FY-3D medium resolution imager image clear sky synthesis method and system

Abstract

This invention discloses a method and system for clear sky synthesis of FY-3D medium-resolution imager images. First, FY-3D MERSI time-series image data is preprocessed, and a time-series clear sky / cloud mask is obtained through fully automated cloud detection based on time-series analysis. Then, based on object-oriented principles, the clear sky / cloud mask is segmented to obtain clear sky and cloud mask objects. The cloud mask objects are then synthesized with FY-3D MERSI time-series data that has undergone radiometric consistency processing to create a clear sky image. Finally, the synthesized clear sky image undergoes post-processing including feathering, color enhancement, and texture enhancement to obtain the final clear sky image. This invention enables fully automated clear sky synthesis of images and boasts high parallel processing efficiency.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing image processing technology, specifically relating to a method and system for synthesizing clear sky images from an FY-3D medium resolution imager. Background Technology

[0002] The Fengyun-3D (FY-3D) satellite was successfully launched from the Taiyuan Satellite Launch Center on November 15, 2017, at 2:35 AM by a Long March 4C carrier rocket. It is a polar-orbiting meteorological satellite providing global observation. Its second-generation Medium Resolution Imager (MERSI) boasts advantages such as high observation frequency and wide coverage, enabling the acquisition of continuous satellite data over large areas and even globally. However, image acquisition is highly susceptible to weather conditions, significantly impacting the quality of ground feature information and reducing data utilization. Cloud cover is one such factor. The presence of clouds severely affects the interpretation of remote sensing images; separating clouds from clear skies is a necessary preprocessing step for retrieving various atmospheric and surface parameters.

[0003] The accuracy of cloud detection results directly affects the inversion results of other parameters. Furthermore, for meteorologists, studying cloud distribution can reveal extreme weather phenomena and their patterns of change. Therefore, an accurate and reliable cloud detection method is extremely valuable. Since this satellite possesses a large amount of time-series data, providing data support for analyzing the reflectivity patterns of cloud pixels and background pixels, this invention proposes a cloud detection method based on time-series analysis.

[0004] Over 66% of the Earth's surface is frequently covered by clouds. Cloud cover attenuates or even destroys the ground feature information we obtain. Therefore, automatically acquiring large-scale clear-sky images, eliminating the influence of clouds, is crucial for subsequent image-based Earth resource surveys, natural disaster prediction, and environmental pollution monitoring. To achieve fully automated clear-sky image acquisition, this invention fully utilizes the temporal characteristics of cloud and clear-sky pixels, proposing a clear-sky image generation method based on object-oriented segmentation and temporal analysis. Summary of the Invention

[0005] This invention addresses the shortcomings of existing technologies by providing a method for synthesizing clear-sky images from an FY-3D medium-resolution imager, comprising the following steps:

[0006] Step 1: Perform atmospheric correction, geometric correction, geometric reprojection, and image stitching preprocessing on the FY-3D MERSI time-series image data;

[0007] Step 2: Perform fully automated cloud detection based on time-series analysis on the preprocessed FY-3D MERSI time-series image to obtain a time-series clear sky / cloud mask;

[0008] Step 3: Perform radiometric consistency processing on the preprocessed FY-3D MERSI time series data;

[0009] Step 4: Perform object-oriented segmentation on the time-series clear sky / cloud mask obtained in Step 2 to obtain clear sky and cloud mask objects;

[0010] Step 5: Based on the optimal strategy, synthesize the clear sky map by combining the cloud mask objects obtained in Step 4 with the FY-3D MERSI time series data processed by radiometric consistency in Step 3.

[0011] Step 6: Perform feathering, color enhancement, and texture enhancement post-processing on the synthesized clear sky image to obtain the final clear sky image.

[0012] Furthermore, in step 1, atmospheric correction is performed on the FY-3D MERSI time-series image data based on NASA's CREFL method to obtain the true surface reflectance in the visible light band. First, the original satellite data, including observation data and geolocation data, is read in. The 1km geolocation data is resampled to 250m to match the spatial resolution of the Earth observation data. The DN values ​​of channels 24 and 25 (red, green, blue, near-infrared, and thermal infrared) from the L1 observation data, along with auxiliary data from the geolocation file, solar and satellite observation angles, and altitude information, are input as parameters for atmospheric correction. Atmospheric correction is performed using the CREFL method based on 6S radiative transfer. Based on the relationship between apparent atmospheric reflectance, path reflectance, and true surface reflectance, the true surface reflectance in the visible light band is calculated. The specific calculation method is as follows:

[0013]

[0014] In the formula, Atmospheric apparent reflectance is calculated using the original DN value; T(μ) represents path reflectivity; s ) and T(μ v ) represents the atmospheric transmittance related to the observation geometry; μ s The solar zenith angle; μ v The zenith angle of the satellite; Where S is the relative azimuth angle; S is the atmospheric albedo. Let be the surface reflectance to be determined.

[0015] Geometric correction was performed using the GLT (Geographic Lookup Table) method. This involved constructing a geographic lookup table using longitude and latitude files from a 1km observation geolocation information file. This table was then used to perform geometric correction on the atmospherically corrected FY-3D MERSI time-series image data. The actual geographic location of a given initial pixel could be retrieved from the lookup table. After geometric correction, the FY-3D MERSI time-series image data underwent geometric reprojection. During reprojection, a nadir distance file was simultaneously generated. When stitching together the atmospherically and geometrically corrected FY-3D MERSI time-series images, if multiple images covered the target area, candidate images with nadir or near-nadir points were prioritized.

[0016] Furthermore, in step 2, for a given pixel, the temporal surface reflectance data obtained in step 1 are sorted from smallest to largest [(X1,T1),(X2,T2),..,(X...]. n ,T n )], X n T represents the surface reflectance. n This represents the time corresponding to the surface reflectance. The first derivative of each time series point is approximated using the backward difference method on the sorted time-series surface reflectance data, i.e.:

[0017]

[0018] In the formula, Let X[i+1] and X[i] represent the reflectivity at the (i+1)th and (i)th time points, respectively, and let T[i+1] and T[i] represent the time corresponding to the (i+1)th and (i)th time points, respectively.

[0019] For the derivative of the last time point of the sorted temporal surface reflectance data, choose the endpoint (X). n ,T n ) and the previous neighboring point (X) n-1 ,T n-1 Substitute into formula (2) for calculation.

[0020] Using pattern matching, key points are identified in the temporal surface reflectance curves of pixels, the temporal surface reflectance curves sorted from smallest to largest, and the sorted temporal surface reflectance curves after first-order differentiation. Key points include local minima and local maxima. Local minima indicate the transition from flat areas to gradually rising areas, while local maxima indicate the transition from gradually rising areas to highly flat areas. Flat areas are relatively low-valued, flat segments in the sorted temporal surface reflectance curves after first-order differentiation, with derivative values ​​close to zero, corresponding to clear skies or cloud shadows. Gradually rising areas are segments where the values ​​gradually increase, with derivative values ​​gradually increasing, corresponding to the transition from clear skies to cloud areas. Highly flat areas are relatively high-valued, flat segments in the sorted temporal surface reflectance curves after first-order differentiation, with derivative values ​​close to 0, corresponding to cloud areas.

[0021] By using key points, the sorted temporal surface reflectance curves after first-order differentiation are divided into flat areas, gently rising areas, and highly flat areas, thus distinguishing between clear skies, cloud shadows, and cloud areas. Since some shadows with lower reflectance exist within the flat areas, the critical value (L) at the boundary between the two sides of the clear sky area is used. min ,L max Move closer to the center, discarding the critical regions on both sides, that is, take [L]. min +0.1(L max -L min ),L max -0.1(L max -L min The time segment within the range is considered a clear sky area to ensure the accuracy of the clear sky area division.

[0022] Furthermore, in step 5, the cloud mask object obtained in step 4 is combined with the FY-3D MERSI time-series image processed for radiometric consistency in step 3 to create a clear sky image. To achieve higher radiometric consistency in the composite result, clear sky objects are selected according to an optimal strategy: ① For adjacent cloud mask objects, clear sky objects from the same or similar dates are selected for filling; ② If ① is satisfied, the clear sky object with the smallest satellite zenith angle and solar zenith angle is selected; ③ If ① and ② are satisfied, the clear sky object with the best image quality is selected. Following the above optimal strategy, clear sky objects are sequentially selected from other time-series images to fill the cloud mask objects, and a composite mask is generated. The value of each pixel in the composite mask represents which image the spectrum of that pixel in the composite clear sky image comes from. Finally, the composite mask is used to fill and generate the clear sky image.

[0023] Furthermore, in step 6, the cloud mask object from step 4 is used to post-process the initial clear sky composite result, outputting the final clear sky image. Post-processing includes: ① Feathering transition processing, using the cloud mask object as auxiliary information to perform a feathering transition along the cloud object boundary, eliminating stitching marks; ② Color enhancement processing, using infrared band information to compensate the green channel, making the vegetation's color tone more natural; ③ Texture enhancement processing, using the DEM to perform texture enhancement processing on the surface reflectance image, extracting texture features from the clear sky image using the gray-level co-occurrence matrix, fitting the texture features as the elevation variable of the DEM to generate a new texture-enhanced DEM, and then applying a local contrast enhancement algorithm to obtain the enhanced image.

[0024] The present invention also provides an FY-3D medium resolution image clear sky synthesis system, used to implement the FY-3D medium resolution image clear sky synthesis method described above.

[0025] Furthermore, it includes a processor and a memory, the memory being used to store program instructions, and the processor being used to call the program instructions in the memory to execute the above-described FY-3D medium resolution image clear sky synthesis method.

[0026] Alternatively, it may include a readable storage medium storing a computer program that, when executed, implements a clear-sky image synthesis method for an FY-3D medium-resolution imager as described above.

[0027] Compared with the prior art, the present invention has the following advantages:

[0028] 1) A fully automatic cloud detection method based on time-series analysis is proposed. By analyzing the surface reflectance characteristics of Fengyun-3 images, a time-series reflectance curve for each pixel is established. Key points are solved using pattern matching. The sorted time-series surface reflectance curves after first-order differentiation are divided into flat areas, gently rising areas, and highly flat areas based on the key points, thereby distinguishing between clear skies, cloud shadows, and cloud areas. This method does not require sample training or threshold selection, and can achieve fully automatic processing with high parallel processing efficiency.

[0029] 2) An object-oriented method for synthesizing clear sky images is proposed. Clear sky and cloud mask objects are obtained through object-oriented segmentation. Based on the optimal strategy, the optimal clear sky object is selected to fill the cloud mask object. The clear sky synthesis process considers the temporal and spatial neighborhood characteristics, reduces speckle noise, and ensures the optimal radiometric consistency of the result. The method is fully automated and directly outputs the synthesized image without manual intervention. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0031] Figure 1 This is a flowchart of the clear sky synthesis method according to an embodiment of the present invention.

[0032] Figure 2 This is a flowchart of the raw data preprocessing process according to an embodiment of the present invention.

[0033] Figure 3 A comparison of the spectral reflectance characteristics of clouds and shadows versus clear sky features.

[0034] Figure 4 This is a temporal surface reflectance sorting transformation and curve matching differentiation diagram of a certain pixel point in an embodiment of the present invention.

[0035] Figure 5 The image shows the result obtained by the fully automatic cloud detection method based on time-series analysis proposed in this invention. The left side is the original image, and the right side is a binary image of cloud and clear sky, where the white area represents clouds and the black area represents clear sky.

[0036] Figure 6 This is a diagram showing the result of multi-scale segmentation of the clear sky / cloud mask according to an embodiment of the present invention. The left side is the time-series clear sky / cloud mask, and the right side is the segmented clear sky and cloud mask objects.

[0037] Figure 7 This is a flowchart of a sky synthesis process based on pairwise segmentation, according to an embodiment of the present invention.

[0038] Figure 8 This is a flowchart illustrating the post-processing of the synthesized clear sky image according to an embodiment of the present invention.

[0039] Figure 9 shows the effect of the clear sky composite method proposed in this invention. Figure 9(a) is the original MERSI-II image of the Belt and Road region on May 12, 2019, and Figure 9(b) is the clear sky composite result of the Belt and Road region in May 2019 obtained by using the method proposed in this invention.

[0040] Figure 10 This is a composite image of clear skies over parts of Europe in May 2019, obtained using this invention.

[0041] Figure 11 This is a composite image of clear skies over parts of Africa in May 2019, obtained using this invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0043] This invention provides a method and system for synthesizing clear sky images from an FY-3D medium-resolution imager. The technical solution of this invention will be further described below with reference to the accompanying drawings.

[0044] Example 1

[0045] like Figure 1 As shown, this embodiment of the invention provides a method for synthesizing clear-sky images from an FY-3D medium-resolution imager, comprising the following steps:

[0046] Step 1: Perform atmospheric correction, geometric correction, geometric reprojection, and image stitching preprocessing on the FY-3D MERSI time-series image data.

[0047] Atmospheric correction was performed on FY-3DMERSI time-series image data based on NASA's CREFL (Corrected Reflectance) method to obtain the true surface reflectance in the visible light band. The preprocessing workflow is as follows: Figure 2 As shown. First, the raw satellite data, including observation data and geolocation data, is read in. The 1km geolocation data is resampled to 250m to match the spatial resolution of the Earth observation data. The DN values ​​of channels 24 and 25 (red, green, blue, near-infrared, and thermal infrared) from the L1 observation data are input, along with auxiliary data from the geolocation file, solar and satellite observation angles, and altitude information, as input parameters for atmospheric correction. Atmospheric correction is performed using the CREFL method based on 6S radiative transfer. Based on the relationship between apparent atmospheric reflectance, path reflectance, and true surface reflectance, the true surface reflectance in the visible light band is calculated. The specific calculation method is as follows:

[0048]

[0049] In the formula, Atmospheric apparent reflectance is calculated using the original DN value; T(μ) represents path reflectivity; s ) and T(μ v ) represents the atmospheric transmittance related to the observation geometry; μ s The solar zenith angle; μ v The zenith angle of the satellite; Where S is the relative azimuth angle; S is the atmospheric albedo. Let be the surface reflectance to be determined.

[0050] Geometric correction was performed using the GLT (Geographic Lookup Table) method. This involved constructing a geographic lookup table using longitude and latitude data from a 1km observation geolocation information file. This table was then used to perform geometric correction on the atmospherically corrected FY-3D MERSI time-series image data. The actual geographic location of a given initial pixel could be retrieved from the lookup table. After geometric correction, the FY-3D MERSI time-series image data underwent geometric reprojection. During reprojection, a nadir distance file was simultaneously generated. When stitching together the atmospherically and geometrically corrected FY-3D MERSI time-series images, if multiple images covered the target area, candidate images with nadir or near-nadir points were prioritized. Because the overlapping areas used pixels closest to the nadir point, the quality of the stitched image was optimized.

[0051] Step 2: Perform fully automated cloud detection based on time-series analysis on the preprocessed FY-3D MERSI time-series image to obtain a time-series clear sky / cloud mask.

[0052] Cloud pixels exhibit reflectance characteristics in the visible, near-infrared, and thermal infrared spectra that are distinctly different from those of clear-sky objects such as vegetation, soil, snow, and water. For example... Figure 3 As shown, clouds have high reflectivity in the visible light range, low reflectivity in their shadows, and low radiative brightness temperature in the thermal infrared range. For a given pixel, the temporal surface reflectivity data obtained in step 1 are sorted from smallest to largest [(X1,T1),(X2,T2),..,(X...]. n ,T n )], X n T represents the surface reflectance. n This represents the time corresponding to the surface reflectance. The first derivative of each time series point is approximated using the backward difference method on the sorted time-series surface reflectance data, i.e.:

[0053]

[0054] In the formula, Let X[i+1] and X[i] represent the reflectivity at the (i+1)th and (i)th time points, respectively, and let T[i+1] and T[i] represent the time corresponding to the (i+1)th and (i)th time points, respectively.

[0055] For the derivative of the last time point of the sorted temporal surface reflectance data, choose the endpoint (X). n ,T n ) and the previous neighboring point (X) n-1 ,T n-1 Substitute into formula (2) for calculation.

[0056] like Figure 4 As shown, after sorting the temporal surface reflectance data of a certain pixel, if the temporal image contains shadows, clear skies, and clouds, then the time series points belonging to shadows will be at the forefront of the curve, with lower reflectance. After one rise to a slightly higher reflectance, the relatively flat time series corresponds to clear skies. Then, a sharp rise to a very high reflectance, followed by a relatively flat time series, corresponds to clouds. Therefore, the pattern of pixel reflectance variation makes it easy to distinguish between clear skies, clouds, and cloud shadows, thus allowing for the synthesis of a clear skies image.

[0057] Using pattern matching, key points are identified from the temporal surface reflectance curves of pixels, the temporal surface reflectance curves sorted from smallest to largest, and the sorted temporal surface reflectance curves after first-order differentiation. Key points include local minima and local maxima. Local minima represent the transition points from flat regions (flat segments with relatively low values ​​in the sorted temporal surface reflectance curves after first-order differentiation, where the derivative value is close to zero, typically corresponding to clear skies or cloud shadows) to gradually rising regions (segments with gradually increasing values ​​in the sorted temporal surface reflectance curves after first-order differentiation, where the derivative value gradually increases, typically corresponding to the transition from clear skies to clouds). Local maxima represent the transition points from gradually rising regions (segments with gradually increasing values ​​in the sorted temporal surface reflectance curves after first-order differentiation, where the derivative value gradually increases, typically corresponding to the transition from clear skies to clouds) to highly flat regions (flat segments with relatively high values ​​in the sorted temporal surface reflectance curves after first-order differentiation, where the derivative value is close to 0, typically corresponding to clouds).

[0058] By using key points, the sorted temporal surface reflectance curves after first-order differentiation are divided into flat regions, gently rising regions, and highly flat regions, thus distinguishing between clear skies, cloud shadows, and cloud areas. Since some shadows with lower reflectance exist within the flat regions, [the text abruptly ends here]. Figure 4 The critical value of the boundary between the clear sky areas on both sides (L) min ,L max Move closer to the center, discarding the critical regions on both sides, that is, take [L]. min +0.1(L max -L min ),L max -0.1(L max -L min The time series within the range of )] is designated as a clear-sky region to ensure the accuracy of clear-sky region delineation. The final time series clear-sky / cloud mask is as follows: Figure 5 As shown.

[0059] Step 3: Perform radiometric consistency processing on the preprocessed FY-3D MERSI time series data.

[0060] Step 4: Perform object-oriented segmentation on the time-series clear sky / cloud mask obtained in Step 2 to obtain clear sky and cloud mask objects.

[0061] like Figure 6 As shown, the left side is the obtained temporal clear sky / cloud mask, and the right side is the clear sky and cloud mask objects obtained after object-oriented segmentation of the temporal clear sky / cloud mask.

[0062] Step 5: Based on the optimal strategy, synthesize the clear sky map by combining the cloud mask object obtained in Step 4 with the FY-3D MERSI time series data processed by radiometric consistency in Step 3.

[0063] The cloud mask objects obtained in step 4 are combined with the FY-3D MERSI time-series image processed for radiometric consistency in step 3 to create a clear sky image. To improve the radiometric consistency of the composite result, clear sky objects are selected according to the optimal strategy: ① For adjacent cloud mask objects, clear sky objects from the same or similar dates are selected for filling; ② If ① is satisfied, the clear sky object with the smallest satellite zenith angle and solar zenith angle is selected; ③ If ① and ② are satisfied, the clear sky object with the best image quality, such as higher sharpness and contrast, and higher information entropy, is selected. Figure 7 As shown, following the optimal strategy described above, clear sky objects are sequentially selected from other time-series images to fill cloud mask objects, and a composite mask is generated. The value of each pixel in the composite mask represents which image the spectrum of that pixel in the composite clear sky image comes from. Finally, the composite mask is used to fill and generate a clear sky image.

[0064] Step 6: Perform feathering, color enhancement, and texture enhancement post-processing on the synthesized clear sky image to obtain the final clear sky image.

[0065] like Figure 8 As shown, the cloud mask object from step 4 is used to post-process the initial clear sky composite result, outputting the final clear sky image. Post-processing includes: ① Feathering transition processing: using the cloud mask object as auxiliary information, feathering transitions are performed along the cloud object boundaries to eliminate stitching artifacts; ② Color enhancement processing: infrared band information is used to compensate the green channel, making the tones of vegetation and other special features more natural; ③ Texture enhancement processing: the surface reflectance image is enhanced using the DEM, and the texture features of the clear sky image are extracted using the gray-level co-occurrence matrix. These texture features are then fitted as elevation variables of the DEM to generate a new texture-enhanced DEM. Finally, the Contrast Limited Adaptive Histogram Equalization (CLAHE) algorithm is applied to obtain the enhanced image.

[0066] Figures 9-11 show the composite results of clear skies in three global regions. The monthly clear skies composite results are obtained by preprocessing, cloud detection, clear skies composite, and subsequent rendering of images from one month (one scene per day). Taking the clear skies composite results of the Belt and Road region in May 2019 as an example, Figure 9(a) shows the original MERSI-II image of the Belt and Road region on May 12, 2019, and Figure 9(b) shows the monthly clear skies result for May obtained using the method proposed in this invention.

[0067] Example 2

[0068] Based on the same inventive concept, the present invention also provides an FY-3D medium resolution image clear sky synthesis system, including a processor and a memory. The memory is used to store program instructions, and the processor is used to call the program instructions in the memory to execute the FY-3D medium resolution image clear sky synthesis method as described above.

[0069] Example 3

[0070] Based on the same inventive concept, the present invention also provides an FY-3D medium resolution image clear sky synthesis system, including a readable storage medium storing a computer program, which, when executed, implements the FY-3D medium resolution image clear sky synthesis method as described above.

[0071] In specific implementation, the method proposed in the technical solution of this invention can be automatically executed by those skilled in the art using computer software technology. System devices for implementing the method, such as computer-readable storage media storing the corresponding computer program of the technical solution of this invention and computer equipment including the computer program running the corresponding computer program, should also be within the protection scope of this invention.

[0072] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to replace them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for synthesizing clear-sky images from an FY-3D medium-resolution imager, characterized in that, Includes the following steps: Step 1: Perform atmospheric correction, geometric correction, geometric reprojection, and image stitching preprocessing on the FY-3D MERSI time-series image data; Step 2: Perform fully automated cloud detection based on time-series analysis on the preprocessed FY-3D MERSI time-series image to obtain a time-series clear sky / cloud mask; Using pattern matching, key points are identified in the temporal surface reflectance curves of pixels, the temporal surface reflectance curves sorted from smallest to largest, and the sorted temporal surface reflectance curves after first-order differentiation. Key points include local minima and local maxima. Local minima indicate the transition from flat areas to gradually rising areas, while local maxima indicate the transition from gradually rising areas to highly flat areas. Flat areas are relatively low-valued segments in the sorted temporal surface reflectance curves after first-order differentiation, with derivative values ​​close to zero, corresponding to clear skies or cloud shadows. Gradually rising areas are segments where the values ​​gradually increase, with derivative values ​​gradually increasing, corresponding to the transition from clear skies to cloud areas. Highly flat areas are relatively high-valued segments in the sorted temporal surface reflectance curves after first-order differentiation, with derivative values ​​close to 0, corresponding to cloud areas. By using key points, the sorted temporal surface reflectance curves after first-order differentiation are divided into flat areas, gently rising areas, and highly flat areas. Since some shadows with low reflectance exist within the flat areas, the critical values ​​of the boundary lines on both sides of the clear-sky areas are determined. Move towards the center, discarding the critical regions on both sides, that is, take... The time period within the range is a clear sky area; Step 3: Perform radiometric consistency processing on the preprocessed FY-3D MERSI time series data; Step 4: Perform object-oriented segmentation on the time-series clear sky / cloud mask obtained in Step 2 to obtain clear sky and cloud mask objects; Step 5: Based on the optimal strategy, synthesize the clear sky map by combining the cloud mask objects obtained in Step 4 with the FY-3D MERSI time series data processed by radiometric consistency in Step 3. The cloud mask objects obtained in step 4 are combined with the FY-3D MERSI time-series images processed for radiometric consistency in step 3 to create a clear sky image. Clear sky objects are selected according to the optimal strategy: ① For adjacent cloud mask objects, clear sky objects from the same or similar dates are selected for filling; ② If ① is satisfied, the clear sky object with the smallest satellite zenith angle and solar zenith angle is selected; ③ If ① and ② are satisfied, the clear sky object with the best image quality is selected. Following the above optimal strategy, clear sky objects are selected from other time-series images to fill the cloud mask objects, and a composite mask is generated. The value of each pixel in the composite mask represents which image the spectrum of that pixel in the composite clear sky image comes from. Finally, the composite mask is used to fill and generate a clear sky image. Step 6: Perform feathering, color enhancement, and texture enhancement post-processing on the synthesized clear sky image to obtain the final clear sky image.

2. The method for synthesizing clear-sky images from an FY-3D medium-resolution imager as described in claim 1, characterized in that: In step 1, atmospheric correction is performed on the FY-3D MERSI time-series image data based on NASA's CREFL method to obtain the true surface reflectance in the visible light band. First, the original satellite data, including observation data and geolocation data, is read in, and the 1km geolocation data is resampled to 250m to match the spatial resolution of the Earth observation data. Input the DN values ​​of channels 24 and 25 (red, green, blue, near-infrared, and thermal infrared) from the L1 observation data, along with auxiliary data from the geolocation file, solar and satellite observation angles, and altitude information, as input parameters for atmospheric correction. Atmospheric correction is performed using the CREFL method based on 6S radiative transfer. The true surface reflectance in the visible light band is calculated based on the relationship between apparent atmospheric reflectance, path reflectance, and true surface reflectance. The specific calculation method is as follows: （1） In the formula, Atmospheric apparent reflectance is calculated using the original DN value; Path reflectivity; and Atmospheric transmittance related to observation geometry; The solar zenith angle; The zenith angle of the satellite; It is the relative azimuth angle; S Atmospheric albedo; Let be the surface reflectance to be determined.

3. The clear sky image synthesis method of FY-3D medium resolution imager as described in claim 2, characterized in that: In step 1, the GLT (Geographic Lookup Table) method is used for geometric correction. This involves constructing a geographic lookup table using the longitude and latitude files from the 1km observation geolocation information file. This table is then used to perform geometric correction on the atmospherically corrected FY-3D MERSI time-series image data. The actual geographic location of a given initial pixel can be retrieved from the lookup table. After geometric correction, the FY-3D MERSI time-series image data also needs to be geometrically reprojected. During geometric reprojection, a nadir distance file is generated simultaneously. When stitching together the atmospherically and geometrically corrected FY-3D MERSI time-series images, if multiple images cover the target area, candidate images with nadir or near-nadir imaging are prioritized.

4. The method for synthesizing clear-sky images from an FY-3D medium-resolution imager as described in claim 1, characterized in that: In step 2, for a given pixel, the temporal surface reflectance data obtained in step 1 are sorted from smallest to largest. , Indicates surface reflectance, Let be the time corresponding to this surface reflectance; the first derivative of each time series point is approximately calculated using the backward difference method on the sorted time series surface reflectance data, i.e.: （2） In the formula, Denotes the first derivative. , Indicates the first i+ 1. i Reflectance at each time point , Indicates the first i+ 1. i The time corresponding to each time series point; For the derivative of the last time-series point of the sorted temporal surface reflectance data, select the endpoint. and the previous adjacent point Substitute into formula (2) to perform the calculation.

5. The clear sky image synthesis method of FY-3D medium resolution imager as described in claim 1, characterized in that: In step 6, the cloud mask object from step 4 is used to post-process the initial clear sky composite result, outputting the final clear sky image. The post-processing includes: ① Feathering transition processing, using the cloud mask object as auxiliary information to perform feathering transition along the boundary of the cloud object to eliminate stitching marks; ② Color enhancement processing, using infrared band information to compensate the green channel, making the vegetation tone more natural; ③ Texture enhancement processing, using the DEM to perform texture enhancement processing on the surface reflectance image, using the gray-level co-occurrence matrix to extract the texture features of the clear sky image, using the texture features as the elevation variable of the DEM to fit, generating a new texture-enhanced DEM, and then applying a local contrast enhancement algorithm to obtain the enhanced image.

6. A clear-sky image synthesis system for an FY-3D medium-resolution imager, characterized in that, It includes a processor and a memory, the memory being used to store program instructions, and the processor being used to call the program instructions in the memory to execute the clear sky image synthesis method of the FY-3D medium resolution imager as described in any one of claims 1-5.

7. A clear-sky image synthesis system for an FY-3D medium-resolution imager, characterized in that, It includes a readable storage medium on which a computer program is stored, and when the computer program is executed, it implements a clear sky synthesis method for FY-3D medium resolution imager as described in any one of claims 1-5.

Images