Method for repairing missing value of satellite zenith angle in VIIRS remote sensing image

By employing techniques such as image fusion, reclassification, and rotation angle estimation, the missing satellite zenith angle values ​​in VIIRS remote sensing images are accurately repaired, solving the problem of poor repair results in existing technologies and improving the processing quality and versatility of remote sensing data.

CN118982489APending Publication Date: 2024-11-19NANTONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410990726.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

Existing technologies are not effective in repairing missing satellite zenith angle values ​​in VIIRS remote sensing images, especially when there are large-scale missing values ​​or high quality requirements. They also rely on external data and are complex to operate, making them unsuitable for real-time processing and lacking in versatility.

Method used

By collecting boundary vector data of the study area and zenith angle raster data to be repaired, a background grid is generated. Image fusion, reclassification, rotation angle estimation, image rotation, and missing value repair are then performed. Linear regression and image cropping techniques are used to achieve accurate repair of missing values.

Benefits of technology

It significantly reduces errors and uncertainties caused by missing data, improves the quality and application effect of remote sensing data processing, and is versatile, applicable to remote sensing image data processing of VIIRS and other sensors.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982489A_ABST
    Figure CN118982489A_ABST
Patent Text Reader

Abstract

The invention provides a method for repairing a missing value of a satellite zenith angle in a VIIRS remote sensing image, and relates to the technical field of noctilucent remote sensing data processing, the method for repairing the missing value of the satellite zenith angle in the VIIRS remote sensing image provided by the invention utilizes linear regression to carry out data repairing based on linear features of zenith angle data in spatial distribution. Therefore, the method can effectively overcome the problem of dependence on external data, and can effectively deal with the problem of large-scale data missing. By using the method, the missing zenith angle information can be accurately predicted and repaired, and the repair error is remarkably reduced, so that the remote sensing data processing quality and the application effect are improved. The method has practicability and operability, and has wide application prospects in the fields of remote sensing image processing, environment monitoring, meteorological analysis and the like.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of night light remote sensing data processing, and in particular to a method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images. Background Art

[0002] In the field of remote sensing image processing, the satellite zenith angle is one of the important parameters that characterize the satellite's attitude. As a new generation of night-light satellite observers, VIIRS (Visible Infrared Imaging Radiometer Suite) has been widely used to collect radiation data of the Earth's land, atmosphere, ice and ocean in the visible and infrared bands. When using VIIRS remote sensing images for data processing, satellite zenith angle data is a very important geometric parameter. For example, when calculating night-light radiation transmission, accurate zenith angle data is required. However, in the downloaded VIIRS remote sensing images, due to various reasons, the satellite zenith angle information may be partially or widely missing, which is bound to seriously affect the availability of data and the accuracy of analysis results.

[0003] At present, the methods for repairing the missing values ​​of satellite zenith angles in VIIRS remote sensing images mainly focus on the following two aspects:

[0004] A common method is to perform interpolation based on known satellite orbit parameters and time information. This method relies on an accurate understanding of the satellite orbit and data collection period, and attempts to estimate the missing zenith angle values ​​through mathematical models or statistical methods, and then fill them in. However, due to the complexity and variability of remote sensing data, a single interpolation method often fails to obtain satisfactory results in all situations, especially when there are many missing data or high quality requirements.

[0005] Another common method is to use synchronous ground observation data or other remote sensing data to assist in repair. By comparing and correcting the relevant parameters in ground measurement data or other remote sensing images with VIIRS images, attempts are made to infer and repair the missing values ​​of the satellite zenith angle. Although these methods can provide some help in some cases, they rely on the availability and accuracy of external data, are complex to operate, and are not easily applicable to a wide range of remote sensing image processing tasks.

[0006] In view of the limitations of existing technologies, it is crucial to develop an innovative method to repair the missing values ​​of satellite zenith angles in VIIRS remote sensing images. Summary of the invention

[0007] The purpose of the present invention is to solve the technical problems that the method of repairing the missing values ​​of satellite zenith angles in VIIRS remote sensing images in the prior art is not effective in processing a large range of missing zenith angle data, and the method in the prior art needs to refer to more parameter information, is highly dependent on the quality of external data, is not suitable for real-time processing needs, and lacks versatility and automated processing capabilities.

[0008] In order to achieve the above object, the present invention adopts the following technical solutions:

[0009] A method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images comprises the following steps:

[0010] S1: Collect the boundary vector data of the study area and the zenith angle raster data to be restored, and generate the background grid;

[0011] S2: Image Fusion:

[0012] The steps are as follows:

[0013] A1: Define basic functions read_tif(), raster_to_geocoord(), geocoord_to_raster();

[0014] A2: Use the read_tif() function to read the background raster grid data and the zenith angle raster data to be repaired, and obtain their single-band image data, geographic conversion parameters and projection information respectively;

[0015] A3: Use the relevant library functions of GDAL to create output raster files and output raster datasets with the same pixel size, geographic transformation parameters and projection information as the background raster data;

[0016] A4: Traverse the background grid data and update the background grid data pixel values ​​according to the validity of the zenith angle grid data pixel values ​​to be repaired;

[0017] A5: Use the WriteArray() function of gdal to write the updated background raster data into the image fusion dataset to generate the final image fusion file;

[0018] S3: Image reclassification: First, set the effective value interval [0,90], filter the effective values ​​of the image fusion dataset, calculate the CDF, calculate the reclassification group boundary, recalculate the histogram according to the reclassification group boundary, define the reclassification rules, and reclassify according to the reclassification rules;

[0019] S4: rotation angle estimation;

[0020] S5: image rotation;

[0021] S6: missing value repair;

[0022] The following steps are included: get the image size; calculate the regression coefficient of each column of the image; calculate the linear regression value of each row corresponding to each column; traverse the image row by row and fill in the missing values;

[0023] S7: Image rotation: the image after missing value repair is reversely rotated according to avg_slope_angle in S4. The rotation process is the same as S5.

[0024] S8: Image clipping: Set the ArcPy workspace and use ArcPy's Clip_management() function to clip the restored image after image rotation according to the study area boundary vector data, where the invalid value is set to -999.

[0025] Preferably, the specific steps of S4 are as follows:

[0026] First, convert the reclassified_data dataset into a grayscale image and perform Canny edge detection;

[0027] Then, use Hough transform to detect straight lines in the image, denoted as lines, and define the rotation angle array slope;

[0028] Traverse lines, get the two endpoints of each line, calculate the slope based on the two endpoints, use the inverse tangent function to calculate the slope angle corresponding to the slope, and append the slope angle to the array slope;

[0029] Next, remove the minimum and maximum values ​​in the rotation angle array slope, and calculate the average value of the remaining angles, recorded as avg_slope_angle.

[0030] Preferably, the specific steps of S5 are as follows:

[0031] First, the size and rotation center of the image to be rotated (i.e., the image fusion dataset) are obtained, which are recorded as height, width, and center respectively;

[0032] Then, according to the rotation center center and the rotation angle avg_slope_angle, use OpenCV's getRotationMatrix2D() function to calculate the rotation matrix, recorded as rotation_matrix;

[0033] Finally, use OpenCV's warpAffine() function to perform affine transformation on the image to be rotated according to the given parameters rotation_matrix, width, and height to generate the rotated image.

[0034] Preferably, the specific steps of A4 are:

[0035] First, the pixel coordinates of the background raster data are converted to geographic coordinates using the raster_to_geocoord() function.

[0036] Then use the geocoord_to_raster() function to convert it into the pixel coordinates of the zenith angle raster data to be repaired;

[0037] Next, it is determined whether the pixel value of the zenith angle grid data to be repaired corresponding to the pixel coordinate is within the valid interval [0,90];

[0038] If it is within the valid range, the pixel value of the background grid data is updated to the pixel value of the zenith angle grid data to be repaired, otherwise it is set to the specified Nodata value (such as -999.0).

[0039] Preferably, the calculation of CDF in S3 is to calculate the lower quantile and the upper quantile according to the CDF;

[0040] The calculation method of the lower quantile is: QL=100 / m / 100, and the calculation method of the upper quantile is: QU=(100-100 / m) / 100, where m is the number of groups.

[0041] Preferably, the specific step of calculating the reclassification grouping boundary in S3 is to extract the lower quantile boundary and the upper quantile boundary from the interval boundary of the histogram according to QL and QU, respectively, and record them as ql and qu;

[0042] According to the grouping boundaries ql and qu, calculate the reclassification grouping intervals bins, where the interval boundaries of the first and mth groups are ql and qu respectively, and the interval boundary of the i-th group is: int(np.ceil(ql+(qu-ql)*i / (m-2))), where i∈[1,m-1).

[0043] Preferably, the reclassification rules defined in S3 use a dictionary method to define the reclassification value of each group: reclass_rules[(bins[i],bins[i+1])]=int((i+1)*255 / m), where i∈[0,m); then create a two-dimensional array with the same shape and type as the image fusion dataset, fill it with 0 values, and record it as the reclassified dataset reclassified_data;

[0044] Traverse reclass_rules to obtain the grouping interval and classification value of each class; according to the grouping interval that the pixel value in the image fusion dataset falls into, set the corresponding pixel value in reclassified_data as the reclassification value.

[0045] Preferably, the specific steps of calculating the regression coefficients of each column of the image in S6 are as follows:

[0046] First, define the regression coefficient dictionary, denoted as regression_results, and set the column number as the index;

[0047] Then, traverse each column, set the valid value interval to [0,90], extract the pixels corresponding to the valid values ​​in each column, record them as valid_pixels; filter all row numbers corresponding to valid_pixels, record them as valid_rows, let valid_rows be the explanatory variable, and valid_pixels be the dependent variable;

[0048] Use the stats.linregress() function to establish a linear regression equation between the two, obtain the slope p and intercept value q, and add the tuple (p,q) to the list regression_results;

[0049] Finally, calculate the average of all slopes and intercepts in the list regression_results, record them as average_slope and average_intercept, and add them to regression_results, that is: regression_results[-1] = (average_slope, average_intercept).

[0050] Preferably, the specific steps of calculating the linear regression values ​​of each row corresponding to each column in S6 are as follows:

[0051] First, define the list of linear regression values ​​corresponding to each column of the current row, denoted as pre_reg;

[0052] Then, traverse all columns corresponding to the current row, and the column number is recorded as col;

[0053] If the regression coefficient corresponding to the current column exists, then extract the corresponding slope and intercept values ​​according to regression_results[col], otherwise extract the average slope and intercept corresponding to regression_results[-1];

[0054] According to the extracted slope and intercept values, calculate the linear regression prediction value corresponding to the current row number, recorded as pre_reg_value;

[0055] Next, the interval constraint is processed for pre_reg_value, that is, pre_reg_value = max(0, min(90, pre_reg_value)), and pre_reg_value is appended to the list pre_reg;

[0056] Finally, extract the upper quartile value in the list pre_reg as the final predicted value of the current row, recorded as pre_value.

[0057] Preferably, the specific steps of traversing the image row by row and repairing the missing values ​​in S6 are as follows:

[0058] First, set the valid value interval to [0,90], extract all pixels corresponding to valid values ​​in the current row, and record them as valid_pixels_row;

[0059] If the number of valid_pixels_row is 0, then set the repair value of all pixels in the current row to pre_value;

[0060] If the number of valid_pixels_row is not 0, then calculate the average value of valid_pixels_row, recorded as avg_value; perform interval constraint processing on avg_value, that is, avg_value = max(0,min(90,avg_value));

[0061] Then, the difference between the average value and the predicted value is determined: if the absolute tolerance between avg_value and pre_value exceeds tol (set to 3), then the repair value corresponding to the current row is: fix_value = (avg_value + pre_value) / 2, otherwise fix_value = avg_value;

[0062] Finally, the invalid value in the current row is set to the fix value fix_value.

[0063] Compared with the prior art, this application has the following beneficial effects:

[0064] (1) Satellite zenith angle is an important parameter of VIIRS remote sensing image data. If there are large-scale missing zenith angle values ​​in the downloaded VIIRS remote sensing data, it will inevitably seriously affect the integrity of the data and the accuracy of the analysis results. The zenith angle repair scheme proposed by this method can accurately predict and repair the missing zenith angle information, which can significantly reduce the errors and uncertainties that may be caused by missing data.

[0065] (2) This method is not only applicable to VIIRS remote sensing image data, but also has universality and can be extended to the processing of remote sensing image data from other sensors. Zenith angle data usually has significant linear characteristics in spatial distribution. By using this feature, this method can accurately repair missing data, thereby improving the quality and application effect of remote sensing data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 A schematic diagram of a process for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images in one embodiment of the present invention;

[0067] Figure 2 This is an image fusion effect diagram of a method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images in one embodiment of the present invention;

[0068] Figure 3 This is an image reclassification effect diagram of a method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images in one embodiment of the present invention;

[0069] Figure 4 This is an image rotation effect diagram of a method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images in one embodiment of the present invention;

[0070] Figure 5 This is a diagram showing the effect of repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images according to an embodiment of the present invention. DETAILED DESCRIPTION

[0071] The present invention is further described in detail below in conjunction with specific embodiments.

[0072] See also Figure 1 , a method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images, comprising the following steps:

[0073] S1: Generate background grid

[0074] In one embodiment, the S1 comprises the following specific steps:

[0075] First, collect the boundary vector data of the study area and the zenith angle raster data to be repaired, such as Figure 2 As shown in a.

[0076] Then, the attribute information of the zenith angle raster data to be repaired is read using the Describe() function of ArcPy. Based on this attribute information, the pixel size is obtained through the meanCellWidth attribute, recorded as cell_size, and the spatial reference information is obtained through the spatialReference attribute, recorded as geo_ref.

[0077] Next, the feature to raster conversion() function of ArcPy is used to convert the vector data of the study area boundary into the background raster data. In the conversion, the raster size is set to cell_size.

[0078] Finally, the DefineProjection_management() function of ArcPy is used to define the spatial reference geo_ref for the background raster data.

[0079] S2: Image Fusion

[0080] In one embodiment, S2 comprises the following steps:

[0081] A1: Define basic functions read_tif(), raster_to_geocoord(), geocoord_to_raster().

[0082] The read_tif() function is used to read raster data in the specified file format GeoTiff and return a triple containing the dataset, geographic transformation parameters, and projection information.

[0083] The raster_to_geocoord() function is used to convert raster pixel coordinates to geospatial coordinates. It calculates the corresponding longitude and latitude based on the input geographic conversion parameters and pixel row and column coordinates.

[0084] The geocoord_to_raster() function is used to convert geospatial coordinates to raster pixel coordinates. It calculates the corresponding pixel row and column coordinates based on the input geographic conversion parameters and longitude and latitude.

[0085] A2: Use the read_tif() function to read the background raster grid data and the zenith angle raster data to be repaired, and obtain their single-band image data, geographic conversion parameters and projection information respectively.

[0086] A3: Use GDAL's related library functions to create output raster files and output raster datasets with the same pixel size, geographic transformation parameters, and projection information as the background raster data.

[0087] The raster file is the image fusion file, and the corresponding raster dataset is the image fusion dataset.

[0088] A4: Traverse the background grid data and update the background grid data pixel values ​​according to the validity of the zenith angle grid data pixel values ​​to be repaired.

[0089] In this step, the pixel coordinates of the background raster data are first converted to geographic coordinates using the raster_to_geocoord() function.

[0090] Then use the geocoord_to_raster() function to convert it into the pixel coordinates of the zenith angle raster data to be repaired.

[0091] Next, it is determined whether the pixel value of the zenith angle grid data to be repaired corresponding to the pixel coordinate is within the valid interval [0,90].

[0092] If it is within the valid range, the pixel value of the background grid data is updated to the pixel value of the zenith angle grid data to be repaired, otherwise it is set to the specified Nodata value (such as -999.0).

[0093] A5: Use the WriteArray() function of gdal to write the updated background grid data into the image fusion dataset to generate the final image fusion file. The generated image fusion effect is as follows: Figure 2 As shown in b.

[0094] S3: Image Reclassification

[0095] In one embodiment, the specific steps of S3 are as follows:

[0096] B1: Set the valid value interval [0,90] to filter the valid values ​​of the image fusion dataset.

[0097] B2: Calculate the corresponding histogram and cumulative distribution function (CDF).

[0098] Specifically, the lower quantile and the upper quantile are calculated based on the CDF.

[0099] The calculation method of the lower quantile is: QL=100 / m / 100, and the calculation method of the upper quantile is: QU=(100-100 / m) / 100, where m is the number of groups.

[0100] B3: Calculate the reclassification group boundaries.

[0101] Specifically, in one embodiment, according to QL and QU, the lower quantile boundary and the upper quantile boundary are extracted from the interval boundaries of the histogram, respectively, and are denoted as ql and qu.

[0102] According to the grouping boundaries ql and qu, calculate the reclassification grouping intervals bins. Among them, the interval boundaries of the first group and the mth group are ql and qu respectively, and the interval boundary of the i-th group is: int(np.ceil(ql+(qu-ql)*i / (m-2))), where i∈[1,m-1).

[0103] B4: Recalculate the histogram based on the reclassification group boundaries. The histogram effect is as follows: Figure 3 As shown in a.

[0104] B5: Define reclassification rules.

[0105] Specifically, in one embodiment, the reclassification value of each group is defined in a dictionary manner as follows: reclass_rules[(bins[i], bins[i+1])] = int((i+1)*255 / m), where i∈[0,m).

[0106] B6: Reclassify according to the reclassification rules.

[0107] Create a two-dimensional array with the same shape and type as the image fusion dataset, fill it with 0 values, and record it as the reclassified dataset reclassified_data. Traverse reclass_rules to obtain the grouping interval and classification value of each class. According to the grouping interval that the pixel value in the image fusion dataset falls into, set the corresponding pixel value in reclassified_data as the reclassified value. The final reclassified image effect is as follows Figure 3 As shown in b.

[0108] S4: Rotation angle estimation

[0109] First, convert the reclassified_data dataset into a grayscale image and perform Canny edge detection.

[0110] Then, Hough transform is used to detect straight lines in the image, denoted as lines. The detected straight lines are as follows: Figure 4 a. Define the rotation angle array slope. Traverse lines, get the two endpoints of each line, calculate the slope based on the two endpoints, and use the inverse tangent function to calculate the slope angle corresponding to the slope. Append the slope angle to the array slope.

[0111] Next, remove the minimum and maximum values ​​in the rotation angle array slope, and calculate the average value of the remaining angles, recorded as avg_slope_angle.

[0112] S5: Image rotation

[0113] First, the size and rotation center of the image to be rotated (i.e., the image fusion dataset) are obtained, which are recorded as height, width, and center respectively.

[0114] Then, according to the rotation center center and the rotation angle avg_slope_angle, the rotation matrix is ​​calculated using the getRotationMatrix2D() function of OpenCV, recorded as rotation_matrix.

[0115] Finally, use OpenCV's warpAffine() function to perform affine transformation on the image to be rotated according to the given parameters rotation_matrix, width, and height to generate the rotated image.

[0116] The effect of rotating the image by avg_slope_angle is as follows Figure 4 As shown in b.

[0117] S6: Missing value repair

[0118] In one embodiment, the method specifically comprises the following steps:

[0119] (1) Get the image size.

[0120] Get the number of rows and columns of the rotated image, recorded as rows and cols respectively.

[0121] (2) Calculate the regression coefficient of each column of the image.

[0122] First, define the regression coefficient dictionary, denoted as regression_results, and set the column number as the index.

[0123] Then, traverse each column, set the valid value interval to [0,90], extract the pixels corresponding to the valid values ​​in each column, and record them as valid_pixels. Filter all row numbers corresponding to valid_pixels, and record them as valid_rows. Let valid_rows be the explanatory variable and valid_pixels be the dependent variable. Use the stats.linregress() function to establish a linear regression equation between the two, and get the slope p and intercept value q. Add the tuple (p,q) to the list regression_results.

[0124] Finally, calculate the average of all slopes and intercepts in the list regression_results, record them as average_slope and average_intercept, and add them to regression_results, that is: regression_results[-1] = (average_slope, average_intercept).

[0125] (3) Calculate the linear regression value of each row corresponding to each column.

[0126] First, define the list of linear regression values ​​corresponding to each column of the current row, denoted as pre_reg.

[0127] Then, traverse all columns corresponding to the current row, and record the column number as col. If the regression coefficient corresponding to the current column exists, extract the corresponding slope and intercept values ​​according to regression_results[col]. Otherwise, extract the average slope and intercept corresponding to regression_results[-1]. Based on the extracted slope and intercept values, calculate the linear regression prediction value corresponding to the current row number, recorded as pre_reg_value.

[0128] Next, the interval constraint is applied to pre_reg_value, that is, pre_reg_value = max(0, min(90, pre_reg_value)). Append pre_reg_value to the list pre_reg.

[0129] Finally, extract the upper quartile value in the list pre_reg as the final predicted value of the current row, recorded as pre_value.

[0130] (4) Traverse the image row by row and fill in missing values.

[0131] First, set the valid value interval to [0,90], extract all pixels corresponding to valid values ​​in the current row, and record them as valid_pixels_row. If the number of valid_pixels_row is 0, then set the repair value of all pixels in the current row to pre_value.

[0132] If the number of valid_pixels_row is not 0, then calculate the average value of valid_pixels_row and record it as avg_value. Perform interval constraint processing on avg_value, that is, avg_value = max(0, min(90, avg_value)).

[0133] Then, the difference between the average value and the predicted value is determined. If the absolute tolerance between avg_value and pre_value exceeds tol (set to 3), the repair value corresponding to the current row is: fix_value = (avg_value + pre_value) / 2, otherwise fix_value = avg_value.

[0134] Finally, the invalid value in the current row is set to the fix value fix_value.

[0135] S7: Image rotation

[0136] After the missing values ​​are repaired, the image is reversely rotated according to the angle -avg_slope_angle in S4. The rotation process is the same as S5.

[0137] The final image after the missing values ​​are restored is as follows: Figure 5 As shown in a.

[0138] S8: Image cropping

[0139] Specifically, in one implementation, the ArcPy workspace is first set.

[0140] Then, ArcPy's Clip_management() function is used to clip the restored image after image rotation according to the boundary vector data of the study area, where the invalid value is set to -999. The final generated missing value repair image effect is as follows Figure 5 As shown in b.

[0141] After the above steps, this method can be used to effectively repair the missing data of satellite zenith angle in VIIRS remote sensing images.

[0142] This application proposes a method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images, which can accurately predict and repair missing zenith angle information, and can significantly reduce the errors and uncertainties that may be caused by missing data. Secondly, the repair method provided by this application is not only applicable to VIIRS remote sensing image data, but also has universality and can be extended to the remote sensing image data processing of other sensors. Zenith angle data usually has significant linear characteristics in spatial distribution. Using this feature, this method can accurately repair missing data, thereby improving the quality and application effect of remote sensing data processing.

Claims

1. A method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images, characterized in that: The following steps are involved: S1: Collect the boundary vector data of the study area and the zenith angle raster data to be restored, and generate the background grid; S2: Image Fusion: The steps are as follows: A1: Define basic functions read_tif(), raster_to_geocoord(), geocoord_to_raster(); A2: Use the read_tif() function to read the background raster grid data and the zenith angle raster data to be repaired, and obtain their single-band image data, geographic conversion parameters and projection information respectively; A3: Use the relevant library functions of GDAL to create output raster files and output raster datasets with the same pixel size, geographic transformation parameters and projection information as the background raster data; A4: Traverse the background grid data and update the background grid data pixel values ​​according to the validity of the zenith angle grid data pixel values ​​to be repaired; A5: Use the WriteArray() function of gdal to write the updated background raster data into the image fusion dataset to generate the final image fusion file; S3: Image reclassification: First, set the effective value interval [0,90], filter the effective values ​​of the image fusion dataset, calculate the CDF, calculate the reclassification group boundary, recalculate the histogram according to the reclassification group boundary, define the reclassification rules, and reclassify according to the reclassification rules; S4: rotation angle estimation; S5: Image rotation; S6: Missing value repair; The following steps are included: get the image size; calculate the regression coefficient of each column of the image; calculate the linear regression value of each row corresponding to each column; traverse the image row by row and fill in the missing values; S7: Image rotation: the image after missing value repair is reversely rotated according to the avg_slope_angle in S4. The rotation process is the same as S5. S8: Image clipping: Set the ArcPy workspace and use ArcPy's Clip_management() function to clip the restored image after image rotation according to the study area boundary vector data, where the invalid value is set to -999.

2. A method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images according to claim 1, characterized in that: The specific steps of S4 are as follows: First, convert the reclassified_data dataset into a grayscale image and perform Canny edge detection; Then, use Hough transform to detect straight lines in the image, denoted as lines, and define the rotation angle array slope; Traverse lines, get the two endpoints of each line, calculate the slope based on the two endpoints, use the inverse tangent function to calculate the slope angle corresponding to the slope, and append the slope angle to the array slope; Next, remove the minimum and maximum values ​​in the rotation angle array slope, and calculate the average value of the remaining angles, recorded as avg_slope_angle.

3. A method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images according to claim 2, characterized in that: The specific steps of S5 are as follows: First, the size and rotation center of the image to be rotated (i.e., the image fusion dataset) are obtained, which are recorded as height, width, and center respectively; Then, according to the rotation center center and the rotation angle avg_slope_angle, use OpenCV's getRotationMatrix2D() function to calculate the rotation matrix, recorded as rotation_matrix; Finally, use OpenCV's warpAffine() function to perform affine transformation on the image to be rotated according to the given parameters rotation_matrix, width, and height to generate the rotated image.

4. The method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images according to claim 1, characterized in that: The specific steps of A4 are: First, the pixel coordinates of the background raster data are converted to geographic coordinates using the raster_to_geocoord() function. Then use the geocoord_to_raster() function to convert it into the pixel coordinates of the zenith angle raster data to be repaired; Next, it is determined whether the pixel value of the zenith angle grid data to be repaired corresponding to the pixel coordinate is within the valid interval [0,90]; If it is within the valid range, the pixel value of the background grid data is updated to the pixel value of the zenith angle grid data to be repaired, otherwise it is set to the specified Nodata value (such as -999.0).

5. The method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images according to claim 1, characterized in that: The calculation of CDF in S3 is to calculate the lower quantile and the upper quantile according to the CDF; The calculation method of the lower quantile is: QL=100 / m / 100, and the calculation method of the upper quantile is: QU=(100-100 / m) / 100, where m is the number of groups.

6. A method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images according to claim 5, characterized in that: The specific steps of calculating the reclassification grouping boundary in S3 are to extract the lower quantile boundary and the upper quantile boundary from the interval boundary of the histogram according to QL and QU, respectively, and record them as ql and qu; According to the grouping boundaries ql and qu, calculate the reclassification grouping intervals bins, where the interval boundaries of the first and mth groups are ql and qu respectively, and the interval boundary of the i-th group is: int(np.ceil(ql+(qu-ql)*i / (m-2))), where i∈[1,m-1).

7. A method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images according to claim 6, characterized in that: The reclassification rules defined in S3 use a dictionary to define the reclassification value of each group: reclass_rules[(bins[i],bins[i+1])]=int((i+1)*255 / m), where i∈[0,m); then create a two-dimensional array with the same shape and type as the image fusion dataset, fill it with 0 values, and record it as the reclassified dataset reclassified_data; Traverse reclass_rules to obtain the grouping interval and classification value of each class; according to the grouping interval that the pixel value in the image fusion dataset falls into, set the corresponding pixel value in reclassified_data as the reclassification value.

8. The method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images according to claim 1, characterized in that: The specific steps of calculating the regression coefficients of each column of the image in S6 are as follows: First, define the regression coefficient dictionary, denoted as regression_results, and set the column number as the index; Then, traverse each column, set the valid value interval to [0,90], extract the pixels corresponding to the valid values ​​in each column, record them as valid_pixels; filter all row numbers corresponding to valid_pixels, record them as valid_rows, let valid_rows be the explanatory variable, and valid_pixels be the dependent variable; Use the stats.linregress() function to establish a linear regression equation between the two, obtain the slope p and intercept value q, and add the tuple (p,q) to the list regression_results; Finally, calculate the average of all slopes and intercepts in the list regression_results, record them as average_slope and average_intercept, and add them to regression_results, that is: regression_results[-1] = (average_slope, average_intercept).

9. A method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images according to claim 8, characterized in that: The specific steps of calculating the linear regression values ​​of each row corresponding to each column in S6 are as follows: First, define the list of linear regression values ​​corresponding to each column of the current row, denoted as pre_reg; Then, traverse all columns corresponding to the current row, and the column number is recorded as col; If the regression coefficient corresponding to the current column exists, then extract the corresponding slope and intercept values ​​according to regression_results[col], otherwise extract the average slope and intercept corresponding to regression_results[-1]; According to the extracted slope and intercept values, calculate the linear regression prediction value corresponding to the current row number, recorded as pre_reg_value; Next, the interval constraint is processed for pre_reg_value, that is, pre_reg_value = max(0, min(90, pre_reg_value)), and pre_reg_value is appended to the list pre_reg; Finally, extract the upper quartile value in the list pre_reg as the final predicted value of the current row, recorded as pre_value.

10. A method for repairing missing values ​​of satellite zenith angles in VIIRS remote sensing images according to claim 9, characterized in that: The specific steps of traversing the image row by row and filling missing values ​​in S6 are as follows: First, set the valid value interval to [0,90], extract all pixels corresponding to valid values ​​in the current row, and record them as valid_pixels_row; If the number of valid_pixels_row is 0, then set the repair value of all pixels in the current row to pre_value; If the number of valid_pixels_row is not 0, then calculate the average value of valid_pixels_row, recorded as avg_value; perform interval constraint processing on avg_value, that is, avg_value = max(0,min(90,avg_value)); Then, the difference between the average value and the predicted value is determined: if the absolute tolerance between avg_value and pre_value exceeds tol (set to 3), then the repair value corresponding to the current row is: fix_value = (avg_value + pre_value) / 2, otherwise fix_value = avg_value; Finally, the invalid value in the current row is set to the fix value fix_value.