images based on digital surface model (DSM) and hyperspectrum

Abstract

The present application belongs to the field of remote sensing image processing, and relates to a multi-modal remote sensing image spectral unmixing method based on a digital surface model (DSM) and hyperspectral joint. The present application solves the problem that the existing method does not consider that the DSM and the hyperspectral data are actually different dimensional information from different sensors of the same scene, does not model the joint imaging of the DSM and the hyperspectral data in the model imaging process, and leads to poor spectral unmixing precision of the remote sensing image. The process is as follows: obtaining multi-modal remote sensing images with the same resolution and covering the same geographical area; the multi-modal remote sensing images include hyperspectral remote sensing image data and DSM remote sensing image data; obtaining geometric component information of a ground object type based on the DSM remote sensing image data; constructing a multi-modal remote sensing image spectral unmixing model based on the DSM and the hyperspectral joint; solving the multi-modal remote sensing image spectral unmixing model based on the DSM and the hyperspectral joint by using an alternating direction multiplier method to obtain abundance information.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of remote sensing image processing, and relates to a multi-modal remote sensing image spectral unmixing method. BACKGROUND

[0002] High-precision spectral unmixing of hyperspectral remote sensing images faces a challenging problem, that is, it is difficult to distinguish different artificial target categories with similar spectra at different heights without height information. A digital surface model (DSM) with the same spatial resolution as the hyperspectral image can provide additional geometric information, which is very useful for the hyperspectral image spectral unmixing task. In addition, in the hyperspectral image, the spectral variability phenomenon cannot be avoided, that is, the spectral curve of the same ground object target will change due to changes in external imaging conditions, which is called the "same object different spectrum" phenomenon. The hyperspectral unmixing model can obtain a shadow component and a reflectance component by introducing an intrinsic decomposition model, wherein the shadow component is the result of the interaction of illumination information and ground surface, and the reflectance component can represent the inherent properties of the ground object target, which is not affected by external imaging conditions. Based on this, the reflectance component can greatly weaken the influence of spectral variability, and then the linear mixing model can be decomposed into endmember information and abundance information. From the perspective of physical modeling, by jointly using the DSM data, the shadow component can be modeled by the interaction of the geometric component of the DSM and the illumination component, and by imposing prior constraint conditions on the reflectance component and the abundance component, the reflectance information can be obtained based on the physical model term and the prior knowledge constraint, and the more accurate abundance information can be estimated.

[0003] Currently, there are two main types of multi-modal joint spectral unmixing methods, one is a deep learning-based method, and the other is a physical model-based method. The deep learning-based method usually inputs the geometric information provided by the DSM as a spatial feature term into the network for feature enhancement, which is equivalent to introducing additional prior knowledge for assistance. The physical model-based method applies weights to improve the separability of different target types with similar spectra at different heights. However, the existing multi-modal remote sensing image spectral unmixing based on DSM and hyperspectral joint does not consider that the DSM data and the hyperspectral data are actually different dimensional information from different sensors observing the same scene, and does not model the joint imaging of the DSM data and the hyperspectral data in the model imaging process, so there is a slight deficiency in model completeness. SUMMARY

[0004] The application aims at solving the problem that the existing multi-modal remote sensing image spectral unmixing based on DSM and hyperspectral joint does not consider that the DSM data and the hyperspectral data are actually different dimensional information from different sensors observing the same scene, does not model the joint imaging of the DSM data and the hyperspectral data in the process of model imaging, and leads to poor spectral unmixing precision of the remote sensing image.

[0005] The multi-modal remote sensing image spectral unmixing method based on DSM and hyperspectral joint has the following specific process:

[0006] Step one, obtaining multi-modal remote sensing images with the same geographical coverage and the same resolution;

[0007] The multi-modal remote sensing images include hyperspectral remote sensing image data and DSM remote sensing image data;

[0008] Obtaining geometric component information of a ground object type based on the DSM remote sensing image data;

[0009] The DSM is a digital surface model;

[0010] Step two, constructing a multi-modal remote sensing image spectral unmixing model based on DSM and hyperspectral joint;

[0011] Step three, solving the multi-modal remote sensing image spectral unmixing model based on DSM and hyperspectral joint constructed in step two by using an alternating direction multiplier method to obtain abundance information.

[0012] The application has the following beneficial effects:

[0013] The application aims at improving the spectral unmixing precision of remote sensing images and proposes a multi-modal remote sensing image spectral unmixing method based on DSM and hyperspectral joint.

[0014] The present application solves the above problems from constructing a multi-modal imaging physical model (at present, there are two main types of multi-modal joint spectral unmixing methods, one is a deep learning-based method, and the other is a physical model-based method. The deep learning-based method usually inputs the geometric information provided by the DSM into the network as a spatial feature item for feature enhancement, which is equivalent to introducing additional prior knowledge for assistance. The physical model-based method applies weights to improve the separability of different target types that exhibit similar spectra at different altitudes. However, the existing multi-modal remote sensing image spectral unmixing based on DSM and hyperspectral joint does not consider that the DSM data and the hyperspectral data are actually different dimensional information from different sensors observing the same scene, and does not model the joint imaging of the DSM data and the hyperspectral data in the model imaging process, so there is a slight deficiency in model completeness.); Considering that the same object may have different spectra in remote sensing images, that is, the spectral curve of the same ground object target type changes due to changes in external imaging conditions, which is not conducive to distinguishing the characteristics of different categories, and also not conducive to unmixing. The design idea of the present application is to embed an eigenvalue decomposition model in the multi-modal spectral unmixing model, which can obtain a reflectivity component with greatly reduced spectral variability, which is only related to the characteristics of the material itself and is not affected by external imaging conditions, thereby obtaining more accurate unmixing results. Specifically, in the construction of the multi-modal spectral unmixing model, the DSM and hyperspectral data are combined, and a multi-modal imaging model term can be constructed using a physical model, which can utilize both the spectral information of the hyperspectral data and the spatial geometric information of the DSM data. The connection between the DSM data and the light is established in the imaging process to obtain the true light-dark component information, and the eigenvalue decomposition model is used to obtain more accurate reflectivity component information. Based on the local similarity of the reflectivity component and the global sparsity prior constraint of the abundance information, the unmixing result with greatly reduced spectral variability can be obtained. That is, the constructed multi-modal remote sensing image spectral unmixing model can theoretically significantly improve the accuracy of hyperspectral image unmixing, and the obtained hyperspectral unmixing result is more realistic and more accurate.

[0015] The method of the present application constructs a multi-modal physical imaging model, combines an eigenvalue image decomposition model, uses a spherical harmonic light model to obtain light-dark components, can obtain reflectivity information with reduced spectral variability, and through the constraint of prior knowledge, optimizes the hyperspectral unmixing result with higher reality and stronger accuracy, and finally realizes high-precision multi-modal remote sensing image spectral unmixing based on the joint of DSM and hyperspectral.

[0016] In order to verify the performance of the present application, a group of real hyperspectral-DSM image pairs are verified, and the experimental results show that, compared with the current representative method, the accuracy of the hyperspectral image unmixing result obtained by the present application is higher, and the reconstruction error is minimum. The experimental results verify the effectiveness of the multi-modal remote sensing image spectral unmixing method based on DSM and hyperspectral joint proposed in the present application. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The implementation flowchart of the present application is shown in the figure.

[0018] Figure 2a The false color synthesis image of the hyperspectral image is shown in the figure.

[0019] Figure 2b The DSM image is shown in the figure.

[0020] Figure 3 The abundance map of different ground object types of different methods is shown in the figure. DETAILED DESCRIPTION

[0021] Specific implementation one: the specific process of the multi-modal remote sensing image spectral unmixing method based on DSM and hyperspectral joint of the present embodiment is as follows:

[0022] Step one, obtaining multi-modal remote sensing images covering the same geographical area and having the same resolution;

[0023] The multi-modal remote sensing images include hyperspectral remote sensing image data and DSM remote sensing image data;

[0024] Obtaining geometric component information of ground object types based on the DSM remote sensing image data;

[0025] The DSM is a digital surface model;

[0026] Step two, constructing a multi-modal remote sensing image spectral unmixing model based on DSM and hyperspectral joint;

[0027] Step three, solving the multi-modal remote sensing image spectral unmixing model based on DSM and hyperspectral joint constructed in step two by using the alternating direction multiplier method, obtaining more accurate abundance information, and reconstructing a hyperspectral image with the same original resolution based on the abundance information.

[0028] Specific implementation two: the difference between the present embodiment and specific implementation one is that, in step one, multi-modal remote sensing images covering the same geographical area and having the same resolution are obtained;

[0029] The multi-modal remote sensing images include hyperspectral remote sensing image data and DSM remote sensing image data;

[0030] Obtaining geometric component information of a ground object type based on DSM remote sensing image data

[0031] The DSM is a digital surface model;

[0032] The specific process is:

[0033] Step one,

[0034] Let And Respectively represent the obtained hyperspectral remote sensing image data and DSM remote sensing image data covering the same geographic area and having the same resolution;

[0035] Wherein, Indicates the spectral characteristics of the kth pixel in the hyperspectral remote sensing image data, k = 1, 2,..., P, k represents the kth pixel of the hyperspectral image, and P represents the total number of pixels of the hyperspectral image;

[0036] λ1 represents the first wavelength, λ2 represents the second wavelength, and λ3 represents the third wavelength, Indicates the wavelength of the first waveband of the kth pixel, Indicates the wavelength of the second waveband of the kth pixel, Indicates the wavelength of the dth waveband of the kth pixel, and the superscript T represents transposition;

[0037] D k Indicates the elevation corresponding to the kth pixel in the DSM remote sensing image data;

[0038] d H Indicates the spectral dimension of the sample in the hyperspectral remote sensing image, d 1 Indicates the spectral dimension of the sample in the DSM remote sensing image, Indicates the domain;

[0039] Step two, calculate the normal of each pixel to obtain the normal vector feature as follows:

[0040]

[0041] Wherein, N represents the normal vector feature, n k Indicates the projection of the normal of the kth pixel on the x, y, and z space coordinate (space rectangular coordinate system) axes, n k = [n k,x ,n k,y ,n k,z ] T ,k = 1, 2,..., P; n k,x Indicates the projection of the normal of the kth pixel on the x space coordinate axis, n k,yn k represents the projection of the normal of the kth pixel on the y spatial coordinate axis, k,z n k represents the projection of the normal of the kth pixel on the z spatial coordinate axis;

[0042] Step III, calculating the geometric components G based on step II:

[0043] G = [G1, G2,..., G k ..., G P ] T

[0044] wherein G k k represents the geometric component of the kth pixel, k = 1, 2,..., P;

[0045] G k =

[0046] [c4, 2c2n k,y , 2c2n k,x , c2n k,z , 2c1n k,x n k,y , 2c1n k,y n k,z , c3n k,z n k,z - c5, 2c1n k,x n k,z , c1(n k, x n k,x - n k,y n k,y ] T wherein c1, c2, c3, c4, and c5 are constants.

[0047] The other steps and parameters are the same as in embodiment I.

[0048] Embodiment III: The difference between this embodiment and embodiment I or II is that the values of the constants c1, c2, c3, c4, and c5 are respectively c1 = 0.429, c2 = 0.512, c3 = 0.743, c4 = 0.886, and c5 = 248.

[0049] The other steps and parameters are the same as in embodiment I or II.

[0050] Embodiment IV: The difference between this embodiment and any one of embodiments I to III is that in step II, a multi-modal remote sensing image spectral unmixing model based on DSM and hyperspectral joint is constructed;

[0051] The specific process is as follows:

[0052] Step two one, introducing intrinsic image decomposition model, intrinsic image decomposition model decomposes hyperspectral remote sensing image Y H into the point product of reflectance component S and shading component R:

[0053] where Y H represents hyperspectral remote sensing image;

[0054] represents shading component, which can represent the change of external imaging conditions;

[0055] represents reflectance component, which can represent the reflectance characteristics of the material itself; R can more accurately represent the reflectance spectral curve of the ground object, reducing the spectral uncertainty under the change of external conditions;

[0056] In the traditional spectral unmixing model, the original hyperspectral image can be decomposed into the joint action of endmember information and abundance information, wherein the endmember represents the type of ground object in the data, and the abundance represents the set of proportions of different ground objects in the hyperspectral image data. Therefore, the purpose of spectral unmixing is to realize the process of extracting and calculating the reflectance of ground object spectral information, that is, endmember extraction and abundance estimation.

[0057] Step two two, introducing linear spectral unmixing model;

[0058] Step two three, based on intrinsic image decomposition model and linear spectral unmixing model, constructing the objective function of multi-modal remote sensing image spectral unmixing model based on DSM and hyperspectral joint.

[0059] Other steps and parameters are the same as one of the first to third embodiments.

[0060] Embodiment five: different from one of the first to fourth embodiments, the linear spectral unmixing model is introduced in step two two; the expression is as follows:

[0061] Y H = EA + N

[0062] wherein, represents endmember matrix, represents abundance matrix, represents measurement error; C represents the number of endmembers.

[0063] The abundance matrix has two basic physical constraints, which are abundance non-negativity constraint and abundance sum-to-one constraint, which means that in natural scenes, the proportion of ground object types is non-negative, and in hyperspectral image data, the sum of the proportions of all ground object types is one;

[0064] The other steps and parameters are the same as one of the first to fourth embodiments.

[0065] The sixth embodiment is different from one of the first to fifth embodiments in that the objective function of the spectral unmixing model based on the DSM and hyperspectral joint multi-modal remote sensing image is constructed based on the intrinsic image decomposition model and the linear spectral unmixing model in the step two three; the objective function is represented as follows:

[0066]

[0067] s.t.A≥0

[0068] wherein, represents the square of the Frobenius norm; α and β are penalty coefficients for balancing the importance of each constraint term; M represents ambient light; G represents a geometric component; 1 P represents a P-dimensional all-one matrix, T represents transposition; Ψ(R) represents the prior constraint term of the reflectance component, Υ(A) represents the prior constraint term of the abundance information, U represents the objective function; ⊙ represents Hadamard product;

[0069] The term multiplied by the coefficient α and β is the constraint term, and the importance of each term is embodied by setting the penalty coefficient of the constraint term.

[0070] The first term in the objective function is the intrinsic image decomposition model term, which is used to decompose the original hyperspectral remote sensing image into a brightness component representing the change of external imaging conditions and a reflectance component representing the reflectance characteristics of the material itself, and to ensure that the error of the point multiplication of the original hyperspectral remote sensing image and the reflectance component S and the brightness component R is as small as possible;

[0071] The second term is the brightness component term, which is used to decompose the original hyperspectral remote sensing image into the interaction of the geometric component G obtained from the digital surface model DSM remote sensing image data and the ambient light M representing the external imaging conditions, and the spherical harmonic light coefficient is introduced for solving;

[0072] The third term is the spectral unmixing imaging model term of the reflectance component replacement, and since the reflectance component is only related to the characteristics of the ground object type itself, the abundance information solved is more real;

[0073] The fourth term Ψ(R) is the prior constraint term of the reflectance component;

[0074] The fifth term Υ(A) is the prior constraint term of the abundance information.

[0075] The other steps and parameters are the same as one of the first to fifth embodiments.

[0076] Specific implementation seven: different from one of the specific implementations one to six, the prior constraint term Ψ (R) of the reflectance component in the objective function is obtained by the following process:

[0077] Constraint one in the prior constraint term Ψ (R) of the reflectance component: adjacent pixels have a greater probability of having similar reflectance, and therefore, the spectrum of one pixel can be linearly combined with the spectra of adjacent pixels;

[0078] R i = w ij R j

[0079] wherein R i and R j represent the reflectance information of pixels i and j in R respectively, w ij represents the similarity measure between pixels i and j in R;

[0080] i = 1, 2,..., P, i represents the i-th pixel of the hyperspectral image, and P represents the total number of pixels of the hyperspectral image;

[0081] j = 1, 2,..., P, j represents the j-th pixel of the hyperspectral image, and i ≠ j;

[0082] Constraint two in the prior constraint term Ψ (R) of the reflectance component: pixels with similar elevation information have a greater probability of having similar reflectance;

[0083] R i = h ij R j

[0084] wherein R i and R j represent the reflectance information of pixels i and j in R respectively, h ij represents the elevation similarity measure between pixels i and j in R;

[0085] In summary, the prior constraint term of the reflectance component is represented as follows:

[0086]

[0087] wherein θ w and θ h represent the weights of the two reflectance priors respectively;

[0088] The prior constraint term Υ (A) of the abundance information is based on the assumption that for a natural scene, the number of ground object types contained in the scene is limited, and therefore, the distribution is sparse on the global.

[0089] The other steps and parameters are the same as one of the specific implementations one to six.

[0090] Specific implementation eight: different from one of the specific implementations one to seven, the prior constraint term Y(A) of the abundance information in the objective function is expressed as follows:

[0091] Y(A) = γ||A|| 1,1

[0092] Wherein, γ represents a coefficient, |||| represents a norm. 1,1

[0093] The other steps and parameters are the same as one of the specific implementations one to seven.

[0094] Specific implementation nine: different from one of the specific implementations one to eight, the alternating direction multiplier method is used in the step three to solve the multi-modal remote sensing image spectral unmixing model based on DSM and hyperspectral joint constructed in the step two, to obtain more accurate abundance information, and the hyperspectral image with the original resolution is reconstructed based on the abundance information; the specific process is as follows:

[0095] The objective function of the multi-modal remote sensing image spectral unmixing model based on DSM and hyperspectral joint is as follows

[0096]

[0097] s.t.A≥0

[0098]

[0099] When the alternating direction multiplier method is used, the parameters R = J and A = K need to be introduced, and the objective function of the multi-modal remote sensing image spectral unmixing method based on DSM and hyperspectral joint is rewritten as follows:

[0100]

[0101] s.t.R = J, A = K, 1 T A = 1

[0102] Wherein, represents a set containing R, A, J, and K.

[0103] J and K represent auxiliary variables.

[0104] The above formula is changed into the Lagrange augmented function of the multi-modal remote sensing image spectral unmixing model based on DSM and hyperspectral joint as follows:

[0105]

[0106] s.t.1 T A = 1

[0107] ​Wherein, Λ1 and Λ2 are Lagrange multipliers, σ1 and σ2 are penalty coefficients; is a Lagrange augmented function;

[0108] The reflectance component R, the auxiliary variable J, the auxiliary variable K and the abundance information A are obtained by optimizing the Lagrange augmented function based on the idea that the zero point of the first derivative of the constraint model is an extreme point.

[0109] The other steps and parameters are the same as one of the first to eighth embodiments.

[0110] The tenth embodiment is different from one of the first to ninth embodiments in that the reflectance component R, the auxiliary variable J, the auxiliary variable K and the abundance information A are obtained by optimizing the Lagrange augmented function based on the idea that the zero point of the first derivative of the constraint model is an extreme point.

[0111] The specific process is as follows:

[0112] 1) Optimize the reflectance component R:

[0113] The objective function of the reflectance component R is expressed as follows:

[0114]

[0115] The above formula is also expressed as follows:

[0116]

[0117] Wherein, T represents an auxiliary variable; S⊙T=Y, The equivalent expression is The solution is as follows:

[0118]

[0119] T0=2W

[0120] T1=R-WR

[0121] T2=2H

[0122] T3=R-HR

[0123] Wherein, T0, T1, T2, T3 represent auxiliary variables;

[0124] 2) Optimize the auxiliary variable J and the auxiliary variable K:

[0125] J=σ1R-Λ1 / σ1

[0126] K=sign(A-Λ2 / σ2)⊙max{0,A-Λ2 / σ2|-1 / σ2}

[0127] 3) Optimize the abundance information A:

[0128] The objective function of the abundance information A is expressed as follows:

[0129]

[0130] The solution is as follows:

[0131] A = (2σ2K - E T R - Λ2I) x (2σ2 - E T E) -1 .

[0132] The other steps and parameters are the same as one of the first to ninth embodiments.

[0133] The beneficial effects of the present application are verified by the following embodiments:

[0134] Embodiment one:

[0135] A multi-modal remote sensing image spectral unmixing method based on DSM and hyperspectral joint, specifically prepared according to the following steps:

[0136] The multi-modal data set used in the experiment includes a hyperspectral image and a DSM image, and the data has been preprocessed by atmospheric correction, geometric correction, etc. The hyperspectral data contains 349x1905 pixels, which is obtained by ITERS CASI-1500 sensor in the campus of Houston University in Texas, USA and the surrounding rural areas. The Hyperion sensor works in the wavelength range of 364-1046 nm, with a spatial resolution of 2.5 m and a total number of 144 bands. The experiment uses the 170x170 size sub-pixel data set after image cropping as the experimental data set.

[0137] The false color synthesis diagram of the hyperspectral image and the DSM data image is shown in Figure 2a and Figure 2b The experiment takes the original large scene original hyperspectral image as the baseline, calculates the reconstruction error and reconstruction signal-to-noise ratio between the abundance estimation result and the data reconstructed by the endmember information and the original image, compares seven representative methods, Figure 3 gives the abundance map of different ground objects, and Table 1 gives the quantitative indicators of different comparison methods. It can be seen that the abundance estimation result is more accurate in visual discrimination, and the reconstruction error and reconstruction data signal-to-noise ratio result are better than other comparison methods. The experimental results verify the effectiveness of the multi-modal remote sensing image spectral unmixing method based on DSM and hyperspectral joint proposed in the present application.

[0138] Table 1 Quantitative indicators of different comparison methods

[0139]

[0140] The present application can also have other various embodiments, and those skilled in the art can make various corresponding changes and modifications according to the present application without departing from the spirit and essential characteristics of the present application, and these corresponding changes and modifications should all belong to the protection scope of the claims of the present application.

Claims

1. A method for spectral unmixing of multimodal remote sensing images based on a combination of digital surface model (DSM) and hyperspectral imaging, characterized by: The specific process of the method is as follows: Step 1: Acquire multimodal remote sensing images with the same geographic coverage and resolution; Geometric component information of land cover types based on DSM remote sensing image data; The DSM is a digital surface model; The multimodal remote sensing images include hyperspectral remote sensing image data and DSM remote sensing image data; Step 2: Construct a multimodal remote sensing image spectral unmixing model based on the joint use of DSM and hyperspectral imaging; the specific process is as follows: Step 2.1: Introduce the intrinsic image decomposition model. The intrinsic image decomposition model will decompose the hyperspectral remote sensing image Y... H Decompose into the dot product of the reflectance component S and the brightness component R: Among them, Y H Represents hyperspectral remote sensing images; Indicates the light and dark components; Represents the reflectivity component; Step 22: Introduce a linear spectral unmixing model; Steps 2 and 3: Based on the intrinsic image decomposition model and the linear spectral unmixing model, construct the objective function of the multimodal remote sensing image spectral unmixing model based on the joint DSM and hyperspectral model; In step two, a linear spectral unmixing model is introduced; the expression is as follows: Yes H =EA+N in, Represents the endmember matrix, Represents the abundance matrix, Indicates measurement error; C represents the number of endmembers; In steps two and three, based on the intrinsic image decomposition model and the linear spectral unmixing model, an objective function is constructed for a multimodal remote sensing image spectral unmixing model based on the joint DSM and hyperspectral methods; the objective function is expressed as follows: stA≥0 in, The square of the Frobenius norm is represented; α and β are penalty coefficients for balancing the importance of each constraint term; M represents ambient lighting; G represents the geometric components; 1 P Let represent a P-dimensional matrix of all 1s, T denotes the transpose; Ψ(R) represents the prior constraint term for the reflectance component, Υ(A) represents the prior constraint term for the abundance information, U represents the objective function; ⊙ represents the Hadmar product. Step 3: Solve the spectral unmixing model of the multimodal remote sensing image based on DSM and hyperspectral combined, constructed in Step 2, using the alternating direction multiplier method to obtain abundance information.

2. The method for spectral unmixing of multimodal remote sensing images based on the combination of digital surface model (DSM) and hyperspectral imaging as described in claim 1, characterized in that: In step one, multimodal remote sensing images with the same geographical coverage area and the same resolution are obtained; Geometric component information of land cover types based on DSM remote sensing image data; The DSM is a digital surface model; The multimodal remote sensing images include hyperspectral remote sensing image data and DSM remote sensing image data; The specific process is as follows: Step 11 make and These represent hyperspectral remote sensing image data and DSM remote sensing image data acquired with the same geographical coverage and resolution, respectively. in, The spectral feature of the k-th pixel in the hyperspectral remote sensing image data is represented by k = 1, 2, ..., P, where k represents the k-th pixel in the hyperspectral image and P represents the total number of pixels in the hyperspectral image. λ1 represents the first wavelength, λ2 represents the second wavelength, and λ3 represents the third wavelength. This represents the wavelength of the first band of the k-th pixel. This represents the wavelength of the second band of the k-th pixel. Let represent the wavelength of the d-th band of the k-th pixel, with the superscript T indicating transpose. D k This represents the elevation corresponding to the k-th pixel in the DSM remote sensing image data; d H d represents the spectral dimension of a sample in a hyperspectral remote sensing image. 1 This represents the spectral dimension of a sample in a DSM remote sensing image. Representation domain; Steps 1 and 2: Calculate the normal vector for each pixel to obtain the normal vector features as follows: Where N represents the characteristics of the normal vector, n k Let n represent the projection of the normal of the k-th pixel onto the x, y, z coordinate axes in space. k =[n k,x ,n k,y ,n k,z ] T k = 1, 2, ..., P; n k,x Let n represent the projection of the normal of the k-th pixel onto the x-axis in space. k,y Let n represent the projection of the normal of the k-th pixel onto the y-space coordinate axis. k,z This represents the projection of the normal of the k-th pixel onto the z-space coordinate axis; Step 13: Calculate geometric component G based on steps 1 and 2: G=[G1,G2,...,G k ,...,G P ] T Among them, G k Let P represent the geometric component of the k-th pixel, where k = 1, 2, ..., P; G k ＝ [c4,2c2n k,y ,2c2n k,x c2n k,z ,2c1n k,x n k,y ,2c1n k,y n k,z c3n k,z n k,z -c5,2c1n k,x n k,z ,c1(n k,x n k,x -n k,y n k,y )] T Where c1, c2, c3, c4, and c5 are all constants.

3. The method for spectral unmixing of multimodal remote sensing images based on the combination of digital surface model (DSM) and hyperspectral imaging as described in claim 2, characterized in that: The values ​​of the constants c1, c2, c3, c4, and c5 are: c1 = 0.429, c2 = 0.512, c3 = 0.743, c4 = 0.886, and c5 = 248.

4. The method for spectral unmixing of multimodal remote sensing images based on the combination of digital surface model (DSM) and hyperspectral imaging as described in claim 3, characterized in that: The process for obtaining the prior constraint term Ψ(R) of the reflectivity component in the objective function is as follows: Constraint 1 in the prior constraint term Ψ(R) of the reflectivity component: R i =w ij R j Among them, R i and R j w represents the reflectance information of pixel i and pixel j in R, respectively. ij This represents a similarity measure between pixel i and pixel j in R; i = 1, 2, ..., P, where i represents the i-th pixel in the hyperspectral image and P represents the total number of pixels in the hyperspectral image; j = 1, 2, ..., P, where j represents the j-th pixel in the hyperspectral image, and i ≠ j; Constraint 2 in the prior constraint term Ψ(R) of the reflectivity component: R i =h ij R j Among them, R i and R j h represents the reflectance information of pixel i and pixel j in R, respectively. ij This represents the elevation similarity measure between pixel i and pixel j in R; In summary, the prior constraints for the reflectivity components are expressed as follows: Where, θ w and θ h These represent the weights of the two reflection priors, respectively.

5. The method for spectral unmixing of multimodal remote sensing images based on the combination of digital surface model (DSM) and hyperspectral imaging as described in claim 4, characterized in that: The prior constraint term Y(A) for abundance information in the objective function is expressed as follows: Y(A)=γ||A|| 1,1 Where γ represents the coefficient, || || 1,1 It represents the first norm.

6. The method for spectral unmixing of multimodal remote sensing images based on the combination of digital surface model (DSM) and hyperspectral imaging as described in claim 5, characterized in that: In step three, the alternating direction multiplier method is used to solve the multimodal remote sensing image spectral unmixing model based on DSM and hyperspectral combined constructed in step two to obtain abundance information; the specific process is as follows: The objective function of the multimodal remote sensing image spectral unmixing model based on DSM and hyperspectral integration is as follows: stA≥0 By introducing parameters R=J and A=K, the objective function of the multimodal remote sensing image spectral unmixing method based on DSM and hyperspectral integration is rewritten in the following form: s.t.R=J,A=K,1 T A=1 in, Represents a set containing R, A, J, and K; J and K represent auxiliary variables; The Lagrange augmented function of the above equation, transformed into a multimodal remote sensing image spectral unmixing model based on DSM and hyperspectral integration, is: s.t.1 T A=1 Where Λ1 and Λ2 are Lagrange multipliers, and σ1 and σ2 are penalty coefficients; It is a Lagrange augmented function; Based on the idea that the zero point of the first derivative of the constraint model is the extreme point, the Lagrange augmented function is optimized to obtain the reflectance component R, auxiliary variable J, auxiliary variable K and abundance information A.

7. The method for spectral unmixing of multimodal remote sensing images based on a digital surface model (DSM) and hyperspectral imaging as described in claim 6, characterized in that: The Lagrange augmented function is optimized based on the idea that the zero point of the first derivative of the constraint model is the extreme point, so as to obtain the reflectance component R, auxiliary variable J, auxiliary variable K and abundance information A. The specific process is as follows: 1) Optimize the reflectivity component R: The objective function for the reflectivity component R is expressed as follows: The above formula can also be expressed as follows: Where T represents an auxiliary variable; S⊙T=Y, Equivalent representation is The solution results are as follows: T0 = ​​2W T1 = R - WR T2 = 2H T3 = R - HR Where T0, T1, T2, and T3 represent auxiliary variables; 2) Optimize auxiliary variables J and K: J=σ1R-Λ1 / σ1 K=sign(A-Λ2 / σ2)⊙max{0,|A-Λ2 / σ2|-1 / σ2} 3) Optimize abundance information A: The objective function for abundance information A is expressed as follows: The solution results are as follows: A=(2σ2K-E T R-Λ2I)×(2σ2-E T E) -1 。

Images