A cloud classification method and system based on meteorological satellite images

Abstract

The application discloses a cloud layer classification method and system based on meteorological satellite images, and belongs to the technical field of data processing.The method first determines a target region, obtains at least two satellite cloud images of the target region in a monitoring period, then cuts the first and second images to form a plurality of first and second sub-images, obtains a plurality of corresponding first and second texture features, matches each first and second texture feature, obtains a matching degree corresponding to each second sub-image according to a matching result, and determines an input weight according to the corresponding matching degree.The application compares the satellite cloud images in the monitoring period according to the matching degree of each sub-image, determines the weight of each input sub-image, and the weight of a sub-image with large change is low, so that the influence degree on the actual result is small, so that the neural network model does not need to identify each satellite cloud image, the identification speed is improved, and the computing power is saved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a cloud layer classification method and system based on meteorological satellite images. BACKGROUND

[0002] In satellite image processing, cloud layer classification is usually used for meteorology, agricultural monitoring or power equipment monitoring (such as photovoltaic power station efficiency evaluation). For example, in power equipment, cloud layers can affect solar power generation prediction, and cloud layer shielding leads to fluctuations in solar radiation, affecting power generation prediction, so it is necessary to monitor the cloud coverage range and movement trend in real time to adjust the power grid scheduling strategy.

[0003] In related art satellite image processing, the way of identifying cloud layer types by artificial intelligence is generally to identify multiple images in a period, and the time required to identify the cloud layer type in each image is relatively long due to the processing capacity of the device, and when there are many images in a period, the identification time may be significantly prolonged, which cannot achieve timeliness, thereby affecting the subsequent prediction process, and when only one image is identified, the credibility of one image cannot be judged in combination with other images, and the accuracy is poor. SUMMARY

[0004] The present application provides a cloud layer classification method and system based on meteorological satellite images to improve the above problems.

[0005] To achieve the above purpose, the present application adopts the following technical solutions:

[0006] In a first aspect, the present application provides a cloud layer classification method based on meteorological satellite images, the method comprising:

[0007] determining a target area, and obtaining at least two satellite cloud images of the target area in a monitoring period; wherein the two satellite cloud images are respectively a first image and a second image sorted by acquisition time;

[0008] cutting the first image to form a plurality of first sub-images, and obtaining a plurality of first texture features corresponding to the plurality of first sub-images;

[0009] cutting the second image to form a plurality of second sub-images, the plurality of second sub-images corresponding one-to-one to the plurality of first sub-images, and obtaining a plurality of second texture features corresponding to the plurality of second sub-images;

[0010] matching each of the first texture features and the second texture features, and obtaining a matching degree corresponding to each of the second sub-images according to the matching result; the matching degree is used to represent the similarity between the first texture features and the corresponding second texture features;

[0011] According to the matching degree, the plurality of second sub-images are divided into a same number of high-approval groups and low-approval groups; wherein the matching degree corresponding to the second sub-image belonging to the high-approval group is greater than the matching degree corresponding to the second sub-image belonging to the low-approval group;

[0012] The position information of each second texture feature and each second sub-image is introduced into a neural network recognition model, and classification data corresponding to a cloud layer type is obtained according to an output result of the neural network recognition model;

[0013] The weight of the second texture feature and the position information corresponding to the second sub-image belonging to the high-approval group is greater than the weight of the second texture feature and the position information corresponding to the second sub-image belonging to the low-approval group when the neural network recognition model is input.

[0014] In combination with the first aspect, in some embodiments, the introduction of the position information of each second texture feature and each second sub-image into the neural network recognition model, and the obtaining of the classification data corresponding to the cloud layer type according to the output result of the neural network recognition model specifically include:

[0015] A training set is obtained, the training set including a plurality of classification data that has been determined, a plurality of training sub-image texture feature data corresponding to the classification data, and training sub-image position data; the training sub-image position data including longitudinal position data and transverse position data;

[0016] An initial neural network recognition model is constructed, the initial neural network recognition model is trained based on the training set, the training of the initial neural network recognition model is stopped when a preset condition is met, and a trained neural network recognition model is output.

[0017] In combination with the first aspect, in some embodiments, the construction of the initial neural network recognition model, the training of the initial neural network recognition model based on the training set, the stopping of the training of the initial neural network recognition model when a preset condition is met, and the output of the trained neural network recognition model specifically include:

[0018] The classification data, the training sub-image texture feature data, the longitudinal position data, and the transverse position data in the training set are introduced into the initial neural network recognition model and iteratively calculated as three different dimensions; wherein the three dimensions constitute a feature vector of the classification data;

[0019] determine a loss function, obtain an error value in an iterative calculation process based on the loss function, stop the iterative calculation when the error value is less than a preset value, and output the neural network recognition model.

[0020] In combination with the first aspect, in some embodiments, the position information of each of the second texture features and each of the second sub-images is introduced into a neural network recognition model, and classification data corresponding to a cloud layer type is obtained according to an output result of the neural network recognition model, specifically including:

[0021] When the second texture features and the position information corresponding to the second sub-images are input into the neural network recognition model, different weights are given to the second texture features and the position information corresponding to the second sub-images belonging to different groups, and the following is met:

[0022] a / b=P1 / P2;

[0023] wherein a is the weight of the second texture features and the position information corresponding to the second sub-images belonging to the high recognition group when input into the neural network recognition model, b is the weight of the second texture features and the position information corresponding to the second sub-images belonging to the low recognition group when input into the neural network recognition model, P1 is the matching degree of the second sub-images belonging to the high recognition group, and P2 is the matching degree of the second sub-images belonging to the low recognition group.

[0024] In combination with the first aspect, in some embodiments, each of the first texture features and the second texture features is matched, and a matching degree corresponding to each of the second sub-images is obtained according to a matching result; the matching degree is used to represent the similarity between the first texture features and the corresponding second texture features, specifically including:

[0025] The first sub-image is converted and segmented to obtain a plurality of first gray scale images, and a first average gray value of the first sub-image is obtained;

[0026] The second sub-image is converted and segmented to obtain a plurality of second gray scale images, and a second average gray value corresponding to the second sub-image is obtained;

[0027] Based on the first average gray value and the second average gray value, an average difference value is determined;

[0028] Each of the first gray scale images is compared with the corresponding second gray scale image, and a plurality of target difference values are obtained;

[0029] comparing the target difference value with the average difference value, if the target difference value is greater than the average difference value, marking the corresponding second gray scale image as a target image;

[0030] determining the matching degree according to a proportional relationship between the target image quantity and the quantity of the second gray scale images.

[0031] In combination with the first aspect, in some embodiments, the determining the matching degree according to a proportional relationship between the target image quantity and the quantity of the second gray scale images specifically comprises:

[0032] P = K (N / n) * 100%;

[0033] wherein P is the matching degree, K is a constant, N is the quantity of the second gray scale images, and n is the target image quantity.

[0034] In a second aspect, the present application provides a cloud layer classification system based on meteorological satellite images, which is configured to:

[0035] determining a target region, and obtaining at least two satellite cloud images of the target region in a monitoring period; wherein the two satellite cloud images are respectively a first image and a second image sorted according to acquisition time;

[0036] segmenting the first image to form a plurality of first sub-images, and obtaining a plurality of first texture features corresponding to the plurality of first sub-images;

[0037] segmenting the second image to form a plurality of second sub-images, the plurality of second sub-images corresponding one-to-one to the plurality of first sub-images, and obtaining a plurality of second texture features corresponding to the plurality of second sub-images;

[0038] matching each of the first texture features with the second texture features, and obtaining a matching degree corresponding to each of the second sub-images according to the matching result; the matching degree is used to represent the similarity between the first texture features and the corresponding second texture features;

[0039] dividing the plurality of second sub-images into a same number of high-approval groups and low-approval groups according to the matching degrees; wherein the matching degree corresponding to the second sub-image belonging to the high-approval group is greater than the matching degree corresponding to the second sub-image belonging to the low-approval group;

[0040] inputting each of the second texture features and the position information of each of the second sub-images into a neural network recognition model, and obtaining classification data corresponding to a cloud layer type according to an output result of the neural network recognition model;

[0041] The weight of the second texture feature and the position information corresponding to the second sub-image belonging to the high-recognized group is greater than the weight of the second texture feature and the position information corresponding to the second sub-image belonging to the low-recognized group when the neural network recognition model is input.

[0042] Optionally, the system is configured to:

[0043] The first texture feature and the second texture feature corresponding to each are matched, and a matching degree corresponding to each second sub-image is obtained according to a matching result; the matching degree is used to represent a similarity degree of the first texture feature and the corresponding second texture feature, and specifically includes:

[0044] The first sub-image is converted and segmented to obtain a plurality of first gray scale images, and a first average gray value of the first sub-image is obtained;

[0045] The second sub-image is converted and segmented to obtain a plurality of second gray scale images, and a second average gray value corresponding to the second sub-image is obtained;

[0046] Based on the first average gray value and the second average gray value, an average difference value is determined;

[0047] Each first gray scale image is compared with the corresponding second gray scale image, and a plurality of target difference values are obtained;

[0048] The plurality of target difference values are compared with the average difference value, and if the target difference value is greater than the average difference value, the corresponding second gray scale image is marked as a target image;

[0049] According to a proportional relationship between the number of target images and the number of second gray scale images, the matching degree is determined.

[0050] The third aspect of the embodiment of the application provides an electronic device, and the electronic device comprises:

[0051] At least one processor and a memory connected in communication with the at least one processor;

[0052] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method provided in the first aspect of the embodiment of the application.

[0053] The fourth aspect of the embodiment of the application provides a computer readable storage medium, and the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method provided in the first aspect of the embodiment of the application.

[0054] In summary, the above method and system have the following technical effects:

[0055] The cloud layer classification method and system based on meteorological satellite images provided by the application first determine a target area, obtain at least two satellite cloud images of the target area in a monitoring period, then split the first image to form a plurality of first sub-images, and obtain a plurality of first texture features corresponding to the plurality of first sub-images, then split the second image to form a plurality of second sub-images, the plurality of second sub-images correspond one-to-one to the plurality of first sub-images, and obtain a plurality of second texture features corresponding to the plurality of first sub-images, then match each corresponding first texture feature and second texture feature, obtain a matching degree corresponding to each second sub-image according to the matching result, and then determine the weight of the second sub-image corresponding data when inputting the neural network recognition model; the cloud layer classification method and system based on meteorological satellite images provided by the application compare the satellite cloud images in the monitoring period, determine the weight of each sub-image inputting the neural network model according to the matching degree of each sub-image, the weight of the sub-image with large changes is low, and the influence degree on the actual result is small, the neural network model does not need to identify each satellite cloud image, and the identification speed is improved while the computing power is saved. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 A flowchart of a cloud layer classification method based on meteorological satellite images is provided for the embodiments of the application. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are some of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the protection scope of the application.

[0058] The application provides a cloud layer classification method based on meteorological satellite images, which is suitable for a cloud layer classification system based on meteorological satellite images. Referring to Figure 1 , the method comprises steps S101-S106, and specifically comprises the following:

[0059] S101, determine a target area, and obtain at least two satellite cloud images of the target area in a monitoring period; wherein the two satellite cloud images are respectively a first image and a second image sorted according to acquisition time;

[0060] It can be understood that in the process of classifying the cloud type, the cloud in a region is identified. The region has a certain range, and the specific range is determined according to the actual size of the cloud, which is not limited in the embodiment. For the satellite cloud picture in the range, multiple images in a certain time can be obtained, and the cloud type is identified based on the multiple images. In the embodiment, two images are taken as an example, and in some other embodiments, the number of satellite cloud pictures can be greater than two or more, which is not limited in the embodiment.

[0061] S102, the first image is divided to form a plurality of first sub-images, and a plurality of first texture features corresponding to the plurality of first sub-images are obtained;

[0062] The texture feature of the cloud is one of the important indicators for judging the cloud type, so in the embodiment, the cloud type can be identified by extracting the texture feature. For how to identify the cloud texture feature, for example, the cloud feature and the distribution of the feature can be introduced into the trained neural network recognition model for identification, or other ways can be used, which is not limited in the application. When extracting the texture feature, the image can be converted to a grayscale image through grayscale processing, which simplifies the subsequent texture analysis. Common methods include weighted average method (such as OpenCV cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)). The specific method is not limited in the embodiment.

[0063] S103, the second image is divided to form a plurality of second sub-images, the plurality of second sub-images correspond to the plurality of first sub-images one by one, and a plurality of second texture features corresponding to the plurality of second sub-images are obtained;

[0064] It can be understood that in this step, it is the same as step S102, which is not described here. It should be noted that in some other embodiments, when the number of satellite cloud pictures in the monitoring period is multiple, the same processing is performed on each satellite cloud picture, and in the subsequent process of identifying by using the neural network recognition model, the recognition accuracy is higher than that of identifying the cloud type of only one image.

[0065] S104, each corresponding first texture feature and second texture feature are matched, and a matching degree corresponding to each second sub-image is obtained according to the matching result; the matching degree is used to represent the similarity between the first texture feature and the corresponding second texture feature;

[0066] It can be understood that in the embodiment, since two images are acquired, in the process of identifying by using the neural network image, in order to be able to combine the data of the two images, the two images can be compared, and the confidence degree is acquired to determine the data of the final input neural network identification model.

[0067] Specifically, in the embodiment, first, the first sub-image can be converted and segmented to acquire a plurality of first gray scale images, and a first average gray value of the first sub-image is acquired.

[0068] It can be understood that for each first sub-image, since the texture is different, the average gray in the corresponding first gray scale image is different, and when the texture changes, the first average gray also changes. The first sub-image is segmented in a grid or other manner, a plurality of first gray scale images can be obtained, and the number of first gray scale images is adjusted according to the ability of the processing device, which is not limited in the embodiment. When the texture changes, each first gray scale image also changes.

[0069] Similarly, the second sub-image also needs to be converted and segmented to acquire a plurality of second gray scale images, and a second average gray value corresponding to the second sub-image is acquired. The average difference value is determined based on the first average gray value and the second average gray value.

[0070] It can be understood that for the corresponding first sub-image and second sub-image, when the change is large, the first gray scale image and the second gray scale image corresponding to the change position have greater difference, and relatively, in the area where the texture changes little, the first gray scale image and the second gray scale image corresponding to the change position have higher similarity, that is, the change of gray is small. Therefore, each first gray scale image can be compared with the corresponding second gray scale image, and a plurality of target difference values are acquired. The actual size of the plurality of target difference values is used to measure the change degree of the position corresponding to the first gray scale image or the second gray scale image.

[0071] In order to determine the change degree, in the embodiment, the plurality of target difference values can be compared with the average difference value. If the target difference value is greater than the average difference value, the corresponding second gray scale image is marked as a target image. According to the proportional relationship between the number of target images and the number of a plurality of second gray scale images, the matching degree is determined.

[0072] Exemplarily, the following can be met:

[0073] P=K(N / n)*100%;

[0074] Wherein: P is the matching degree, K is a constant, N is the number of second gray scale images, and n is the number of target images.

[0075] In some other embodiments, it can also be determined whether it belongs to the target image by setting a preset value and comparing the preset value with the target difference value.

[0076] S105, divide the plurality of second sub-images into a same number of high-approval groups and low-approval groups according to the matching degrees; wherein the matching degrees corresponding to the second sub-images belonging to the high-approval groups are greater than the matching degrees corresponding to the second sub-images belonging to the low-approval groups;

[0077] It can be understood that in the above steps, the matching degrees of the plurality of second sub-images and the first sub-image are obtained, and the second sub-image with a higher matching degree also has a higher credibility. Therefore, when the data corresponding to the second sub-image with a higher matching degree is input into the neural network recognition model, it can be given a higher weight. Relatively, for the second sub-image with a lower matching degree, it can be given a lower weight.

[0078] In the embodiment, the same number means that the high-approval groups and the low-approval groups are divided by the intermediate value. For example, for parameters 1, 2, 3, and 4, in the case of being divided into two groups, 1 and 2 belong to the low-approval group, and 3 and 4 belong to the high-approval group.

[0079] In some other embodiments, the weight of the data corresponding to each second sub-image can also be directly determined according to the confidence.

[0080] S106, import each second texture feature and position information of each second sub-image into a neural network recognition model, and obtain classification data corresponding to the cloud layer type according to an output result of the neural network recognition model;

[0081] The weight of the second texture feature and the position information corresponding to the second sub-image belonging to the high-approval group when inputting into the neural network recognition model is greater than the weight of the second texture feature and the position information corresponding to the second sub-image belonging to the low-approval group when inputting into the neural network recognition model.

[0082] It can be understood that in the embodiment, the cloud layer features and the distribution of the features can be introduced into the trained neural network recognition model for recognition. When directly recognizing multiple images, the recognition speed is slow and a large amount of computing resources are occupied, and multiple data are obtained, and the credibility of each data cannot be determined. At the same time, the more images, the longer the processing time. Therefore, in the embodiment, before inputting the neural network recognition model, the confidence of each data inputting the neural network recognition model can be determined by the matching degree, that is, the corresponding weight, and in the process of using the neural network recognition model, only the determined data is input, and the processing speed is faster than that of inputting multiple image data.

[0083] For how to obtain the neural network recognition model, exemplarily, a training set including multiple classification data that have been determined, multiple training sub-image texture feature data corresponding to the classification data, and training sub-image position data can be obtained first, then an initial neural network recognition model is constructed, the initial neural network recognition model is trained based on the training set, and the training of the initial neural network recognition model is stopped when the initial neural network recognition model meets a preset condition and the trained neural network recognition model is output.

[0084] Specifically, the classification data, the training sub-image texture feature data, the longitudinal position data and the transverse position data in the training set can be introduced into the initial neural network recognition model as three different dimensions and iterative calculation is performed, wherein the three dimensions constitute a feature vector of the classification data.

[0085] A loss function is determined, an error value in the iterative calculation process is obtained based on the loss function, and the iterative calculation is stopped when the error value is less than a preset value and the neural network recognition model is output.

[0086] In some other embodiments, the trained neural network recognition model can also be directly obtained, or the model can be obtained in other ways disclosed in some existing technologies, which are not limited in the embodiment.

[0087] It can be understood that in the embodiment, when the second texture feature and the position information corresponding to the second sub-image are input into the neural network recognition model, the second texture feature and the position information corresponding to the second sub-image belonging to different groups are given different weights, and the following conditions are met:

[0088] a / b=P1 / P2;

[0089] Wherein, a is the weight of the second texture feature and the position information corresponding to the second sub-image belonging to the high recognition group when inputting the neural network recognition model, b is the weight of the second texture feature and the position information corresponding to the second sub-image belonging to the low recognition group when inputting the neural network recognition model, P1 is the matching degree corresponding to the second sub-image belonging to the high recognition group, and P2 is the matching degree corresponding to the second sub-image belonging to the low recognition group.

[0090] In some other embodiments, the weight of the data corresponding to each second sub-image can also be determined directly according to the confidence, and of course, the confidence and the weight need to satisfy a positive proportional relationship.

[0091] The cloud layer classification method based on meteorological satellite images provided in the embodiments of the present application first determines a target region, obtains at least two satellite cloud images of the target region in a monitoring period, then splits the first image to form a plurality of first sub-images, obtains a plurality of first texture features corresponding to the plurality of first sub-images, then splits the second image to form a plurality of second sub-images, the plurality of second sub-images correspond one-to-one to the plurality of first sub-images, and obtains a plurality of second texture features corresponding to the plurality of first sub-images, then matches each corresponding first texture feature and second texture feature, obtains a matching degree corresponding to each second sub-image according to the matching result, and determines the weight of the data corresponding to the second sub-image when inputting the neural network recognition model according to the corresponding matching degree. The cloud layer classification method based on meteorological satellite images provided in the present application compares the satellite cloud images in the monitoring period first, determines the weight of each sub-image inputting the neural network model according to the matching degree of each sub-image, the weight of the sub-image with large variation is low, and the influence degree on the actual result is small. The neural network model does not need to identify each satellite cloud image, which improves the identification speed and saves the computing power.

[0092] Based on the same inventive concept, the present application provides a cloud layer classification system based on meteorological satellite images, which is configured to:

[0093] Determine a target region, and obtain at least two satellite cloud images of the target region in a monitoring period; wherein the two satellite cloud images are a first image and a second image sorted according to the acquisition time, respectively;

[0094] Split the first image to form a plurality of first sub-images, and obtain a plurality of first texture features corresponding to the plurality of first sub-images;

[0095] cutting the second image to form a plurality of second sub-images, the plurality of second sub-images corresponding to the plurality of first sub-images one by one, and obtaining a plurality of second texture features corresponding to the plurality of second sub-images;

[0096] matching each of the first texture features with the second texture features, and obtaining a matching degree corresponding to each of the second sub-images according to a matching result;

[0097] dividing the plurality of second sub-images into a same number of high-approval groups and low-approval groups according to the matching degrees, wherein the matching degree corresponding to the second sub-image belonging to the high-approval group is greater than the matching degree corresponding to the second sub-image belonging to the low-approval group;

[0098] inputting each of the second texture features and position information of each of the second sub-images into a neural network recognition model, and obtaining classification data corresponding to a cloud layer type according to an output result of the neural network recognition model;

[0099] wherein the weight of the second texture features and the position information corresponding to the second sub-image belonging to the high-approval group is greater than the weight of the second texture features and the position information corresponding to the second sub-image belonging to the low-approval group when inputting into the neural network recognition model.

[0100] Optionally, the system is configured to:

[0101] the inputting each of the second texture features and position information of each of the second sub-images into a neural network recognition model, and obtaining classification data corresponding to a cloud layer type according to an output result of the neural network recognition model, specifically includes:

[0102] obtaining a training set, the training set including a plurality of classification data that has been determined, a plurality of training sub-image texture feature data corresponding to the classification data, and training sub-image position data; the training sub-image position data including longitudinal position data and transverse position data;

[0103] constructing an initial neural network recognition model, training the initial neural network recognition model based on the training set, stopping training when the initial neural network recognition model meets a preset condition, and outputting a trained neural network recognition model.

[0104] Optionally, the system is configured to:

[0105] The initial neural network recognition model is constructed, the initial neural network recognition model is trained based on the training set, the training of the initial neural network recognition model is stopped when the initial neural network recognition model meets a preset condition, and a trained neural network recognition model is output. Specifically, the method comprises the following steps:

[0106] The classification data, the training sub-image texture feature data, the longitudinal position data and the transverse position data in the training set are taken as three different dimensions, and are imported into the initial neural network recognition model and iteratively calculated; wherein the three dimensions constitute a feature vector of the classification data;

[0107] A loss function is determined, an error value in the iterative calculation process is obtained based on the loss function, and the iterative calculation is stopped and the neural network recognition model is output when the error value is less than a preset value.

[0108] Optionally, the system is configured to:

[0109] The second texture feature of each second sub-image and the position information of each second sub-image are imported into the neural network recognition model, and the classification data corresponding to the cloud layer type is obtained according to the output result of the neural network recognition model. Specifically, the method comprises the following steps:

[0110] When the second texture feature corresponding to the second sub-image and the position information are input into the neural network recognition model, different weights are given to the second texture feature corresponding to the second sub-image and the position information belonging to different groups, and the following conditions are met:

[0111] a / b=P1 / P2;

[0112] Wherein, a is the weight of the second texture feature corresponding to the second sub-image and the position information belonging to the high recognition group when inputting the neural network recognition model, b is the weight of the second texture feature corresponding to the second sub-image and the position information belonging to the low recognition group when inputting the neural network recognition model, P1 is the matching degree corresponding to the second sub-image belonging to the high recognition group, and P2 is the matching degree corresponding to the second sub-image belonging to the low recognition group.

[0113] Optionally, the system is configured to:

[0114] The first texture feature and the second texture feature corresponding to each second sub-image are matched, and the matching degree corresponding to each second sub-image is obtained according to the matching result; the matching degree is used to represent the similarity between the first texture feature and the corresponding second texture feature. Specifically, the method comprises the following steps:

[0115] convert and segment the first sub-image to obtain a plurality of first gray scale images and obtain a first average gray value of the first sub-image;

[0116] convert and segment the second sub-image to obtain a plurality of second gray scale images and obtain a second average gray value corresponding to the second sub-image;

[0117] determine an average difference value based on the first average gray value and the second average gray value;

[0118] compare each of the first gray scale images with the corresponding second gray scale image and obtain a plurality of target difference values corresponding thereto;

[0119] compare the plurality of target difference values with the average difference value, and if the target difference value is greater than the average difference value, mark the corresponding second gray scale image as a target image;

[0120] determine the matching degree according to a proportional relationship between the number of target images and the number of the plurality of second gray scale images.

[0121] Optionally, the system is configured to:

[0122] determine the matching degree according to a proportional relationship between the number of target images and the number of the plurality of second gray scale images, specifically comprising:

[0123] P = K (N / n) * 100%;

[0124] wherein P is the matching degree, K is a constant, N is the number of second gray scale images, and n is the number of target images.

[0125] The cloud layer classification system based on meteorological satellite images provided in the embodiments of the present application first determines a target region, acquires at least two satellite cloud images of the target region in a monitoring period, then cuts the first image to form a plurality of first sub-images, and acquires a plurality of first texture features corresponding to the plurality of first sub-images, then cuts the second image to form a plurality of second sub-images, the plurality of second sub-images correspond one-to-one to the plurality of first sub-images, and acquires a plurality of second texture features corresponding to the plurality of first sub-images, then matches each corresponding first texture feature and second texture feature, acquires a matching degree corresponding to each second sub-image according to the matching result, and then determines a weight when the second sub-image corresponding data is input into a neural network recognition model according to the corresponding matching degree. The cloud layer classification system based on meteorological satellite images provided in the present application compares the satellite cloud images in the monitoring period first, determines the weight of each sub-image input into the neural network model according to the matching degree of each sub-image, and the weight of the sub-image with large change is low, and the influence degree on the actual result is small. The neural network model does not need to recognize each satellite cloud image, which improves the recognition speed and saves the computing power.

[0126] Based on the same inventive concept, the embodiments of the present application also provide an electronic device, which comprises:

[0127] at least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the cloud layer classification method based on meteorological satellite images according to the embodiments of the present application.

[0128] In addition, to achieve the above-mentioned purpose, the embodiments of the present application also provide a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the cloud layer classification method based on meteorological satellite images according to the embodiments of the present application.

[0129] The various constituent components of the electronic device will be specifically introduced as follows:

[0130] The processor is the control center of the electronic device, and can be one processor or a combination of multiple processing elements. For example, the processor is one or more central processing units (CPUs), application specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present application, such as one or more microprocessors (digital signal processors (DSPs)), or one or more field programmable gate arrays (FPGAs).

[0131] Optionally, the processor can execute various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0132] The memory is used to store software programs for implementing the solutions of the present application, and is controlled by the processor to execute. The specific implementation can refer to the above method embodiments, and will not be described here.

[0133] Optionally, the memory can be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, an optical disc storage (including a compact disc, a laser disc, an optical disc, a digital versatile disc, a Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program codes in the form of instructions or data structures and accessible by a computer, but not limited to. The memory can be integrated with the processor or exist independently and coupled with the processor through the interface circuit of the electronic device, and the embodiments of the present application do not make specific limitations in this regard.

[0134] The transceiver is used to communicate with the network device or the terminal device.

[0135] Optionally, the transceiver can include a receiver and a transmitter. The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0136] Optionally, the transceiver can be integrated with the processor, or exist independently and be coupled with the processor through the interface circuit of the router, and the embodiments of the present application do not make specific limitation hereon.

[0137] In addition, the technical effects of the electronic device can refer to the technical effects of the data transmission method of the above-mentioned method embodiments, which will not be described here again.

[0138] It should be understood that the processor in the embodiments of the present application can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0139] It should also be understood that the memory in the embodiments of the present application can be volatile or nonvolatile memory, or can include both volatile and nonvolatile memory. The nonvolatile memory can be read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), electrically EPROM (EEPROM), or flash memory. The volatile memory can be random access memory (RAM) used as external cache. By way of example, and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0140] The above-described embodiments can be implemented in part or in whole through software, hardware (e.g., circuitry), firmware, or any combination thereof. When implemented in software, the above-described embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When loaded and executed by a computer, the computer instructions or computer programs can generate the processes or functions according to the embodiments of the present application, in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable apparatus. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, such as from a website site, a computer, a server, or a data center to another website site, a computer, a server, or a data center, via a wired (e.g., infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0141] It should be understood that the term "and / or" used herein is merely an association relationship between the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone, where A and B can be singular or plural. In addition, the character " / " herein generally represents an "or" relationship between the associated objects before and after it, but it can also represent an "and / or" relationship, which can be understood in the context before and after it.

[0142] In the present application, "at least one" means one or more, and "multiple" means two or more. "At least one of the following" or the like means any combination of the items, including any combination of single or multiple items. For example, at least one of a, b, or c can mean a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, and c can be single or multiple.

[0143] It should be understood that in various embodiments of the present application, the size of the sequence number of the above-described processes does not mean the order of execution, and the execution order of the processes should be determined by their functions and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0144] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on specific applications and design constraints. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

Claims

1. A cloud classification method based on meteorological satellite images, characterized in that, The method comprises: determining a target area, and acquiring at least two satellite cloud images of the target area in a monitoring period; the two satellite cloud images are respectively a first image and a second image sorted by acquisition time; cutting the first image to form a plurality of first sub-images, and acquiring a plurality of first texture features corresponding to the plurality of first sub-images; cutting the second image to form a plurality of second sub-images, the plurality of second sub-images corresponding one-to-one to the plurality of first sub-images, and acquiring a plurality of second texture features corresponding to the plurality of second sub-images; matching each of the first texture features and the second texture features, and acquiring a matching degree corresponding to each of the second sub-images according to a matching result; the matching degree is used to represent a similarity degree of the first texture features and the corresponding second texture features; dividing the plurality of second sub-images into a same number of high-approval groups and low-approval groups according to the matching degrees; the matching degree corresponding to the second sub-image belonging to the high-approval group is greater than the matching degree corresponding to the second sub-image belonging to the low-approval group; inputting each of the second texture features and position information of each of the second sub-images into a neural network recognition model, and acquiring classification data corresponding to a cloud layer type according to an output result of the neural network recognition model; the weight of the second texture features and the position information corresponding to the second sub-image belonging to the high-approval group is greater than the weight of the second texture features and the position information corresponding to the second sub-image belonging to the low-approval group when the second texture features and the position information are input into the neural network recognition model.

2. The cloud layer classification method based on meteorological satellite images according to claim 1, characterized in that: the step of inputting each of the second texture features and position information of each of the second sub-images into a neural network recognition model, and acquiring classification data corresponding to a cloud layer type according to an output result of the neural network recognition model, specifically comprises: acquiring a training set, the training set comprising a plurality of the classification data that have been determined, a plurality of training sub-image texture feature data corresponding to the classification data, and training sub-image position data; the training sub-image position data comprises longitudinal position data and transverse position data; constructing an initial neural network recognition model, training the initial neural network recognition model based on the training set, stopping training when the initial neural network recognition model meets a preset condition, and outputting a trained neural network recognition model.

3. The cloud layer classification method based on meteorological satellite images according to claim 2, characterized in that: the step of constructing an initial neural network recognition model, training the initial neural network recognition model based on the training set, stopping training when the initial neural network recognition model meets a preset condition, and outputting a trained neural network recognition model, specifically comprises: The classification data, training sub-image texture feature data, longitudinal position data and transverse position data in the training set are imported into the initial neural network recognition model as three different dimensions and iterative calculation is performed; wherein the three dimensions constitute a feature vector of the classification data; A loss function is determined, an error value in the iterative calculation process is obtained based on the loss function, and the iterative calculation is stopped and the neural network recognition model is output when the error value is less than a preset value.

4. The cloud classification method based on meteorological satellite images according to claim 3, characterized in that: The second texture feature of each second sub-image and the position information of each second sub-image are input into the neural network recognition model, and the classification data corresponding to the cloud type is obtained according to the output result of the neural network recognition model, specifically comprising: When the second texture feature and the position information of the second sub-image are input into the neural network recognition model, different weights are given to the second texture feature and the position information of the second sub-image belonging to different groups, and the following conditions are met: a / b=P1 / P2; Wherein a is the weight of the second texture feature and the position information of the second sub-image belonging to the high recognition group when input into the neural network recognition model, b is the weight of the second texture feature and the position information of the second sub-image belonging to the low recognition group when input into the neural network recognition model, P1 is the matching degree of the second sub-image belonging to the high recognition group, and P2 is the matching degree of the second sub-image belonging to the low recognition group.

5. The cloud classification method based on meteorological satellite images according to claim 1, characterized in that: The first texture feature and the second texture feature corresponding to each second sub-image are matched, and the matching degree corresponding to each second sub-image is obtained according to the matching result; The matching degree is used to represent the similarity between the first texture feature and the corresponding second texture feature, specifically comprising: The first sub-image is converted and segmented to obtain a plurality of first gray scale images, and a first average gray value of the first sub-image is obtained; The second sub-image is converted and segmented to obtain a plurality of second gray scale images, and a second average gray value corresponding to the second sub-image is obtained; Based on the first average gray value and the second average gray value, an average difference value is determined; Each first gray scale image is compared with the corresponding second gray scale image, and a plurality of target difference values are obtained; The plurality of target difference values are compared with the average difference value, and if the target difference value is greater than the average difference value, the corresponding second gray scale image is marked as a target image; The matching degree is determined according to the proportional relationship between the number of target images and the number of second gray scale images.

6. The cloud classification method based on meteorological satellite images according to claim 5, characterized in that: The matching degree is determined according to a proportional relationship between the target image quantity and the quantity of the plurality of second gray scale images, and specifically includes: P = K (N / n) * 100%; Wherein: P is the matching degree, K is a constant, N is the quantity of the second gray scale image, and n is the target image quantity.

7. A cloud classification system based on meteorological satellite images, characterized in that, The system is configured to: Determine a target region, and obtain at least two satellite cloud images of the target region in a monitoring period; the two satellite cloud images are respectively a first image and a second image sorted according to acquisition time; The first image is cut to form a plurality of first sub-images, and a plurality of first texture features corresponding to the plurality of first sub-images are obtained; The second image is cut to form a plurality of second sub-images, the plurality of second sub-images correspond one-to-one to the plurality of first sub-images, and a plurality of second texture features corresponding to the plurality of second sub-images are obtained; Each of the first texture features and the second texture features is matched, and a matching degree corresponding to each of the second sub-images is obtained according to the matching result; the matching degree is used to represent the similarity of the first texture features and the corresponding second texture features; According to the matching degree, the plurality of second sub-images are divided into a plurality of high-approval groups and a plurality of low-approval groups with the same quantity; the matching degree corresponding to the second sub-image belonging to the high-approval group is greater than the matching degree corresponding to the second sub-image belonging to the low-approval group; Each of the second texture features and the position information of each of the second sub-images is input into a neural network recognition model, and classification data corresponding to a cloud layer type is obtained according to an output result of the neural network recognition model; The weight of the second texture features and the position information corresponding to the second sub-image belonging to the high-approval group when inputting the neural network recognition model is greater than the weight of the second texture features and the position information corresponding to the second sub-image belonging to the low-approval group when inputting the neural network recognition model.

8. The cloud classification system based on meteorological satellite images according to claim 7, characterized in that, The system is configured to: Each of the first texture features and the second texture features is matched, and a matching degree corresponding to each of the second sub-images is obtained according to the matching result; The matching degree is used to represent the similarity of the first texture features and the corresponding second texture features, and specifically includes: The first sub-image is converted and segmented to obtain a plurality of first gray scale images, and a first average gray scale value of the first sub-image is obtained; The second sub-image is converted and segmented to obtain a plurality of second gray scale images, and a second average gray scale value corresponding to the second sub-image is obtained; An average difference value is determined based on the first average gray scale value and the second average gray scale value; Each of the first gray scale images is compared with the corresponding second gray scale image, and a plurality of target difference values are obtained; The plurality of target difference values are compared with the average difference value, and if the target difference value is greater than the average difference value, the corresponding second gray scale image is marked as a target image; According to a proportional relationship between the target image quantity and a quantity of the second gray scale images, the matching degree is determined.

9. An electronic device, comprising: The method comprises: at least one processor, and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method as claimed in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, A computer program is stored thereon, and the program is executed by a processor to implement the method as claimed in any one of claims 1-6.

Images