A hyperspectral satellite image fusion method based on spectrum analysis and its application

Abstract

The present invention provides a hyperspectral satellite image fusion method and application based on spectrum analysis. Aiming at the demand for quantitative analysis of surface elements, the present invention introduces harmonic analysis (HA), guided filtering (GF) and Gram‑Schmidt (GS) algorithm based on hyperspectral satellite images and high spatial resolution multispectral satellite images from the perspective of time-frequency analysis to construct an image fusion method (HGF), which can obtain remote sensing images with both high spatial and high spectral resolution (spatial spectrum). The present invention takes the pixel spectral curve as the processing unit, considers the differences in the characteristic bands of different surface elements and the optimal spatial spectrum fusion method for their matching, improves the spatial resolution of hyperspectral satellite images and the spectral fidelity of fused images, provides a new method for efficient fusion of hyperspectral satellite images, and enhances the applicability of quantification of surface elements of hyperspectral satellite images.

Description

Technical Field

[0001] The invention relates to a hyperspectral satellite image fusion method based on spectrum analysis and application, and belongs to the field of hyperspectral satellite remote sensing image processing. Background Art

[0002] The continuous launch of a large number of remote sensing satellites has further improved the ability to obtain high-resolution remote sensing data. The available hyperspectral satellite images include Hyperion, HJ-1A, Gaofen-5 (GF-5) and Resource-1-02D ​​(ZY1-02D), etc. The optimal spatial resolution is 30m. Therefore, it is very difficult to obtain image data with high spatial and spectral resolution at the same time, which, to a certain extent, limits its development in earth observation technology and also hinders the quantitative monitoring of surface elements. In order to alleviate the contradiction between the spatial resolution and spectral resolution of satellite imaging systems, with the help of the concept of remote sensing image data fusion and the study of its theoretical methods, the spatial-spectral fusion method is applied to improve the resolution of the image, aiming to obtain remote sensing images with both high spatial resolution and high spectral resolution, so as to achieve more comprehensive and accurate land surface monitoring.

[0003] Traditional image fusion methods include component replacement, multi-resolution analysis, and Bayesian matrix decomposition. Component replacement and multi-resolution analysis fusion methods are the two most commonly used methods, and the multi-resolution analysis fusion method has better spectral fidelity. Afterwards, panchromatic-hyperspectral image fusion based on models and mixed pixel decomposition has also been developed. With the rapid improvement of the spatial resolution of satellite-borne images, the spatial resolution ratio between images has gradually increased. In order to solve the problem of the difficulty in effectively integrating the spatial-spectral information of images under large spatial resolution differences, a strategy of step-by-step fusion with high-resolution images of adjacent spatial resolution is proposed. In recent years, deep learning methods have shown excellent application prospects and have been widely used in the fusion of hyperspectral images, but this method needs to be based on a large number of training samples. From the perspective of the spectral curve of a single pixel, combined with time-frequency analysis technology, good results have been achieved in the fusion of panchromatic and hyperspectral image data from the same satellite.

[0004] The current algorithms and models are mainly used for ground feature interpretation, and their application in the quantitative service of surface elements needs further exploration. The quantitative research of surface elements requires high spatial resolution and spectral fidelity of hyperspectral satellite images, and the optimal image fusion methods matched to the research of different surface elements are different, and the applicable wavelength ranges are also different. Therefore, considering the needs of quantitative research of different surface elements, it is extremely important to make full use of remote sensing image data from different satellites from the perspective of single pixel spectral analysis and build an efficient satellite hyperspectral image fusion model. Summary of the invention

[0005] The technical problem to be solved by the present invention is: to explore the application potential of time-frequency analysis technology in the fusion of remote sensing images from different satellites and the optimal fusion method suitable for different surface elements, and to propose a method for spectrum fusion and its applicability evaluation to surface elements. This method can efficiently fuse remote sensing images from different satellites, and at the same time evaluate the optimal fusion method for the quantitative needs of soil, vegetation and water bodies. The operation is relatively convenient and reliable, and can promote the development of hyperspectral satellite image processing technology and its application in the quantitative inversion of surface elements.

[0006] In order to solve the above technical problems, the present invention proposes a hyperspectral satellite image fusion method based on spectrum analysis. The hyperspectral satellite image fusion method based on spectrum analysis of the present invention, aimed at the needs of quantitative analysis of surface elements, based on hyperspectral satellite images and high spatial resolution multispectral satellite images, from the perspective of time-frequency analysis, introduces harmonic analysis (HA), guided filtering (GF) and Gram-Schmidt (GS) algorithm to construct an image fusion method (HGF), which can obtain remote sensing images with both high spatial and high spectral resolution (spatial spectrum), and evaluates the fusion effect in three band intervals according to the spectral response of soil, vegetation and water surface elements.

[0007] The hyperspectral satellite image fusion method based on spectrum analysis of the present invention comprises the following steps:

[0008] Step 1: Data preparation

[0009] Hyperspectral satellite images and high-resolution multispectral satellite images covering the same area are prepared. The hyperspectral image is a ZY1-02D ​​satellite image with a spectral range of 0.40-2.50 μm, a spectral resolution of 10 nm and 20 nm in the visible near-infrared band and short-wave infrared band, and a spatial resolution of 30 m. The high-resolution multispectral satellite image is a blue light band of Sentinel-2B with a spatial resolution of 10 m.

[0010] Step 2: Data preprocessing

[0011] The image in step 1 is calibrated by radiation, atmospheric, orthorectified, and geometrically corrected. The geometric correction is based on the ZY1-02D ​​satellite image, and the blue light band of the Sentinel-2B satellite image is geometrically aligned. The image is then cropped using the target area vector file.

[0012] Step 3: Harmonic decomposition of ZY1-02D ​​hyperspectral satellite imagery

[0013] The single pixel spectrum of the ZY1-02D ​​satellite image processed in step 2 is used as the processing unit. According to the definition of harmonic analysis (HA), the HA of the pixel spectrum is approximately understood as representing the spectrum curve in the form of superposition of sine (cosine) curves. The HA of the spectrum is to decompose the spectrum curve into multiple spectra of different frequencies, and then superimpose the spectra of different frequencies to represent the original spectrum, that is, first perform harmonic decomposition of the pixel spectrum.

[0014] According to the Fourier expansion form of the periodic waveform, for the pixel spectral curve in the ZY1-02D ​​satellite image, it can be expressed in the form of Fourier series after transformation and expansion.

[0015]

[0016] Where: is the hth harmonic component, A h , B h , C h , The calculation formula is

[0017]

[0018] Where: x(n) is the discrete spectral curve, n represents the band number, L represents the total number of bands, A 0 / 2 represents the harmonic remainder, h represents the decomposition number, C h represents the amplitude of the hth harmonic component, It represents the phase of the hth harmonic component. After harmonic transformation, the dimension of the spectrum curve is W=2h+1.

[0019] It can be seen from the above formula that each pixel spectrum is composed of a series of sine (cosine) component curves, and each sine (cosine) curve is composed of the harmonic remainder A. 0 / 2, Amplitude C h and Phase composed of.

[0020] Step 4: Harmonic reconstruction of ZY1-02D ​​hyperspectral satellite imagery

[0021] The harmonic reconstruction parameters of ZY1-02D ​​satellite hyperspectral image are divided into three parts: A 0 / 2 and Sentinel-2B satellite image blue band GS transformation fusion data, C h and

[0022] According to formulas (1) and (2), A 0 / 2 is a constant that does not affect the waveform of the spectral curve. It is the comprehensive response of the ground object to the reflection of electromagnetic waves and also contains the spatial information of the image. 0 / 2 can achieve space-spectrum fusion by inverse harmonic reconstruction, but the blue band of Sentinel-2B satellite image is There are huge differences between pixel grayscale values.

[0023] Therefore, for A 0 / 2 and Sentinel-2B satellite images by GS transformation to improve A 0 / 2 spatial resolution, and at the same time make the pixel gray value of the fused image after GS transformation closer to A 0 / 2, and then transform the GS fusion image, C h and Perform inverse harmonic reconstruction transform to complete the fusion of high-spectral and high-spatial resolution images that are closer to the real spectral reflectance of the ground objects.

[0024] Step 5: Optimization of spatial-spectral fusion images

[0025] Guided filtering can not only reduce noise, but also has the function of edge preservation when the guided image is the original image, so it becomes an edge preservation filter. It has corresponding applications in image enhancement, target detection, image defogging and image classification. The principle of guided filtering is as follows:

[0026] The harmonic reconstructed inverse transform image p obtained in step 4 is input, and the output image O is obtained after filtering through the guide image I, which is the guided filtering. For the pixel point at position i, the obtained filtering output is a weighted average, and the guided filtering is shown as follows:

[0027]

[0028] Where i and j represent pixel subscripts. ij is a filter kernel that is only related to the guidance image I. The filter is linear with respect to p.

[0029] An important assumption of guided filtering is that the output image O and the guided image I are in a window w centered at pixel k and with a filter radius r. k There is a local linear relationship on :

[0030]

[0031] In the formula, w k is a square window of size (2r+1)×(2r+1), i represents the pixel index, a k and b kis the coefficient, in window w k remains unchanged.

[0032] Step 6: Accuracy evaluation of fused images

[0033] The evaluation indicators of hyperspectral image fusion accuracy can be roughly divided into two categories, spatial information integration and spectral information fidelity. The spatial information integration includes correlation coefficient (CC), standard deviation (STD), peak signal-to-noise ratio (PSNR), structural similarity and relative global error, and the spectral information fidelity includes spectral angle (SAM), deviation index, etc. In order to verify the effectiveness of the HGF image fusion algorithm proposed in this invention, it is evaluated from qualitative and quantitative aspects. The qualitative evaluation adopts visual effect for evaluation, and the quantitative evaluation takes the original hyperspectral image as the standard and adopts four indicators of CC, STD, PSNR and SAM for evaluation. The calculation formula is as follows:

[0034]

[0035] Where M, N are the width and height of the image; I H and I W are the fused image and the original image respectively; and are the pixel averages of the summed image and the original image, respectively.

[0036]

[0037] Where M, N are the width and height of the image; μ is the mean; I is the pixel value of the image at position i, j.

[0038]

[0039]

[0040] In the formula, L represents the difference between the maximum and minimum grayscale values ​​of the ideal reference image, and the difference is usually 255; MSE is the mean square error; M and N are the image width and height; I(i,j) and K(i,j) are the pixel values ​​at the corresponding positions.

[0041]

[0042] In the formula, is the spectral vector of the (i, j)th pixel of the fused image, is the spectral vector of the (i, j)th pixel of the original image.

[0043] The distribution of spectral characteristic bands for surface element research is analyzed. For example, water body research is mainly in the visible light band, vegetation research is mainly concentrated in the visible light and near-infrared bands, and soil research is in the visible light, near-infrared and mid-infrared bands. Therefore, in the quantitative evaluation, considering the needs of the later quantification of surface elements, the water vapor absorption band and the number of bands of hyperspectral satellite images, the entire spectral range is divided into three band intervals (band 1: 390-730nm, band 2: 730-1400nm, band 3: 1400-2260nm), and the soil, vegetation and water bodies are evaluated separately in the spectral angle index, so as to select the optimal algorithm for the quantitative analysis of surface elements in different band intervals.

[0044] Furthermore, the present invention also provides an application of a hyperspectral satellite image fusion method based on spectrum analysis in the quantification of surface elements in hyperspectral satellite images. The surface elements include soil, vegetation and water bodies.

[0045] The present invention has the following beneficial effects:

[0046] 1. The present invention takes the pixel spectral curve as the processing unit, takes into account the differences in the characteristic bands of different surface elements and their matching optimal spatial-spectral fusion methods, improves the spatial resolution of hyperspectral satellite images and the spectral fidelity of fused images, provides a new method for the efficient fusion of hyperspectral satellite images, and enhances the applicability of hyperspectral satellite image surface element quantification.

[0047] 2. The method of the present invention introduces the advantages of harmonic analysis, guided filtering and GS transformation, constructs a HGF fusion model based on pixel spectral curves, and effectively improves the effect of hyperspectral satellite image fusion.

[0048] 3. The method of the present invention realizes the remote sensing image fusion of the blue light band of ZY1-02D ​​hyperspectral satellite image and Sentinel-2B satellite image, strengthens the application of fusion of different satellite images, and effectively promotes the development of hyperspectral satellite image fusion.

[0049] 4. The present invention has practical applicability. The method of the present invention can effectively improve the fusion effect of spectral satellite images from the perspective of spectrum conversion, and adopts the method of sub-band evaluation to enhance the applicability of hyperspectral satellite fusion images in the quantification of surface elements, and improve the utilization rate of hyperspectral fusion images in the quantitative analysis of soil, vegetation and water bodies. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a flow chart of the method for fusion of hyperspectral satellite images based on spectrum analysis and its applicability assessment to surface elements of the present invention.

[0051] Figure 2It is the blue light band (2-b) of the ZY1-02D ​​hyperspectral satellite image (2-a) and the Sentinel-2B multispectral satellite image involved in the present invention.

[0052] Figure 3 The fused images are obtained by using PCA, GS and HGF respectively. 3-a is the PCA fused image; 3-b is the GS fused image; 3-c is the HGF fused image.

[0053] Figure 4 The following are the local magnification images of the fused images obtained by using PCA, GS and HGF respectively. 4-a is the local magnification image of the original hyperspectral satellite image of ZY1-02D; 4-b is the local magnification image of the PCA fused image; 4-c is the local magnification image of the GS fused image; 4-d is the local magnification image of the HGF fused image.

[0054] Figure 5 It is a spectrum comparison diagram of the fused image of the present invention. Among them, 5-a is a spectrum comparison diagram of pure soil pixels of the fused image; 5-b is a spectrum comparison diagram of pure vegetation pixels of the fused image; 5-c is a spectrum comparison diagram of pure water pixels of the fused image. DETAILED DESCRIPTION

[0055] The present invention is described in detail below in conjunction with the accompanying drawings to make the technical route and operation steps of the present invention clearer.

[0056] The flowchart of a method for fusion of hyperspectral satellite images based on spectrum analysis proposed in an embodiment of the present invention is as follows: Figure 1 As shown, the following steps are included:

[0057] Step 1. Data preparation. Prepare hyperspectral satellite images and high-resolution multispectral satellite images covering the same area. The hyperspectral image is a ZY1-02D ​​satellite image acquired on June 28, 2020, with a spectral range of 0.40-2.50 μm, a spectral resolution of 10 nm and 20 nm in the visible near-infrared band and short-wave infrared band, and a spatial resolution of 30 m. The high-resolution multispectral satellite image is a blue band of Sentinel-2B, acquired on June 20, 2020, with a spatial resolution of 10 m. The original data of the two images are as follows: Figure 2 shown.

[0058] Step 2: Data preprocessing: The image in step 1 is subjected to radiometric calibration, atmospheric correction, orthorectification and geometric correction. The geometric correction is based on the ZY1-02D ​​satellite image, the blue light band of the Sentinel-2B satellite image is geometrically aligned, and the image is cropped using the target area vector file.

[0059] Step 3: Use PCA, GS and HGF to perform fusion

[0060] 3-1: Principal component analysis (PCA) fusion of ZY1-02D ​​satellite images.

[0061] First, principal component analysis is performed on the original ZY1-02D ​​hyperspectral satellite image, and the principal component components are calculated according to the eigenvalues ​​and eigenvectors of the correlation matrix between the image bands; the first principal component component is replaced with the blue band of the high spatial resolution Sentinel-2B multispectral satellite image, and the principal component inverse transformation is performed on the other principal component components to obtain the PCA fusion image ( Figure 3 3-a).

[0062] 3-2: ZY1-02D ​​satellite image GS fusion.

[0063] The low spatial resolution ZY1-02D ​​hyperspectral satellite image is converted to orthogonal space through GS transformation, and the first component is replaced by the blue band of the high spatial resolution Sentinel-2B multispectral satellite image; all hyperspectral bands are fused and the loss of spectral information of the fused image is reduced. The GS fused image is as follows: Figure 3 As shown in 3-b.

[0064] 3-3: Harmonic decomposition of ZY1-02D ​​hyperspectral satellite imagery.

[0065] The single pixel spectrum of the ZY1-02D ​​satellite image processed in step 2 is used as the processing unit. According to the definition of harmonic analysis (HA), the HA of the pixel spectrum is approximately understood as representing the spectrum curve in the form of superposition of sine (cosine) curves. The HA of the spectrum is to decompose the spectrum curve into multiple spectra of different frequencies, and then superimpose the spectra of different frequencies to represent the original spectrum, that is, first perform harmonic decomposition of the pixel spectrum.

[0066] According to the Fourier expansion form of the periodic waveform, for the pixel spectral curve in the ZY1-02D ​​satellite image, it can be expressed in the form of Fourier series after transformation and expansion.

[0067]

[0068] Where: is the hth harmonic component, A h , B h , C h , The calculation formula is

[0069]

[0070] Where: x(n) is the discrete spectral curve, n represents the band number, L represents the total number of bands, A 0 / 2 represents the harmonic remainder, h represents the decomposition number, C h represents the amplitude of the hth harmonic component, It represents the phase of the hth harmonic component. After harmonic transformation, the dimension of the spectrum curve is W=2h+1.

[0071] From the above formula, it can be seen that each pixel spectrum is composed of a series of sine (cosine) component curves, and each sine (cosine) curve is composed of the harmonic remainder A. 0 / 2, Amplitude C h and Phase composed of;

[0072] 3-4: Harmonic reconstruction of ZY1-02D ​​hyperspectral satellite imagery.

[0073] According to formulas (1) and (2), A 0 / 2 is a constant that does not affect the waveform of the spectral curve. It is the comprehensive response of the ground object to the reflection of electromagnetic waves and also contains the spatial information of the image. 0 / 2 can achieve space-spectrum fusion by inverse harmonic reconstruction, but the blue band of Sentinel-2B satellite image is There are huge differences between pixel grayscale values.

[0074] Therefore, for A 0 / 2 and Sentinel-2B satellite images by GS transformation to improve A 0 / 2 spatial resolution, and at the same time make the pixel gray value of the fused image after GS transformation closer to A 0 / 2, and then transform the GS fusion image, C h and Perform inverse harmonic reconstruction to complete the fusion of high-spectral and high-spatial-resolution images that are closer to the real spectral reflectance of the ground objects. HGF fused images such as Figure 3 As shown in 3-c;

[0075] Step 4: Optimization of spatial-spectral fusion images. Guided filtering can not only reduce noise, but also has the function of edge preservation when the guided image is the original image, so it becomes an edge preservation filter. It has corresponding applications in image enhancement, target detection, image defogging and image classification. The principle of guided filtering is as follows:

[0076] The harmonic reconstructed inverse transform image p obtained by the above steps is input, and the output image O is obtained after filtering through the guide image I, which is the guided filtering. For the pixel point at position i, the obtained filtering output is a weighted average, and the guided filtering is shown as follows:

[0077]

[0078] Where i and j represent pixel subscripts. ij is a filter kernel that is only related to the guidance image I. The filter is linear with respect to p.

[0079] An important assumption of guided filtering is that the output image O and the guided image I are in a window w centered at pixel k and with a filter radius r. k There is a local linear relationship on :

[0080]

[0081] In the formula, w k is a square window of size (2r+1)×(2r+1), i represents the pixel index, a k and b k is the coefficient, in window w k remains unchanged;

[0082] Step 5: Accuracy evaluation of fused images.

[0083] The evaluation indicators of hyperspectral image fusion accuracy can be roughly divided into two categories, spatial information integration and spectral information fidelity. The spatial information integration includes correlation coefficient (CC), standard deviation (STD), peak signal-to-noise ratio (PSNR), structural similarity and relative global error, and the spectral information fidelity includes spectral angle (SAM), deviation index, etc. In order to verify the effectiveness of the HGF image fusion algorithm, it is evaluated from both qualitative and quantitative aspects. The qualitative evaluation uses visual effects for evaluation, and the quantitative evaluation uses the original hyperspectral image as the standard, using CC, STD, PSNR and SAM as the four indicators for evaluation. The calculation formula is as follows:

[0084]

[0085] Where M, N are the width and height of the image; I H and I W are the fused image and the original image respectively; and are the pixel averages of the summed image and the original image, respectively.

[0086]

[0087] Where M, N are the width and height of the image; μ is the mean; I is the pixel value of the image at position i, j.

[0088]

[0089]

[0090] In the formula, L represents the difference between the maximum and minimum grayscale values ​​of the ideal reference image, and the difference is usually 255; MSE is the mean square error; M and N are the width and height of the image; I(i, j) and K(i, j) are the pixel values ​​at the corresponding positions.

[0091]

[0092] In the formula, is the spectral vector of the (i, j)th pixel of the fused image, is the spectral vector of the (i, j)th pixel of the original image.

[0093] The distribution of spectral characteristic bands for surface element research is analyzed. For example, water body research is mainly in the visible light band, vegetation research is mainly concentrated in the visible light and near-infrared bands, and soil research is in the visible light, near-infrared and mid-infrared bands. Therefore, in the quantitative evaluation, considering the needs of the later quantification of surface elements, the water vapor absorption band and the number of bands of hyperspectral satellite images, the entire spectral range is divided into three band intervals (band 1: 390-730nm, band 2: 730-1400nm, band 3: 1400-2260nm), and the soil, vegetation and water bodies are evaluated separately in the spectral angle index, so as to select the optimal algorithm for the quantitative analysis of surface elements in different band intervals.

[0094] The fused images and local location magnification images obtained using the three methods of PCA, GS and HGF are shown in the figure. Figure 3 and Figure 4 As shown in the figure, qualitative analysis shows that the HGF method proposed in the present invention is better. It can retain the original image details and improve the image quality to the maximum extent, and at the same time, it has the effect of edge-preserving filtering and noise reduction. Quantitative comparison shows that the HGF and GS fusion methods can be used for the quantitative inversion of water bodies, and the HGF fusion method is the best for the quantitative study of vegetation and soil. Figure 5 The pure pixels at the same position are selected from the four images of ZY1-02D, PCA, GS and HGF, and the extracted spectral curve can be used to show the fusion effect. Figure 5 The specific analysis is as follows:

[0095] If the soil, vegetation and water bodies in the range of 390-730nm are studied, the GS fusion method is better in terms of CC (0.79), STD (2.4) and PSNR (20.76). By comparing the SAM index of the pure pixel spectrum at the same position of the original image and the fused image, it is found that the HGF method is the best, which are 0.02, 0.01 and 0.31 respectively, which are smaller than the SAM index of the PCA and GS fusion images and the original images ( Figure 5 ).

[0096] In the study of soil, vegetation and water information in the 730-1400nm range, HGF fusion image is the best, with CC of 0.82, STD of 3.91 and PSNR of 16.78, all of which are greater than the corresponding indicators of PCA and GS fusion images. By comparing the SAM index of the pure pixel spectrum at the same position of the original image and the fused image, it is found that the HGF method is the best, with SAM of 0.01, 0.01 and 0.78 respectively, all of which are less than the SAM index of PCA and GS fusion images ( Figure 5 ).

[0097] In the range of 1400-2260nm, for further research on soil, vegetation and water bodies, HGF fusion has a better effect, with CC of 0.80, STD of 4.59 and PSNR of 17.62, all of which are greater than the corresponding indicators of PCA and GS fusion images. By comparing the SAM index of the pure pixel spectrum at the same position of the original image and the fused image, it is found that the HGF method is the best, with SAM of 0.02, 0.01 and 0.87 respectively, all of which are less than the SAM index of PCA and GS fusion images ( Figure 5 ).

[0098] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A hyperspectral satellite image fusion method based on spectrum analysis, Its characteristics are: The following steps are involved: Step 1: Data preparation Prepare hyperspectral satellite images and high-resolution multispectral satellite images covering the same area; Step 2: Data preprocessing Perform radiometric calibration, atmospheric correction, orthorectification and geometric correction on the image in step 1; Step 3: Harmonic decomposition of hyperspectral satellite images Taking the single pixel spectrum of the hyperspectral satellite image processed in step 2 as the processing unit, perform frequency domain decomposition; Step 4: Harmonic reconstruction of hyperspectral satellite images Use high-resolution multispectral image replacement and frequency domain reconstruction inverse transformation to achieve spatial-spectral fusion; Harmonic reconstruction parameters of ZY1-02D ​​satellite hyperspectral images A 0 / 2 and Sentinel-2B satellite images by GS transformation to improve A 0 / 2 spatial resolution, and make the pixel grayscale value of the fused image closer to A 0 / 2, and then transform the GS fusion image, C h and Perform inverse harmonic reconstruction transformation to complete the fusion of high-spectral and high-spatial resolution images that are closer to the spectral reflectance of real objects; Step 5: Optimization of spatial-spectral fusion images The spatial-spectral fusion image obtained in step 4 is filtered to improve the image quality while maintaining the edge.

2. A hyperspectral satellite image fusion method based on spectrum analysis as claimed in claim 1, Its characteristics are: In step 1, the hyperspectral image is a ZY1-02D ​​satellite image with a spectral range of 0.40-2.50 μm, a spectral resolution of 10 nm and 20 nm in the visible near-infrared band and the short-wave infrared band, and a spatial resolution of 30 m; the multispectral image is a blue light band of the Sentinel-2B satellite image with a spatial resolution of 10 m.

3. A hyperspectral satellite image fusion method based on spectrum analysis as claimed in claim 2, Its characteristics are: In step 2, the geometric correction is to geometrically align the blue light band of the Sentinel-2B satellite image with the ZY1-02D ​​satellite image as a reference, and to crop the image using the target area vector file.

4. The hyperspectral satellite image fusion method based on spectrum analysis as claimed in claim 3, Its characteristics are: In step 3, a single pixel spectrum of the ZY1-02D ​​satellite image processed in step 2 is used as a processing unit, and the pixel spectrum is approximately represented as a superposition of sine and / or cosine curves. The frequency domain decomposition of the ZY1-02D ​​image is to decompose the pixel spectrum curve into multiple spectra of different frequencies. Each pixel spectrum is composed of a series of sine and / or cosine component curves, and each sine or cosine curve is composed of the remainder A 0 / 2, Amplitude C h and Phase The single pixel spectral curve is used as the processing unit to perform harmonic decomposition on the ZY1-02D ​​satellite hyperspectral image. h = 81.

5. The hyperspectral satellite image fusion method based on spectrum analysis as claimed in claim 4, Its characteristics are: In step 5, guided filtering is used to perform edge-preserving optimization on the spatial-spectral fusion image, the guided filtering radius is set to 7, and the number of decomposition components is set to 4.

6. Application of a hyperspectral satellite image fusion method based on spectrum analysis as described in any one of claims 1 to 5 in quantification of surface elements in hyperspectral satellite images.

7. The use according to claim 6, Its characteristics are: The surface elements include soil, vegetation and water bodies.

8. The use according to claim 7, Its characteristics are: According to the spectral response of soil, vegetation and water surface elements, the effect is evaluated in three bands.

9. The use according to claim 8, Its characteristics are: The three bands are: Band 1: 390-730 nm, Band 2: 730-1400 nm, and Band 3: 1400-2260 nm.

Images