A method and system for generating synthetic cloud images based on meteorological satellite data

Abstract

The present invention discloses a method and system for generating a synthetic cloud map based on meteorological satellite data, and relates to the field of image processing. The method comprises: obtaining a target multi-star cloud map mosaic image set; performing image preprocessing on the image set to obtain a cloud map mosaic image set to be extracted; collecting multiple regional points and matching the initial threshold value set; calculating the sun angle based on the geographical location to obtain multiple first correction coefficients; retrieving regional elevation data and matching it with multiple regional points to obtain multiple second correction coefficients; correcting the initial threshold value set to obtain a corrected threshold value set; extracting the cloud map mosaic image set to be extracted based on the corrected threshold value set, and superimposing it on the clear sky map of the corresponding month to obtain a target synthetic cloud map set. The technical problem of low efficiency and accuracy of cloud map extraction in the prior art is solved, and the technical effect of improving the accuracy of cloud map data processing is achieved.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a method and system for generating a synthetic cloud image based on meteorological satellite data. Background Art

[0002] Meteorological satellites play a vital role in monitoring and predicting weather systems. As an important part of meteorological satellite data, synthetic cloud images can provide large-scale and continuous cloud distribution information, which is of great significance to weather forecasting, climate change research and other fields. Meteorological satellite data covers cloud image information at multiple time points and multiple regional points. These data need to be effectively combined and processed to generate synthetic cloud images that can reflect real weather conditions. In addition, factors such as the geographical location and elevation of different regions will also affect the generation and extraction of cloud images. Existing methods are difficult to accurately extract cloud features and perform precise classification, resulting in low accuracy of the generated synthetic cloud images. Summary of the invention

[0003] The embodiments of the present application provide a method and system for generating a synthetic cloud image based on meteorological satellite data, which solves the technical problems of low efficiency and accuracy of cloud image extraction in the prior art.

[0004] In view of the above problems, an embodiment of the present application provides a method and system for generating a synthetic cloud image based on meteorological satellite data.

[0005] A first aspect of an embodiment of the present application provides a method for generating a synthetic cloud image based on meteorological satellite data, the method comprising:

[0006] Acquire a target multi-nebula image mosaic set, wherein the target multi-nebula image mosaic has a month mark;

[0007] Performing image preprocessing on the target multi-star cloud image mosaic image set to obtain a star cloud image mosaic image set to be extracted;

[0008] Collect multiple regional points at the current moment and match them with the initial threshold set, where each initial threshold corresponds to one regional point;

[0009] Calculate the sun angle based on the geographical locations of the multiple regional points to obtain multiple first correction coefficients;

[0010] Retrieving regional elevation data, matching it with the plurality of regional points, and obtaining a plurality of second correction coefficients;

[0011] Modifying the initial threshold set based on the multiple first correction coefficients and the multiple second correction coefficients to obtain a modified threshold set;

[0012] The cloud image mosaic image set to be extracted is extracted based on the correction threshold set, and superimposed on the clear sky image of the corresponding month to obtain a target synthetic cloud image set.

[0013] A second aspect of the embodiment of the present application provides a synthetic cloud image generation system based on meteorological satellite data, the system comprising:

[0014] An image acquisition module, the image acquisition module is used to acquire a target multi-star cloud image mosaic image set, wherein the target multi-star cloud image mosaic image has a month mark;

[0015] A processing module, the processing module is used to perform image preprocessing on the target multi-star cloud image mosaic image set to obtain a star cloud image mosaic image set to be extracted;

[0016] A collection module, the collection module is used to collect multiple regional points at the current moment and match the initial threshold set, wherein each initial threshold corresponds to a regional point;

[0017] A calculation module, the calculation module is used to calculate the sun angle based on the geographical locations of the multiple regional points to obtain multiple first correction coefficients;

[0018] A matching module, the matching module is used to retrieve the regional elevation data, match it with the multiple regional points, and obtain multiple second correction coefficients;

[0019] A correction module, the correction module is used to correct the initial threshold set based on the multiple first correction coefficients and the multiple second correction coefficients to obtain a corrected threshold set;

[0020] The superposition module is used to extract the cloud image mosaic image set to be extracted based on the correction threshold set, and superimpose it on the clear sky map of the corresponding month to obtain a target synthetic cloud image set.

[0021] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0022] The cloud map data collected from different satellite resources are integrated and spliced ​​to form a target multi-star cloud map splicing image set with month identification. Subsequently, the target multi-star cloud map splicing image set is converted into a cloud map splicing image set to be extracted after image preprocessing. By collecting multiple regional points at the current moment and matching an initial threshold for each regional point, a preliminary threshold set is obtained. Next, the sun angle is calculated according to the geographical location of each regional point to obtain multiple first correction coefficients. At the same time, the regional elevation data matching the multiple regional points is retrieved to obtain multiple second correction coefficients. Combining the above two correction coefficients, the initial threshold set is dynamically corrected to obtain a more accurate correction threshold set. Using the corrected threshold set, the cloud map splicing image set to be extracted is accurately extracted. Then, the extracted cloud map information is superimposed on the clear sky map of the corresponding month, and finally the target synthetic cloud map set is obtained, which solves the technical problems of low efficiency and accuracy of cloud map extraction in the prior art and achieves the technical effect of improving the accuracy of cloud map data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 A schematic diagram of a flow chart of a method for generating a synthetic cloud image based on meteorological satellite data provided in an embodiment of the present application;

[0025] Figure 2 A schematic diagram of the structure of a synthetic cloud image generation system based on meteorological satellite data provided in an embodiment of the present application.

[0026] Explanation of the reference numerals: image acquisition module 11 , processing module 12 , collection module 13 , calculation module 14 , matching module 15 , correction module 16 , superposition module 17 . DETAILED DESCRIPTION

[0027] The embodiments of the present application solve the technical problems of low efficiency and accuracy of cloud image extraction in the prior art by providing a method and system for generating a synthetic cloud image based on meteorological satellite data.

[0028] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0029] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0030] Embodiment 1

[0031] like Figure 1 As shown, the embodiment of the present application provides a method for generating a synthetic cloud image based on meteorological satellite data, wherein the method comprises:

[0032] Acquire a target multi-nebula image mosaic set, wherein the target multi-nebula image mosaic has a month mark;

[0033] Collect target multi-star cloud map mosaic images from meteorological satellites, and attach a month label to each image to form a target multi-star cloud map mosaic image set. The target multi-star cloud map mosaic image is a comprehensive image obtained by integrating and mosaicking cloud map data from different satellites.

[0034] Performing image preprocessing on the target multi-star cloud image mosaic image set to obtain a star cloud image mosaic image set to be extracted;

[0035] Image preprocessing is performed on each image in the target multi-star cloud image stitching image set to convert it into a grayscale image. The images after image preprocessing constitute the cloud image stitching image set to be extracted.

[0036] Further, the method comprises:

[0037] Traversing the target multi-star cloud image mosaic image set to perform image denoising processing to obtain an initial denoised star cloud image mosaic image set;

[0038] Using a preset filter to perform mean filtering on the initial denoised cloud image mosaic image set to obtain a stage denoised cloud image mosaic image set;

[0039] Performing clarity identification on the denoised cloud image mosaic image set in the above stage respectively, and counting the number of images whose clarity is less than a preset clarity;

[0040] If the number of images is less than or equal to the preset number of images, the denoised cloud image mosaic image set of the stage is used as the cloud image mosaic image set to be extracted;

[0041] If the number of images is greater than the preset number of images, feedback information is generated to optimize the size of the preset filter. Since satellite images may be interfered by various noises during acquisition and transmission, such as sensor noise, signal processing noise, etc., noise reduction is required to ensure data quality. Traverse each image in the target multi-star cloud image mosaic image set, apply an appropriate noise reduction algorithm to each image, such as median filtering, Gaussian filtering or non-local mean filtering, etc., to reduce the noise in the image, and obtain an initial denoised cloud image mosaic image set after noise reduction. Then, the initial denoised cloud image mosaic image set is subjected to mean filtering using a preset filter. Specifically, according to the size of the preset filter (i.e., the pixel range involved in the mean calculation), for each pixel in the image, the average value of the pixel grayscale value in its neighborhood is calculated, and the original pixel value is replaced by this average value, and this process is repeated until the entire image is processed. For pixels at the edge of the image, methods such as no processing, mirror filling or zero filling can be used to deal with the boundary effect. After mean filtering, a stage denoised cloud image mosaic image set is obtained. After obtaining the stage denoised cloud image mosaic set, the clarity of each stage denoised cloud image mosaic can be evaluated, and the number of images with clarity lower than the preset clarity can be counted. If the number of images is less than the preset number of images, the stage denoised cloud image mosaic set is used as the cloud image mosaic set to be extracted. If the number of images is greater than the preset number of images, feedback information is generated to optimize the size of the preset filter.

[0042] Furthermore, the method further comprises:

[0043] Acquire a plurality of sample target multi-star cloud image mosaic image sets and a plurality of sample initial denoised star cloud image mosaic image sets as a training sample data set;

[0044] Dividing the training sample data set into n groups of training sample data sets, and training the network framework constructed based on the convolutional neural network in sequence;

[0045] After a set of training sample data sets are trained, determine whether the output accuracy meets the requirements. If not, adjust the network parameters of the network framework;

[0046] If yes, then the next set of training sample data sets is trained until the output reaches convergence, and the trained denoising network layer is obtained;

[0047] The target multi-star cloud image mosaic image set is input into the denoising network layer to obtain the initial denoising cloud image mosaic image set.

[0048] Preferably, a convolutional neural network (CNN) is used to perform denoising of cloud image mosaics. A target multi-star cloud image mosaic image set of multiple samples and an initial denoised cloud image mosaic image set of multiple samples are obtained, and these two sets are used as training sample data sets. The entire training sample data set is divided into n groups, each group containing a certain number of input images and their corresponding expected output images. Initialize a network framework based on a convolutional neural network (CNN), train n groups of training sample data sets in turn, and use them to iteratively update the network framework. After each group of training sample data sets is trained, the output accuracy of the network is evaluated. If the output accuracy does not meet the preset requirements, the parameters of the network framework are adjusted, such as the learning rate, weight, etc., and then the next round of training is continued. If the output accuracy meets the requirements, the next group of training sample data sets is trained until all groups have completed training and the network reaches a convergence state. The target multi-star cloud image mosaic image set is input into the denoising network layer that has been trained, and the denoising network layer is used to process the denoising task of the target multi-star cloud image mosaic image. Through the denoising process of the denoising network layer, the corresponding initial denoised cloud map mosaic image set is obtained.

[0049] Collect multiple regional points at the current moment and match them with the initial threshold set, where each initial threshold corresponds to one regional point;

[0050] The time when the image was collected and the regional points in the image are obtained. The regional points correspond to different regions in the image, and an initial threshold is matched for each regional point.

[0051] Further, the method comprises:

[0052] Collect the geographical locations of multiple regional points and match multiple empirical threshold sets;

[0053] Performing density analysis on the multiple empirical threshold sets respectively to determine multiple center thresholds;

[0054] An initial threshold mean analysis is performed based on the multiple center thresholds to obtain the initial threshold set.

[0055] Determine multiple regional points where data needs to be collected, and obtain their geographical location information. For each regional point, match it according to the existing empirical threshold set. Compare the data of the regional point with the threshold in each empirical threshold set to determine which empirical threshold set the regional point belongs to. Perform density analysis on each matched empirical threshold set, calculate the density of the data in each empirical threshold set, and determine its central threshold. The central threshold is the threshold that best represents each regional point when dividing the cloud layer. Based on the obtained central threshold, further analyze it, calculate the average of multiple empirical thresholds within the preset analysis step around the central threshold, and obtain the initial threshold.

[0056] Furthermore, the method further comprises:

[0057] Selecting a first empirical threshold set from the multiple empirical threshold sets;

[0058] Calculating the mean of the first empirical threshold set to obtain a first initial threshold mean point;

[0059] Counting the aggregation threshold value within a preset analysis step length around the first initial threshold mean point to obtain a first aggregation threshold value;

[0060] Aggregate threshold quantity statistics are performed on multiple first empirical thresholds within the preset analysis step to obtain multiple first empirical aggregate threshold quantity sets.

[0061] One of the multiple empirical threshold sets is selected as the first empirical threshold set, and the thresholds in the first empirical threshold set are averaged to obtain a first initial threshold mean point. Within the preset analysis step length around the first initial threshold mean point, the aggregation threshold amount is counted. The aggregation threshold amount may refer to the number of empirical thresholds in the area constructed with the first initial threshold mean point as the center and the preset analysis step length as the radius. The larger the aggregation threshold amount, the more empirical thresholds are clustered around the corresponding first initial threshold mean point, and the more representative the first initial threshold mean point is. Aggregation threshold amount statistics are performed on multiple first empirical thresholds within the preset analysis step length to obtain multiple first empirical aggregation threshold amount sets, which can be used to represent the aggregation of multiple empirical thresholds within a preset range. Among them, the preset analysis step length is a distance pre-set by a technician in this field when performing aggregation threshold amount statistics.

[0062] Furthermore, the method further comprises:

[0063] respectively calculating increments of the first aggregation threshold value and the plurality of first experience aggregation threshold value sets to determine a leading direction;

[0064] Performing incremental analysis based on the leading direction until the incremental value is less than or equal to a preset incremental value, thereby obtaining a first center threshold value;

[0065] A density analysis is performed based on the multiple empirical threshold sets to obtain multiple center thresholds.

[0066] Compare the first aggregation threshold value with each set of first experience aggregation threshold values, and calculate the increment between them. The increment can be a numerical difference or change. Analyze these increments to determine which direction has a larger or more significant change, and this direction is the leading direction. Based on the leading direction, perform incremental analysis. According to the preset incremental standard, gradually adjust the threshold and observe the incremental change. The goal of incremental analysis is to find a suitable threshold so that the increment is less than or equal to the preset increment. When the incremental analysis meets the preset conditions (the increment is less than or equal to the preset increment), it indicates that the area with the most dense distribution of empirical thresholds has been reached. In order to save computing resources, the empirical threshold at this time is used as the first central threshold. Perform density analysis on multiple empirical threshold sets to determine which thresholds are more central or representative. Based on the results of density analysis, multiple central thresholds are obtained.

[0067] Calculate the sun angle based on the geographical locations of the multiple regional points to obtain multiple first correction coefficients;

[0068] Based on the geographic locations of multiple regional points, the sun's angle is calculated using the given date and time and the latitude and longitude information of each regional point using the calculation formulas for the sun's altitude and azimuth. The sun's altitude is the angle between the sun's rays and the horizontal plane, and the sun's azimuth is the angle between the sun's rays projected on the horizontal plane and the north direction. Based on the calculated sun's angle, the first correction coefficient is calculated at each regional point. The first correction coefficient can reflect the effect of the sun's angle on the threshold.

[0069] Retrieving regional elevation data, matching it with the plurality of regional points, and obtaining a plurality of second correction coefficients;

[0070] The elevation data of the region is obtained from the geographic information system. The elevation data is used to represent the altitude of the surface. The geographical locations of multiple regional points are matched with the obtained regional elevation data. For example, the corresponding longitude and latitude coordinates can be used to find the corresponding elevation values. According to the matched elevation data, a second correction coefficient is calculated for each regional point. The second correction coefficient can reflect the influence of terrain factors on the threshold.

[0071] Further, the method comprises:

[0072] Determine snow mountain altitude data based on matching the regional elevation data with the geographical locations of the plurality of regional points;

[0073] A plurality of second correction coefficients are generated according to the gray value range matched with the snow mountain altitude data.

[0074] On the basis of matching, it is determined whether each regional point is a snow mountain, and the corresponding altitude data is obtained. According to the snow mountain altitude data, the corresponding gray value range is determined. For each regional point, a second correction coefficient is generated according to its gray value range, and the second correction coefficient can reflect the influence of the altitude on the threshold. Optionally, according to the gray value range of the snow mountain image corresponding to different snow mountain altitudes, the influence range value on the threshold is determined, so as to correct the initial threshold set, thereby providing a basis for accurately extracting clouds from the image.

[0075] Modifying the initial threshold set based on the multiple first correction coefficients and the multiple second correction coefficients to obtain a modified threshold set;

[0076] By correcting the initial threshold set based on the first correction coefficient and the second correction coefficient, a more accurate and reliable corrected threshold set can be obtained. Ensure that the first correction coefficient and the second correction coefficient of multiple regional points have been obtained. These coefficients reflect the influence of the sun angle and terrain factors on the threshold respectively. For each threshold in the initial threshold set, correction is performed according to the first correction coefficient and the second correction coefficient. Optionally, by constructing a threshold correction module, the multiple first correction coefficients and the multiple second correction coefficients are input into the threshold correction module, and the corrected threshold set is output. By obtaining multiple sample first correction coefficients, multiple sample second correction coefficients and multiple sample corrected threshold sets as training data, supervised training is performed on the framework constructed based on the convolutional neural network until the output reaches convergence, and the threshold correction module that has been trained is obtained. Collect all the corrected thresholds to form a corrected threshold set. The corrected threshold set should reflect the threshold that takes into account the influence of the sun angle and terrain factors. The threshold is to distinguish between clouds and non-clouds in the spliced ​​image by dividing the grayscale value.

[0077] The cloud image mosaic image set to be extracted is extracted based on the correction threshold set, and superimposed on the clear sky image of the corresponding month to obtain a target synthetic cloud image set.

[0078] Using the correction threshold set as a guide, the qualified images are extracted from the cloud map mosaic image set to be extracted. The grayscale value of each pixel in the cloud map mosaic image set to be extracted can be compared with the threshold in the correction threshold set to determine whether the pixel is a cloud area. If the grayscale value exceeds a certain threshold, the pixel can be considered to be a cloud area. The corrected cloud map image is superimposed on the clear sky image of the corresponding month. According to the superimposed image, the target synthetic cloud map set is obtained.

[0079] In summary, the embodiments of the present application have at least the following technical effects:

[0080] The cloud map data collected from different satellite resources are integrated and spliced ​​to form a target multi-star cloud map splicing image set with month identification. Subsequently, the target multi-star cloud map splicing image set is converted into a cloud map splicing image set to be extracted after image preprocessing. By collecting multiple regional points at the current moment and matching an initial threshold for each regional point, a preliminary threshold set is obtained. Next, the sun angle is calculated according to the geographical location of each regional point to obtain multiple first correction coefficients. At the same time, the regional elevation data matching the multiple regional points is retrieved to obtain multiple second correction coefficients. Combining the above two correction coefficients, the initial threshold set is dynamically corrected to obtain a more accurate correction threshold set. Using the corrected threshold set, the cloud map splicing image set to be extracted is accurately extracted. Then, the extracted cloud map information is superimposed on the clear sky map of the corresponding month, and finally the target synthetic cloud map set is obtained, which solves the technical problems of low efficiency and accuracy of cloud map extraction in the prior art and achieves the technical effect of improving the accuracy of cloud map data processing.

[0081] Embodiment 2

[0082] Based on the same inventive concept as the method for generating a synthetic cloud image based on meteorological satellite data in the aforementioned embodiment, Figure 2 As shown, the present application provides a synthetic cloud image generation system based on meteorological satellite data, and the system and method embodiments in the present application embodiments are based on the same inventive concept. The system includes:

[0083] An image acquisition module 11, wherein the image acquisition module 11 is used to acquire a target multi-nebula image mosaic image set, wherein the target multi-nebula image mosaic image has a month mark;

[0084] A processing module 12, the processing module 12 is used to perform image preprocessing on the target multi-star cloud image mosaic image set to obtain a star cloud image mosaic image set to be extracted;

[0085] A collection module 13, the collection module 13 is used to collect multiple regional points at the current moment and match the initial threshold set, wherein each initial threshold corresponds to a regional point;

[0086] A calculation module 14, the calculation module 14 is used to calculate the sun angle based on the geographical locations of the multiple regional points to obtain multiple first correction coefficients;

[0087] A matching module 15, the matching module 15 is used to retrieve the regional elevation data, match it with the multiple regional points, and obtain multiple second correction coefficients;

[0088] A correction module 16, wherein the correction module 16 is used to correct the initial threshold value set based on the multiple first correction coefficients and the multiple second correction coefficients to obtain a corrected threshold value set;

[0089] The superposition module 17 is used to extract the cloud image mosaic image set to be extracted based on the correction threshold set, and superimpose it on the clear sky image of the corresponding month to obtain a target synthetic cloud image set.

[0090] Furthermore, the processing module 12 is used to execute the following method:

[0091] Traversing the target multi-star cloud image mosaic image set to perform image denoising processing to obtain an initial denoised star cloud image mosaic image set;

[0092] Using a preset filter to perform mean filtering on the initial denoised cloud image mosaic image set to obtain a stage denoised cloud image mosaic image set;

[0093] Performing clarity identification on the denoised cloud image mosaic image set in the above stage respectively, and counting the number of images whose clarity is less than a preset clarity;

[0094] If the number of images is less than or equal to the preset number of images, the denoised cloud image mosaic image set of the stage is used as the cloud image mosaic image set to be extracted;

[0095] If the number of images is greater than the preset number of images, feedback information is generated to optimize the size of the preset filter.

[0096] Furthermore, the processing module 12 is used to execute the following method:

[0097] Acquire a plurality of sample target multi-star cloud image mosaic image sets and a plurality of sample initial denoised star cloud image mosaic image sets as a training sample data set;

[0098] Dividing the training sample data set into n groups of training sample data sets, and training the network framework constructed based on the convolutional neural network in sequence;

[0099] After a set of training sample data sets are trained, determine whether the output accuracy meets the requirements. If not, adjust the network parameters of the network framework;

[0100] If yes, then the next set of training sample data sets is trained until the output reaches convergence, and the trained denoising network layer is obtained;

[0101] The target multi-star cloud image mosaic image set is input into the denoising network layer to obtain the initial denoising cloud image mosaic image set.

[0102] Furthermore, the acquisition module 13 is used to perform the following method:

[0103] Collect the geographical locations of multiple regional points and match multiple empirical threshold sets;

[0104] Performing density analysis on the multiple empirical threshold sets respectively to determine multiple center thresholds;

[0105] An initial threshold mean analysis is performed based on the multiple center thresholds to obtain the initial threshold set.

[0106] Furthermore, the acquisition module 13 is used to perform the following method:

[0107] Selecting a first empirical threshold set from the multiple empirical threshold sets;

[0108] Calculating the mean of the first empirical threshold set to obtain a first initial threshold mean point;

[0109] Counting the aggregation threshold value within a preset analysis step length around the first initial threshold mean point to obtain a first aggregation threshold value;

[0110] Aggregate threshold quantity statistics are performed on multiple first empirical thresholds within the preset analysis step to obtain multiple first empirical aggregate threshold quantity sets.

[0111] Furthermore, the acquisition module 13 is used to perform the following method:

[0112] respectively calculating increments of the first aggregation threshold value and the plurality of first experience aggregation threshold value sets to determine a leading direction;

[0113] Performing incremental analysis based on the leading direction until the incremental value is less than or equal to a preset incremental value, thereby obtaining a first center threshold value;

[0114] A density analysis is performed based on the multiple empirical threshold sets to obtain multiple center thresholds.

[0115] Furthermore, the matching module 15 is used to perform the following method:

[0116] Determine snow mountain altitude data based on matching the regional elevation data with the geographical locations of the plurality of regional points;

[0117] A plurality of second correction coefficients are generated according to the gray value range matched with the snow mountain altitude data.

[0118] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. Other embodiments are within the scope of the attached claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0119] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0120] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for generating a synthetic cloud image based on meteorological satellite data, characterized in that: The method comprises: Acquire a target multi-nebula image mosaic set, wherein the target multi-nebula image mosaic has a month mark; Performing image preprocessing on the target multi-star cloud image mosaic image set to obtain a star cloud image mosaic image set to be extracted; Collect multiple regional points at the current moment and match them with the initial threshold set, where each initial threshold corresponds to one regional point; Calculate the sun angle based on the geographical locations of the multiple regional points to obtain multiple first correction coefficients; Retrieving regional elevation data, matching it with the plurality of regional points, and obtaining a plurality of second correction coefficients; Modifying the initial threshold set based on the multiple first correction coefficients and the multiple second correction coefficients to obtain a modified threshold set; Extracting the cloud image mosaic image set to be extracted based on the correction threshold set, and superimposing it on the clear sky image of the corresponding month to obtain a target synthetic cloud image set; Traversing the target multi-star cloud image mosaic image set to perform image denoising processing to obtain an initial denoised star cloud image mosaic image set; Using a preset filter to perform mean filtering on the initial denoised cloud image mosaic image set to obtain a stage denoised cloud image mosaic image set; Performing clarity identification on the denoised cloud image mosaic image set in the above stage respectively, and counting the number of images whose clarity is less than a preset clarity; If the number of images is less than or equal to the preset number of images, the denoised cloud image mosaic image set of the stage is used as the cloud image mosaic image set to be extracted; If the number of images is greater than the preset number of images, feedback information is generated to optimize the size of the preset filter; The method of modifying the initial threshold set based on the plurality of first modification coefficients and the plurality of second modification coefficients to obtain the modified threshold set includes: Constructing a threshold correction module, inputting the plurality of first correction coefficients and the plurality of second correction coefficients into the threshold correction module, and outputting the correction threshold set; Obtaining a plurality of sample first correction coefficients, a plurality of sample second correction coefficients and a plurality of sample correction threshold sets as training data, performing supervised training on a framework constructed based on a convolutional neural network until the output reaches convergence, and obtaining the threshold correction module that has completed training; Collecting all the modified thresholds to form a modified threshold set; The correction threshold set should reflect the threshold that takes into account the influence of the sun angle and terrain factors. The threshold is to distinguish between the cloud layer and the non-cloud layer in the stitched image by dividing the gray value; The cloud image mosaic image set to be extracted is extracted based on the correction threshold set, and superimposed on the clear sky image of the corresponding month to obtain the target synthetic cloud image set, including: Using the modified threshold set as a guide, extracting images that meet the conditions from the set of cloud image mosaics to be extracted; Compare the grayscale value of each pixel in the cloud image mosaic set to be extracted with the threshold in the correction threshold set to determine whether the pixel is a cloud area; if the grayscale value exceeds a certain threshold, the pixel is considered to be a cloud area; superimposing the corrected cloud image onto the clear sky image of the corresponding month; According to the superimposed images, a set of target synthetic cloud images is obtained.

2. The method according to claim 1, characterized in that The method further comprises: Acquire a plurality of sample target multi-star cloud image mosaic image sets and a plurality of sample initial denoised star cloud image mosaic image sets as a training sample data set; Dividing the training sample data set into n groups of training sample data sets, and training the network framework constructed based on the convolutional neural network in sequence; After a set of training sample data sets are trained, determine whether the output accuracy meets the requirements. If not, adjust the network parameters of the network framework; If yes, then the next set of training sample data sets is trained until the output reaches convergence, and the trained denoising network layer is obtained; The target multi-star cloud image mosaic image set is input into the denoising network layer to obtain the initial denoising cloud image mosaic image set.

3. The method according to claim 1, characterized in that The method further comprises: Collect the geographical locations of multiple regional points and match multiple empirical threshold sets; Performing density analysis on the multiple empirical threshold sets respectively to determine multiple center thresholds; An initial threshold mean analysis is performed based on the multiple center thresholds to obtain the initial threshold set.

4. The method according to claim 3, characterized in that The method further comprises: Selecting a first empirical threshold set from the multiple empirical threshold sets; Calculating the mean of the first empirical threshold set to obtain a first initial threshold mean point; Counting the aggregation threshold value within a preset analysis step length around the first initial threshold mean point to obtain a first aggregation threshold value; Aggregate threshold quantity statistics are performed on multiple first empirical thresholds within the preset analysis step to obtain multiple first empirical aggregate threshold quantity sets.

5. The method according to claim 4, characterized in that The method further comprises: respectively calculating increments of the first aggregation threshold value and the plurality of first experience aggregation threshold value sets to determine a leading direction; Performing incremental analysis based on the leading direction until the incremental value is less than or equal to a preset incremental value, thereby obtaining a first center threshold value; A density analysis is performed based on the multiple empirical threshold sets to obtain multiple center thresholds.

6. The method according to claim 1, characterized in that The method further comprises: Determine snow mountain altitude data based on matching the regional elevation data with the geographical locations of the plurality of regional points; A plurality of second correction coefficients are generated according to the gray value range matched with the snow mountain altitude data.

7. A synthetic cloud image generation system based on meteorological satellite data, characterized in that: A method for generating a synthetic cloud image based on meteorological satellite data according to any one of claims 1 to 6, the system comprising: An image acquisition module, the image acquisition module is used to acquire a target multi-star cloud image mosaic image set, wherein the target multi-star cloud image mosaic image has a month mark; A processing module, the processing module is used to perform image preprocessing on the target multi-star cloud image mosaic image set to obtain a star cloud image mosaic image set to be extracted; A collection module, the collection module is used to collect multiple regional points at the current moment and match the initial threshold set, wherein each initial threshold corresponds to a regional point; A calculation module, the calculation module is used to calculate the sun angle based on the geographical locations of the multiple regional points to obtain multiple first correction coefficients; A matching module, the matching module is used to retrieve the regional elevation data, match it with the multiple regional points, and obtain multiple second correction coefficients; A correction module, the correction module is used to correct the initial threshold set based on the multiple first correction coefficients and the multiple second correction coefficients to obtain a corrected threshold set; The superposition module is used to extract the cloud image mosaic image set to be extracted based on the correction threshold set, and superimpose it on the clear sky map of the corresponding month to obtain a target synthetic cloud image set.

Images