Kernel function enhanced satellite image cloud recognition and removal method

Abstract

The invention relates to the technical field of satellite image processing, in particular to a kernel function enhanced satellite image cloud recognition and removal method, which comprises the following steps that: a satellite image and a historical image corresponding to the satellite image are acquired, the satellite image contains cloud, and the historical image does not contain cloud; carrying out matching processing on the satellite image and the historical image by adopting an image matching algorithm, and carrying out matching transformation on the historical image to obtain the historical image after matching transformation; performing color histogram analysis on the satellite image to obtain a color model, and coloring the matched and transformed historical image based on the color model to obtain a colored historical image; determining a pixel fusion kernel function, determining a pixel-level fusion weight by the pixel fusion kernel function, and performing pixel-level fusion on the satellite image and the coloring historical image by the pixel-level fusion weight to obtain a cloud-removed satellite image; according to the invention, the cloud and mist removal effect of the satellite image can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite image processing, and in particular to a kernel function enhanced satellite image cloud and fog recognition and removal method. Background Art

[0002] With the rapid development of remote sensing technology, satellite images have been widely used in many fields such as meteorology, agriculture, environmental monitoring, and urban planning. However, in practical applications, atmospheric phenomena such as clouds and fog have a significant impact on the quality of satellite images. The presence of clouds and fog not only blocks ground information, but also causes color distortion and reduced contrast of the image. In severe cases, it can even completely cover the target area, thus affecting subsequent image analysis, feature extraction, and target detection tasks.

[0003] Traditional satellite image cloud removal methods mainly rely on physical models and image processing techniques. Early physical models, such as the atmospheric radiative transfer model (RTM), predicted the impact of clouds and fog by simulating the scattering and absorption of electromagnetic waves by the atmosphere. However, these models often require more complex parameter settings and lack sufficient flexibility in practical applications, especially when faced with dynamically changing cloud layers. In addition, traditional image processing techniques, such as histogram equalization and contrast stretching, can improve image quality to a certain extent, but cannot effectively remove color distortion and information occlusion caused by clouds and fog.

[0004] In recent years, with the development of deep learning technology, cloud removal methods based on deep learning models such as convolutional neural networks (CNN) and generative adversarial networks (GAN) have received widespread attention. Deep learning models can learn the characteristics of cloud layers from a large number of satellite images and effectively remove them. However, existing deep learning methods still face some challenges in cloud removal. First, the types and influencing factors of clouds are very complex, and a single deep learning model is often difficult to cope with various types of clouds. Second, deep learning methods usually require a large amount of labeled data for training, and in practical applications, obtaining labeled data is costly and difficult, which limits its widespread application.

[0005] In addition, current cloud removal technologies usually ignore the multi-layer and spatial distribution characteristics of clouds and fog, lack effective cross-layer and multi-scale model design, resulting in unsatisfactory removal effects. In the context of increasing demand for real-time processing, how to improve the computational efficiency and processing speed of cloud removal algorithms remains an urgent problem to be solved. Therefore, existing satellite image cloud removal technologies have certain limitations in terms of accuracy, adaptability, and efficiency, and further research and innovation are needed. Summary of the invention

[0006] In view of the above problems, the present invention provides a kernel function enhanced satellite image cloud and fog recognition and removal method, which solves the technical problems in the prior art such as poor cloud and fog removal effect, low computational efficiency, and inability to accurately restore image details.

[0007] The present invention provides a kernel function enhanced satellite image cloud and fog recognition and removal method, comprising the following steps:

[0008] Step S1, obtaining a satellite image to be defogged and a historical image corresponding to the satellite image to be defogged, wherein the satellite image to be defogged contains fog, and the historical image is consistent with the geographical location where the satellite image to be defogged was taken and does not contain fog;

[0009] Step S2, using an image matching algorithm to match the satellite image to be declouded and the historical image, performing a matching transformation on the historical image to obtain a historical image after matching transformation;

[0010] Step S3, performing color histogram analysis on the satellite image to be declouded to obtain a color model, and coloring the historical image after the matching transformation based on the color model to obtain a colored historical image;

[0011] Step S4, determining a pixel fusion kernel function, wherein the pixel fusion kernel function determines a pixel-level fusion weight based on a difference in image grayscale texture between the satellite image to be declouded and the historical image after matching transformation;

[0012] The satellite image to be declouded and the colored historical image are fused at the pixel level using the pixel level fusion weight to obtain a satellite image with declouds removed.

[0013] Preferably, the step S2 specifically includes:

[0014] Step S2-1, detecting a plurality of feature points in both the satellite image to be defogging and the historical image, calculating a feature descriptor for each feature point, and calculating a matching degree between the feature descriptors in the satellite image to be defogging and the historical image;

[0015] Step S2-2: determining a mapping function for matching transformation of the historical image based on the degree of matching between the satellite image to be declouded and the feature descriptors in the historical image, and performing a matching transformation on the historical image by using the mapping function to obtain a historical image after matching transformation.

[0016] Preferably, the calculation expression of the matching degree between the feature descriptors in step S2-1 is:

[0017]

[0018] Wherein, match(d1, d2) represents the matching degree between two feature descriptors d1 and d2, d1 represents a feature descriptor in the satellite image to be declouded, d2 represents a feature descriptor in the historical image, N represents the dimension of the feature descriptor, d 1k ,d 2k denote the kth component of feature descriptors d1 and d2 respectively, and min(·,·) indicates the calculated minimum value;

[0019] In step S2-2, the mapping function is used to perform matching transformation on the historical image, and the expression of the historical image after matching transformation is obtained as follows:

[0020]

[0021] Wherein, M2 represents the historical image, Ψ(M2) represents the mapping function of the matching transformation of the historical image, Represents the historical image after matching transformation.

[0022] Preferably, the step S3 specifically includes:

[0023] Step S3-1, calculating the color histogram of the satellite image to be declouded, and calculating the cumulative distribution function of the color histogram as the color model;

[0024] Step S3-2, obtaining a mapping relationship based on the color model, wherein the mapping relationship can map the pixel values ​​of the historical image after the matching transformation to new values, so that the mapped image is consistent with the color style of the satellite image to be declouded;

[0025] Step S3-3: Apply the mapping relationship to the historical image after the matching transformation to obtain a colored historical image.

[0026] Preferably, the step S4 specifically includes:

[0027] Step S4-1, determining a cloud region screening matrix based on a three-channel mean matrix and a three-channel standard deviation matrix of the satellite image to be declouded;

[0028] Step S4-2, determining a pixel fusion weight matrix based on the absolute value of the grayscale difference between each pixel in the satellite image to be declouded and the historical image after matching transformation and the cloud area screening matrix;

[0029] Step S4-3, using a bilateral filtering process to process the pixel fusion weight matrix to obtain a smoothed pixel fusion weight matrix;

[0030] Step S4-4: Based on the smoothed pixel fusion weight matrix, the satellite image to be declouded and the colored historical image are fused at the pixel level to obtain a satellite image with declouds removed.

[0031] Preferably, in step S4-1, the cloud and fog area screening matrix Q is expressed as:

[0032]

[0033] Among them, C i,j Represents the value of the three-channel mean matrix C at the i, j coordinate position, Represents C i,j The amount of change, V i,j represents the value of the three-channel standard deviation matrix V at the i, j coordinate position, ν represents the element mean of the three-channel standard deviation matrix, Indicates V i,j The amount of change, Q i,j It represents the value of matrix Q at the i,j coordinate position, 1 represents a potential cloud area, and 0 represents an area that is unlikely to be a cloud area.

[0034] Preferably, in step S4-2, the expression of the pixel fusion weight matrix is:

[0035]

[0036] W = tanh((Δ-m) 2 *100) / 2⊙Q

[0037] Among them, Δ represents the absolute difference matrix of grayscale values, G 1,i,j Represents the grayscale value of the satellite image to be declouded at the i, j coordinate position, represents the gray value of the historical image at the i, j coordinate position after matching transformation, |·| represents the calculation of the absolute value, represents a real number field with dimension w×h, w and h are the maximum values ​​of the horizontal and vertical coordinates respectively, W represents the pixel fusion weight matrix, m represents the 1 / 3 median of the elements in the Δ matrix, Q is the cloud area screening matrix, and represents the Hadamard product operation.

[0038] Preferably, in step S4-3, the smoothed pixel fusion weight matrix is ​​expressed as:

[0039]

[0040] in, represents the smoothed pixel fusion weight matrix, x represents the first element position in the W matrix, x′ represents the second element position in the W matrix, W(x) and W(x′) represent the values ​​of the W matrix at the x and x′ element positions, respectively, xx′ represents the difference between the first element position x and the second element position x′, C(x) represents the normalization constant at the x element position, Represents the spatial distance as z s The spatial Gaussian weights between the elements when ; Indicates that the element difference is z r The range Gaussian weight between elements when , σ s and σ r They are used to calculate the standard deviation of the spatial Gaussian weight and the range Gaussian weight respectively.

[0041] Preferably, in step S4-4, the expression of pixel-level fusion is:

[0042]

[0043] Among them, c represents the color channel, represents the value of the color channel c of the satellite image with the cloud removed, M 1,c Indicates the value of the color channel c of the satellite image to be declouded. Represents the value of the color channel c of the historical image after matching transformation, and ⊙ represents the Hadamard product operation.

[0044] Compared with the prior art, the present invention has at least the following beneficial effects:

[0045] (1) The present invention performs color sampling and coloring through a histogram method, making the color difference between the foggy area and the non-foggy area more prominent, providing more accurate data support for subsequent image matching and texture analysis, ensuring that different areas can be accurately distinguished when processing images, thereby improving the accuracy of the overall defogging effect.

[0046] (2) The present invention can more accurately identify foggy areas by combining image matching and texture feature analysis. Compared with traditional methods, this combination avoids recognition failures caused by the high similarity between foggy areas and other image areas, and can effectively remove foggy areas in complex image scenes, ensuring image clarity and detail integrity.

[0047] (3) The kernel function provided by the present invention can better remove fog and restore details by adjusting the ratio of pixel fusion according to the difference between the matching image and the texture features, thereby ensuring a natural defogging effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The drawings are only for the purpose of illustrating particular embodiments and are not to be construed as limiting the invention.

[0049] Figure 1 A flow chart of the kernel function enhanced satellite image cloud and fog recognition and removal method provided by the present invention;

[0050] Figure 2 Schematic diagram of satellite images and historical images provided by the present invention.

[0051] Figure 3 This is a schematic diagram of the image matching overlapping area provided by the present invention.

[0052] Figure 4 A schematic diagram of the overlapping area and the corresponding fusion weights provided by the present invention.

[0053] Figure 5 This is a schematic diagram of the fusion effect provided by the present invention. DETAILED DESCRIPTION

[0054] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. In addition, the present invention can also be implemented in other ways different from those described herein, and therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0055] The present invention first performs color sampling and coloring on satellite images, then identifies the differences between images through image matching methods, especially utilizes texture features to identify cloud and fog areas, and designs kernel functions to fuse image pixels, thereby achieving efficient and accurate cloud and fog removal and restoring the ground information and details of the image.

[0056] In order to illustrate the effectiveness of the method proposed by the present invention, the above technical solution of the present invention is described in detail below through a specific embodiment. Figure 1 As shown, a kernel function enhanced satellite image cloud and fog recognition and removal method is disclosed, and the specific implementation steps are as follows:

[0057] Step S1, obtaining a satellite image to be defogged and a historical image corresponding to the satellite image to be defogged, wherein the satellite image to be defogged contains fog, and the historical image is consistent with the geographical location where the satellite image to be defogged was taken and does not contain fog.

[0058] like Figure 2As shown, the present invention first obtains a satellite image of the satellite image to be defogged taken by a satellite, wherein the satellite image contains fog. The satellite image to be defogged has a corresponding historical image, and the historical image and the satellite image to be defogged are shot at the same geographical location and do not contain fog, but the color style of the historical image may be inconsistent with that of the satellite image to be defogged. The historical image can be used as a reference for the satellite image to be defogged for subsequent processing.

[0059] Step S2: using an image matching algorithm to match the satellite image to be declouded and the historical image, performing a matching transformation on the historical image to obtain a historical image after matching transformation.

[0060] The present invention uses an image matching algorithm to compare the satellite image of the cloud to be removed with the historical image. The historical image is used as a reference image, and its content is similar to the target image but does not contain the cloud. Through image matching, the cloud area in the satellite image and the corresponding area in the historical image can be accurately located.

[0061] The matching process is completed by calculating the similarity between the images or using, for example, a matching algorithm based on feature points. In some embodiments, the satellite image to be defogging and the historical image are matched by the SURF algorithm, and the historical image is matched and transformed to the corresponding position in the satellite image to be defogging according to the best matching result. The SURF algorithm detects multiple feature points in the two images, calculates a feature descriptor for each feature point, calculates the matching degree between the feature descriptors of the two images, determines the mapping function of the matching transformation according to the best matching result, and finally completes the matching transformation of the historical image. The expression is:

[0062]

[0063] Wherein, match(d1,d2) represents the matching degree between two feature descriptors d1 and d2, d1 represents a feature descriptor in the satellite image to be declouded, d2 represents a feature descriptor in the historical image, M2 represents the historical image, Ψ(M2) represents the mapping function of the matching transformation of the historical image, Represents the historical image after matching transformation.

[0064] The SURF algorithm detects multiple feature points in the image and calculates a feature descriptor for each feature point. These descriptors can be high-dimensional vectors such as 64 or 128 dimensions. N represents the dimension of the feature descriptor, d 1k ,d 2k They represent the kth component of feature descriptors d1 and d2 respectively, and min(·,·) means calculating the minimum value.

[0065] Figure 3 The overlapping area of ​​the historical image after matching transformation in the satellite image to be declouded is shown.

[0066] Step S3: performing color histogram analysis on the satellite image to be declouded to obtain a color model, and coloring the historical image after the matching transformation based on the color model to obtain a colored historical image.

[0067] The present invention performs histogram analysis on the satellite image to be declouded and obtains a color model. Specifically, the satellite image can be first converted into a color space such as HSV or Lab, and a histogram analysis is performed on the image in the color space. The histogram shows the frequency information of the color distribution in the image, and the overall color characteristics of the image can be represented by counting the number of pixels of each color component.

[0068] The step of performing color histogram analysis on the satellite image to be declouded to obtain a color model specifically includes:

[0069] First, the satellite image color histogram of the cloud region to be processed is calculated, and then the cumulative distribution function (CDF) of the histogram is calculated as the color model.

[0070] The CDF is used to regulate the histogram of the historical image. Specifically, a mapping relationship is obtained based on the CDF of the satellite image to be declouded. The mapping relationship can map the pixel values ​​of the historical image colored after the matching transformation to the new value, so that the CDF of the colored historical image is as close as possible to the CDF of the satellite image of the cloud area to be processed, thereby realizing the coloring of the historical image after the matching transformation, so that the color style of the colored historical image is consistent with that of the satellite image of the cloud area to be processed.

[0071] Finally, the mapping relationship is applied to the historical image after the matching transformation to obtain a colored historical image.

[0072] Step S4, determining a pixel fusion kernel function, wherein the pixel fusion kernel function determines a pixel-level fusion weight based on the image grayscale texture difference between the satellite image to be defogging and the historical image after the matching transformation, and the satellite image to be defogging and the colored historical image are fused at the pixel level according to the pixel-level fusion weight to obtain a satellite image with defogging.

[0073] In this step, the cloud and fog area screening matrix Q used for preliminary screening of whether it is a cloud and fog area is first determined. The specific process is as follows:

[0074] Calculate the three-channel mean matrix C of the satellite image to be declouded, and perform the following operations:

[0075]

[0076] in, Represents C i,j The amount of change, C i,j Represents the value of the three-channel mean matrix C at the i, j coordinate position.

[0077] Calculate the three-channel standard deviation matrix V of the satellite image to be declouded and the element mean ν of the three-channel standard deviation matrix, and perform the following operations:

[0078]

[0079] in, Indicates V i,j The amount of change, V i,j represents the value of the three-channel standard deviation matrix V at the i, j coordinate position, and ν represents the element mean of the three-channel standard deviation matrix.

[0080] The cloud area screening matrix Q is obtained. Q is a 0-1 screening matrix for de-merging operation:

[0081]

[0082] Among them, Q i,j It represents the value of matrix Q at the i,j coordinate position, 1 represents a potential cloud area, and 0 represents an area that is unlikely to be a cloud area.

[0083] The present invention provides a kernel function, which determines the pixel fusion weight matrix of two images according to the matching conditions of the images and the difference in grayscale textures. The expression is:

[0084]

[0085] Among them, Δ represents the absolute difference matrix of grayscale values, G 1,i,j Represents the grayscale value of the satellite image to be declouded at the i, j coordinate position, represents the gray value of the historical image at the i, j coordinate position after matching transformation, |·| represents the calculation of the absolute value, represents a real number domain with dimension w×h, w and h are the maximum values ​​of horizontal and vertical coordinates respectively, W represents the pixel fusion weight matrix, and m represents the 1 / 3 median of the elements in the Δ matrix. The tanh function is used to map the difference value to a range, so that pixels with larger differences get larger weights and pixels with smaller differences get smaller weights. Q is the cloud region screening matrix, and ⊙ represents the Hadamard product operation.

[0086] In this way, the kernel function assigns weights to each pixel according to the grayscale texture difference between the foggy area and the fog-free area, thereby performing pixel-level fusion of the image. In this way, the influence of the fog layer can be removed while retaining the details of the original image.

[0087] Since the fusion ratio of each pixel is estimated separately, there is a certain amount of noise in the estimation. Therefore, W is processed through the bilateral filtering process to obtain the final smoothed pixel fusion weight matrix This makes the final fusion effect more natural, and the expression is:

[0088]

[0089] in, represents the smoothed pixel fusion weight matrix, x represents the first element position in the W matrix, x′ represents the second element position in the W matrix, W(x) and W(x′) represent the values ​​of the W matrix at the x and x′ element positions, respectively, and xx′ represents the difference between the first element position x and the second element position x′. C(x) represents the normalization constant at the x element position. Represents the spatial distance as z s The spatial Gaussian weights between the elements when ; Indicates that the element difference is z r The range Gaussian weight between elements when σ s and σ r Calculate the standard deviation of the spatial and range Gaussian weights, respectively.

[0090] Figure 4 The overlapping area of ​​two images and the corresponding smoothed fusion matrix are shown.

[0091] Based on the smoothed pixel fusion weight matrix The satellite image to be de-clouded and the historical image after matching transformation are fused at pixel level to obtain the satellite image with de-clouded. The expression is:

[0092]

[0093] Among them, c represents the color channel, represents the value of the color channel c of the satellite image with the cloud removed, M 1,c The value of the color channel c of the satellite image representing the area where the clouds are to be removed. Represents the value of the color channel c of the historical image after matching transformation, and ⊙ represents the Hadamard product operation.

[0094] In the above manner, the present invention performs color sampling and coloring through a histogram method, highlighting the color difference between foggy and non-foggy areas to support image matching and texture analysis. The image matching method is combined to correspond historical images and satellite images. The kernel function provided by the present invention adjusts the pixel fusion ratio according to image matching and grayscale texture differences, effectively removes foggy and restores details, and achieves a natural defogging effect.

[0095] Figure 5 The satellite image with the final cloud and fog removed is shown. It can be seen from the figure that the method provided by the present invention can effectively match the overlapping area and effectively remove the cloud and fog area, and can also effectively remove the shadow area caused by the cloud and fog. In addition, although the color style of the satellite image is inconsistent with the reference image, the present invention ensures that the color style of the removed area is consistent with the original satellite image while ensuring accurate removal of the cloud and fog.

[0096] Although the specific embodiments of the present invention depict various actions or steps in a specific order, this should be understood as requiring such actions or steps to be performed in the specific order shown or in a sequential order, or requiring all illustrated actions or steps to be performed to obtain the desired results. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Certain features described in the context of a separate embodiment can also be implemented in a single implementation in combination. On the contrary, the various features described in the context of a single implementation can also be implemented in multiple implementations individually or in any suitable sub-combination. The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or replacements that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

[0097] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A kernel function enhanced satellite image cloud and fog recognition and removal method, characterized in that: The following steps are involved: Step S1, obtaining a satellite image to be defogged and a historical image corresponding to the satellite image to be defogged, wherein the satellite image to be defogged contains fog, and the historical image is consistent with the geographical location where the satellite image to be defogged was taken and does not contain fog; Step S2, using an image matching algorithm to match the satellite image to be declouded and the historical image, performing a matching transformation on the historical image to obtain a historical image after matching transformation; Step S3, performing color histogram analysis on the satellite image to be declouded to obtain a color model, and coloring the historical image after the matching transformation based on the color model to obtain a colored historical image; Step S4, determining a pixel fusion kernel function, wherein the pixel fusion kernel function determines a pixel-level fusion weight based on a difference in image grayscale texture between the satellite image to be declouded and the historical image after matching transformation; The satellite image to be declouded and the colored historical image are fused at the pixel level using the pixel level fusion weight to obtain a satellite image with declouds removed.

2. The kernel function enhanced satellite image cloud and fog recognition and removal method according to claim 1, characterized in that: The step S2 specifically includes: Step S2-1, detecting a plurality of feature points in both the satellite image to be defogging and the historical image, calculating a feature descriptor for each feature point, and calculating a matching degree between the feature descriptors in the satellite image to be defogging and the historical image; Step S2-2: determining a mapping function for matching transformation of the historical image based on the degree of matching between the satellite image to be declouded and the feature descriptors in the historical image, and performing a matching transformation on the historical image using the mapping function to obtain a historical image after matching transformation.

3. The kernel function enhanced satellite image cloud and fog recognition and removal method according to claim 2 is characterized in that: The calculation expression of the matching degree between the feature descriptors in step S2-1 is: Wherein, match(d1, d2) represents the matching degree between two feature descriptors d1 and d2, d1 represents a feature descriptor in the satellite image to be declouded, d2 represents a feature descriptor in the historical image, N represents the dimension of the feature descriptor, d 1k ,d 2k denote the kth component of feature descriptors d1 and d2 respectively, and min(·,·) indicates the calculated minimum value; In step S2-2, the mapping function is used to perform matching transformation on the historical image, and the expression of the historical image after matching transformation is obtained as follows: Wherein, M2 represents the historical image, Ψ(M2) represents the mapping function of the matching transformation of the historical image, Represents the historical image after matching transformation.

4. The kernel function enhanced satellite image cloud and fog recognition and removal method according to claim 3 is characterized in that: The step S3 specifically includes: Step S3-1, calculating the color histogram of the satellite image to be declouded, and calculating the cumulative distribution function of the color histogram as the color model; Step S3-2, obtaining a mapping relationship based on the color model, wherein the mapping relationship can map the pixel values ​​of the historical image after the matching transformation to new values, so that the mapped image is consistent with the color style of the satellite image to be declouded; Step S3-3: Apply the mapping relationship to the historical image after the matching transformation to obtain a colored historical image.

5. The kernel function enhanced satellite image cloud and fog recognition and removal method according to claim 4 is characterized in that: The step S4 specifically includes: Step S4-1, determining a cloud region screening matrix based on a three-channel mean matrix and a three-channel standard deviation matrix of the satellite image to be declouded; Step S4-2, determining a pixel fusion weight matrix based on the absolute value of the grayscale difference between each pixel in the satellite image to be declouded and the historical image after matching transformation and the cloud area screening matrix; Step S4-3, using a bilateral filtering process to process the pixel fusion weight matrix to obtain a smoothed pixel fusion weight matrix; Step S4-4: Based on the smoothed pixel fusion weight matrix, the satellite image to be declouded and the colored historical image are fused at the pixel level to obtain a satellite image with declouds removed.

6. The kernel function enhanced satellite image cloud and fog recognition and removal method according to claim 5, characterized in that: In step S4-1, the expression of the cloud and fog area screening matrix Q is: Among them, C i,j Represents the value of the three-channel mean matrix C at the i, j coordinate position, C i,j Represents C i,j The amount of change, V i,j represents the value of the three-channel standard deviation matrix V at the i, j coordinate position, ν represents the element mean of the three-channel standard deviation matrix, V i,j Indicates V i,j The amount of change, Q i,j It represents the value of matrix Q at the i,j coordinate position, 1 represents a potential cloud area, and 0 represents an area that is unlikely to be a cloud area.

7. The kernel function enhanced satellite image cloud and fog recognition and removal method according to claim 6, characterized in that: In step S4-2, the expression of the pixel fusion weight matrix is: Among them, Δ represents the absolute difference matrix of grayscale values, G 1,i,j Represents the grayscale value of the satellite image to be declouded at the i, j coordinate position, represents the gray value of the historical image at the i, j coordinate position after matching transformation, |·| represents the calculation of the absolute value, represents a real number field with dimension w×h, w and h are the maximum values ​​of the horizontal and vertical coordinates respectively, W represents the pixel fusion weight matrix, m represents the 1 / 3 median of the elements in the Δ matrix, Q is the cloud area screening matrix, and ⊙ represents the Hadamard product operation.

8. The kernel function enhanced satellite image cloud and fog recognition and removal method according to claim 7, characterized in that: In step S4-3, the expression of the smoothed pixel fusion weight matrix is: in, represents the smoothed pixel fusion weight matrix, x represents the first element position in the W matrix, x′ represents the second element position in the W matrix, W(x) and W(x′) represent the values ​​of the W matrix at the x and x′ element positions, respectively, xx′ represents the difference between the first element position x and the second element position x′, C(x) represents the normalization constant at the x element position, Represents the spatial distance as z s The spatial Gaussian weights between the elements when ; Indicates that the element difference is z r The range Gaussian weight between elements when , σ s and σ r They are used to calculate the standard deviation of the spatial Gaussian weight and the range Gaussian weight respectively.

9. The kernel function enhanced satellite image cloud and fog recognition and removal method according to claim 8, characterized in that: In step S4-4, the expression of the pixel-level fusion is: Among them, c represents the color channel, represents the value of the color channel c of the satellite image with the cloud removed, M 1,c Indicates the value of the color channel c of the satellite image to be declouded. Represents the value of the color channel c of the historical image after matching transformation, and ⊙ represents the Hadamard product operation.

Images