Multi-temporal cloud picture prediction method and prediction model based on multi-channel satellite data

By building a multi-channel multi-time phase cloud map prediction network and processing multi-channel satellite data, the crosstalk effect problem in multi-channel cloud map prediction is solved, and accurate cloud map prediction under multi-channel data is realized, which is suitable for long-term series prediction and disaster monitoring.

CN118941977AActive Publication Date: 2024-11-12NANJING UNIV OF INFORMATION SCI & TECH

Patent Information

Application Number
CN202410944032.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-15
Publication Date
2024-11-12
Estimated Expiration
2044-07-15

AI Technical Summary

Technical Problem

The existing cloud map prediction methods have crosstalk effects in multi-channel data processing, resulting in inaccurate prediction results, and traditional methods are difficult to effectively simulate the nonlinear motion of cloud clusters, resulting in problems such as fuzzy prediction and short time.

Method used

Using the multi-time phase cloud map prediction method based on multi-channel satellite data, by downloading satellite cloud maps of different bands, performing splicing and bright temperature calculations, using logarithmic normalization processing, and building a multi-channel multi-time phase cloud map prediction network, including selection attention Force module, iterative downsampling module, iterative upsampling module and channel separation output module, learn multi-scale depth features and cloud cluster structure in the image, and generate accurate multi-channel satellite cloud map prediction results.

Benefits of technology

It effectively alleviates the crosstalk problem of multi-channel data during deep learning network layer processing, maintains the prediction accuracy of long-term series, generates more realistic satellite cloud map prediction results, and can better predict cloud group movement trends and precipitation. Suitable for disaster monitoring and other fields.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118941977A_ABST
    Figure CN118941977A_ABST
Patent Text Reader

Abstract

The invention discloses a multi-temporal cloud picture prediction method and prediction model based on multi-channel satellite data. The method specifically comprises the following steps: step 1, downloading satellite cloud pictures of different wavebands; 2, splicing the satellite cloud pictures of different wavebands to form a multi-channel satellite cloud picture; 3, calculating a brightness temperature value of each pixel in each satellite cloud picture, and mapping the satellite cloud picture to a brightness temperature value form; 4, performing logarithmic normalization processing on the brightness temperature value in each satellite cloud picture; 5, constructing a multi-channel multi-temporal cloud picture prediction network, and training the multi-channel multi-temporal cloud picture prediction network by taking the normalized multi-channel satellite cloud picture in the step 4 as a training set to obtain a trained multi-channel multi-temporal cloud picture prediction network; and 6, performing prediction by adopting the trained multi-channel multi-temporal cloud picture prediction network. The satellite cloud picture prediction result generated by the method is more real.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of meteorological processing, and in particular relates to a multi-temporal cloud image prediction method and a prediction model based on multi-channel satellite data. Background Art

[0002] In modern society, there are many ways to obtain meteorological data, such as ground meteorological observation stations, radars, and geosynchronous satellites. However, the construction of ground meteorological observation stations is limited by terrain conditions, and radars have a small coverage area and are easily interfered by electromagnetic waves. Only geosynchronous satellites have the advantages of large coverage area, no spatial restrictions, and high temporal resolution. Using cloud images taken by geosynchronous satellites to predict cloud images in future time periods is of great significance to people's research on weather changes, monitoring of disaster weather, etc.

[0003] Most traditional cloud prediction methods are based on block matching and optical flow. The block matching method effectively utilizes the motion information of clouds, but as a linear model, it cannot accurately simulate the nonlinear motion of clouds; the optical flow method calculates the motion speed through the change of pixel points, realizing nonlinear simulation, but the grayscale of the object will also change with the change of time, resulting in inaccurate calculation results. Modern people have begun to apply deep learning methods to cloud prediction, and have achieved better results than traditional methods, but there are still problems such as fuzzy prediction and short prediction time. Especially in multi-channel cloud prediction, channel crosstalk effects often occur, affecting the model to generate accurate predictions. Summary of the invention

[0004] Purpose of the invention: In order to solve the problems existing in the above-mentioned prior art, the present invention provides a multi-temporal cloud image prediction method and prediction model based on multi-channel satellite data.

[0005] Technical solution: The present invention discloses a multi-temporal cloud image prediction method based on multi-channel satellite data, comprising the following steps:

[0006] Step 1: Download satellite cloud images of different bands;

[0007] Step 2: Splice satellite cloud images of different bands to form a multi-channel satellite cloud image;

[0008] Step 3: Calculate the brightness temperature value of each pixel in each satellite cloud image, thereby mapping the satellite cloud image into the form of brightness temperature value;

[0009] Step 4: Logarithmically normalize the brightness temperature value in each satellite cloud image;

[0010] Step 5: construct a multi-channel multi-temporal cloud image prediction network, and use the multi-channel satellite cloud image normalized in step 4 as a training set to train the multi-channel multi-temporal cloud image prediction network to obtain a trained multi-channel multi-temporal cloud image prediction network;

[0011] Step 6: Use the trained multi-channel multi-phase cloud image prediction network for prediction.

[0012] Furthermore, in step 3, the brightness temperature value of each pixel in each satellite cloud image is calculated using the following formula:

[0013]

[0014] Among them, T b Represents the brightness temperature value of the bth pixel in the satellite cloud image, Image b represents the pixel value between the satellite and the bth pixel in the image, f1 and f2 represent the coefficients of the Planck function, b1 and b2 are the spectral response function offset and scale correction terms, respectively.

[0015] Furthermore, in step 4, a logarithmic function is used for normalization, and the expression is as follows:

[0016]

[0017] in, represents the normalized brightness temperature value of the bth pixel in the satellite cloud image, I min Indicates the minimum brightness temperature, T b It represents the brightness temperature of the bth pixel in the satellite cloud image, and s' is the scale size.

[0018] Furthermore, the multi-channel multi-phase cloud map prediction network includes a selective attention module, an iterative downsampling module, an iterative upsampling module and a channel separation output module;

[0019] The selective attention module includes a first stage and a second stage, wherein the first stage is used to extract multi-scale deep features of the original input image, and the multi-scale deep features are input into the second stage to establish correlations between features of different scales to obtain channel enhancement features, and then the channel enhancement features are convolved to the same size as the input feature size to obtain spatial channel enhancement features, and the spatial channel enhancement features are input into an iterative downsampling module to gradually extract long-term and short-term time information contained in the spatial channel enhancement features, while learning the complex structure and motion pattern of the cloud clusters implied in the image; then the output features of the iterative downsampling module are input into an iterative upsampling module to restore the detailed features of the image, and the output features of the iterative upsampling module are input into a channel separation module, and the channel separation module uses C groups of independent filters to predict a single-channel satellite cloud image, where C is the number of channels of the original input image input into the multi-channel multi-phase cloud image prediction network, and then the predicted single-channel cloud image is spliced ​​in the channel dimension to obtain a predicted multi-channel satellite cloud image.

[0020] Furthermore, the first stage of the selective attention module includes n pooling layers, n upsampling modules, n element-by-element multiplication modules and a splicing module. The i-th pooling layer is connected to the i-th upsampling module. The output of the i-th upsampling module is multiplied by the original image input to the selective attention module through the i-th element-by-element multiplication module. The outputs of all element-by-element multiplication modules are spliced ​​through the splicing module to obtain multi-scale deep features; the second stage includes the first to fourth reshaping modules, the first and second matrix multiplication modules, the softmax module, the addition module and the first convolution module. The first to third reshaping modules reshape the feature map input to the second stage into feature maps of different dimensions, multiply the features output by the second and third reshaping modules through the first matrix multiplication module and input them into the softmax module. The softmax module multiplies the output features with the output features of the first reshaping module through the second matrix multiplication module and inputs them into the fourth reshaping module. The features output by the fourth reshaping module are added to the features input to the second stage to obtain channel enhancement features, and the obtained channel enhancement features are input to the first convolution module to obtain spatial channel enhancement features.

[0021] Furthermore, the structures of the C groups of filters in the channel separation output module are the same, and all include a second convolution module, a first deconvolution module, a third convolution module and a fourth convolution module connected in sequence.

[0022] Furthermore, the loss function of the multi-channel multi-phase cloud image prediction network is As shown below:

[0023]

[0024] Among them, Y represents the real multi-channel satellite cloud image, represents the predicted value of Y, i represents the i-th sample, j represents the j-th frame, N represents the batch size, and S represents the total number of frames of the cloud map in a sample.

[0025] A prediction model for multi-temporal cloud images based on multi-channel satellite data, comprising: a data acquisition module, a data processing module, a model building module and a model training module;

[0026] The data acquisition module is used to download satellite cloud images of different bands;

[0027] The data processing module is used to execute steps 2 to 4;

[0028] The model building module is used to build a multi-channel and multi-phase cloud image prediction network;

[0029] The model training module is used to train the constructed multi-channel and multi-phase cloud image prediction network.

[0030] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of a multi-temporal cloud image prediction method based on multi-channel satellite data are implemented.

[0031] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, a multi-temporal cloud image prediction method based on multi-channel satellite data is executed.

[0032] Beneficial effects: The present invention adopts a channel separation output module to generate multi-channel satellite cloud image prediction results, which can effectively alleviate the crosstalk problem caused by different channel data when processed in the deep learning network layer, and uses a selective attention module to learn the interdependence of multi-scale deep features in both spatial and channel dimensions, fully integrating the spatial features provided by different channels, so that the model can still maintain accurate prediction capabilities in the prediction of long time series, and generate more realistic satellite cloud image prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flow chart of the present invention.

[0034] Figure 2 This is a network structure diagram for predicting multi-channel and multi-phase cloud images of the present invention.

[0035] Figure 3 The figure is a comparison diagram between the predicted satellite cloud image sequence and the real satellite cloud image sequence obtained by the method of the present invention.

[0036] Figure 4 It is a structural schematic diagram of an attention-based multi-channel satellite cloud image prediction device of the present invention.

[0037] Figure 5 The present invention is a hardware structure diagram of an attention-based multi-channel satellite cloud image prediction device. DETAILED DESCRIPTION

[0038] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0039] Figure 1 A flowchart of a multi-channel satellite cloud image prediction method based on attention is provided in this embodiment, such as Figure 1 As shown, this embodiment specifically includes the following steps:

[0040] Step S1, download satellite cloud images of different bands. In this embodiment, GOES-16ABI L1b Radiance Data is downloaded from the NOAA website. The bands are C08 (6.17 μm), C09 (6.93 μm), C10 (7.34 μm), and C13 (10.33 μm).

[0041] Satellite cloud images contain a variety of information closely related to atmospheric precipitation, such as cloud height and thickness, cloud top temperature, water vapor content, etc., which can help meteorologists estimate the precipitation in the target area. The satellite cloud image prediction problem can be expressed as follows: Assume that the input sequence X = (X t-s ,…,X t-1 ,X t ), the sequence length is s frames, X t ∈R w×h×c is the tth frame in the cloud image sequence, where w, h and c are the width, height and number of channels respectively. The satellite cloud image prediction problem can be defined as deriving the next future frame sequence with a time length of m from the input sequence X. This embodiment selects GOES-16 of the GOES series of geostationary environmental observation satellites for fixed-point observation of the continental United States as the observation satellite, uses the reflectivity signal of the satellite's advanced reference imager as the data source, and selects the water vapor band (C08, C09, C13) and infrared band (C13) that are strongly correlated with precipitation for accurate prediction, which can further effectively estimate precipitation and better respond to disastrous weather.

[0042] Step S2, refer to the GOES-R satellite product manual document to read the variables, parameters and other information contained in the NetCDF format data.

[0043] The downloaded data source is stored in the network universal data format NetCDF, which is widely used to store data in meteorological science. NetCDF provides different interfaces for different development languages. This embodiment uses the netCDF4 library in Python3 and reads the variables, parameters and other information contained in the metadata with reference to the GOES-16 satellite product manual.

[0044] Step S3: Use the Planck function and the band passband to correct the reflectance data to obtain the brightness temperature data of the corresponding band.

[0045] The data downloaded from the website is the Radiance data of L1b of GOES-16 / ABI. In order to obtain the Brightness Temperature (T b ), the brightness temperature value of the corresponding band can only be obtained after using the Planck function and band pass correction. The calculation formula is as follows:

[0046]

[0047] Among them, T b Represents the brightness temperature value of the bth pixel in the satellite cloud image, Image b represents the pixel value of the satellite and the bth pixel in the image, f1 and f2 represent the coefficients of the Planck function, which can be obtained by reading the variables "planck_fk1" and "planck_fk2" in the NC file in step 2. b1 and b2 are the spectral response function offset and scale correction terms, which can be obtained by reading the variables "planck_bc1" and "planck_bc2" in the NC file in step 2. b1 and b2 are the spectral response function offset and scale correction terms.

[0048] Step S4: Perform data normalization using a normalization method based on a logarithmic function.

[0049] In the training process of deep learning, each input feature often has different dimensions and dimensional units. If it is input into the model for training without processing, it will affect the model's data analysis results. In order to eliminate the influence of singular samples, normalization processing is often performed to limit the input features to the range of 0 to 1. Since this embodiment uses two different bands of water vapor and infrared, there are large differences between the bands, so it is necessary to preserve the differences between the bands during the normalization process. Therefore, this embodiment uses a normalization method based on a logarithmic function. The specific calculation formula is as follows:

[0050]

[0051] in, represents the normalized brightness temperature value of the bth pixel in the satellite cloud image, I min Indicates the minimum brightness temperature, T b represents the brightness temperature value of the bth pixel in the satellite cloud image, and s' is the scaling size. Step S5: construct a multi-channel and multi-phase cloud image prediction network, and use the data that has undergone the above preprocessing operations to train the network.

[0052] Multi-channel multi-temporal cloud image prediction network Figure 2 As shown, it includes a selection attention module, an iterative downsampling module, an iterative upsampling module and a channel separation output module.

[0053] In the existing research on using deep learning models to predict satellite cloud images, single-channel satellite cloud images are generally used, that is, only the shooting data of one band of the satellite is included, but the sensors carried by the satellite can often capture meteorological data of multiple bands at the same time. The data of different bands can both verify and complement each other. Therefore, when predicting satellite cloud images, multi-channel satellite cloud images from multiple bands are used, and the single-channel satellite cloud images are fused with other channel information. The other channel images and the single-channel (band) images are spliced ​​in the channel dimension to obtain more accurate prediction results. However, if the satellite cloud image data of multiple channels are processed at the same time, the signal of one channel may leak or couple to other channels, and serious crosstalk problems may occur, which not only fails to achieve the effect of improving the prediction accuracy, but also may affect the satellite cloud image prediction of the single channel. Therefore, this embodiment builds a selective attention module. In the first stage, the input multi-channel cloud image is pooled at different spatial scales to obtain a set of pooled features with different spatial resolutions, and then the pooled features are resampled using upsampling operations at the same resolution. After synthesis in the channel dimension, multi-scale deep features that retain more useful information are obtained, and the second stage is entered. Each channel map input to the second stage can be regarded as a mapping of a specific scale, and there is correlation between feature maps of different scales. In order to utilize this correlation, the input features of the second stage are reshaped into different shapes (RESHAPE1, RESHAPE 2, and RESHAPE 3 modules transform the dimensions of the input data into S*4C*HW, S*4C*HW, and S*HW*4C, respectively), and matrix multiplication and softmax operations are performed between them to establish the correlation between features of different scales. Finally, after being reshaped to the input size, they are added to the original input features of the second stage to obtain channel enhancement features. At this time, each channel in the feature is the weighted sum of all channels, which can simulate the long-distance dependency between multi-scale feature maps. Finally, the channel enhancement features are convolved to the same size as the input feature size to obtain spatial channel enhancement features.

[0054] The spatial channel enhanced features enter the iterative downsampling module composed of alternating 3D convolutional layers and 3D-SwinTransformer layers, gradually extracting the long-term and short-term time information contained in the multi-channel cloud image sequence, while learning the complex structure and motion pattern of the clouds hidden in the image; then enter the iterative upsampling module composed of 3D transposed convolutional layers to restore the detailed features of the image.

[0055] If the prediction channel is generated directly through the 3D transposed convolution layer, that is, all output channels are generated by the same filter, then each conversion will directly consider the cross-channel information, disrupting the model's prediction. Therefore, a channel separation module is added to use C groups of independent filters to generate the predicted single-channel cloud map, so that the cross-channel information is only used in the first conversion, and then becomes indirect information, gradually reducing side effects, and finally splicing the generated single-channel image in the channel dimension. C is the number of channels of the original input image input to the multi-channel multi-phase cloud map prediction network; the filter includes a layer of convolution module, a layer of deconvolution and two layers of convolution.

[0056] When the model predicts time series data, the previous data has the most historical information to refer to. As the prediction time increases, the amount of reference information gradually decreases and the prediction error increases, so a clear preceding frame and a blurred subsequent frame are generated. To combat this characteristic, when calculating the error between the predicted image and the generated image, a logarithmic function is used to dynamically increase the weight of the subsequent frame in the total loss, forcing the model to pay attention to the prediction of the subsequent frame, forming a time-skewed loss, and improving the model's full-sequence prediction accuracy.

[0057] This implementation inputs the data that has undergone the above preprocessing operations to train the multi-channel multi-phase cloud map prediction network, and uses the time tilt loss function to guide the model to generate a cloud map sequence that is more similar to the real cloud map sequence. The calculation formula is:

[0058]

[0059] Among them, Y represents the real multi-channel satellite cloud image, Represents the multi-channel satellite cloud image predicted by the model, N represents the batch size, S represents the number of frames of a sample, u and j represent the i-th sample and j-th frame respectively.

[0060] Step S6: Use the optimal parameters obtained through training to encapsulate and create a multi-channel and multi-phase cloud map prediction model.

[0061] In this embodiment, the ADAM optimizer is selected to use the above-mentioned loss function to iterate 100 epochs to train the above-mentioned multi-channel multi-phase cloud map prediction network, and the parameters that obtain the lowest loss on the validation set are selected as the final model parameters, which are packaged into a multi-channel multi-phase cloud map prediction model that can be directly used.

[0062] Step S7: Use the multi-channel multi-temporal cloud image prediction model to complete the real-time multi-channel satellite cloud image prediction.

[0063] The real-time multi-channel satellite cloud image is subjected to the above-mentioned preprocessing operation and input into the multi-channel multi-temporal cloud image prediction model to obtain the predicted multi-channel satellite cloud image.

[0064] Depend on Figure 3 It can be seen that the present invention can more accurately predict the long-term satellite cloud image movement trend and cloud changes, providing important references for precipitation estimation, disaster monitoring and other fields.

[0065] Figure 4 A functional module diagram of a multi-temporal cloud image prediction device based on multi-channel satellite data provided by an embodiment of the present invention, the device mainly includes: an acquisition module 10, a preprocessing module 20, a model training module 30 and a prediction module 40, wherein:

[0066] Acquisition module 10 is used to download GOES-16ABI L1b Radiance Data from the NOAA website. The bands are C08 (6.17μm), C09 (6.93μm), C10 (7.34μm), and C13 (10.33μm).

[0067] The preprocessing module 20 is used to preprocess the metadata, including reading the variables, parameters and other information contained in the NetCDF format data with reference to the GOES-R satellite product manual document, using the Planck function and the band passband to correct the reflectivity data to obtain the corresponding band brightness temperature data, and using the normalization method based on the logarithmic function to normalize the data;

[0068] A model training module 30 is used to train a multi-channel multi-phase cloud map prediction network using the data that has undergone the aforementioned preprocessing operation, and encapsulate a multi-channel multi-phase cloud map prediction model using the optimal parameters obtained through the training;

[0069] The prediction module 40 is used to complete the real-time multi-channel satellite cloud image prediction using the multi-channel multi-temporal cloud image prediction model.

[0070] Figure 5 A schematic diagram of the hardware structure of a multi-temporal cloud image prediction device based on multi-channel satellite data provided in an embodiment of the present invention, the electronic device comprises: a processor 60, a memory 61, a bus 62 and a communication interface 63, wherein the processor 60, the communication interface 63 and the memory 61 are connected via the bus 62; the processor 60 is used to execute an executable module stored in the memory 61, such as a computer program.

[0071] The memory 61 includes a high-speed random access memory (RAM), a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 63 (which can be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. can be used. The bus 62 can be an ISA bus, a PCI bus or an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0072] Among them, the memory 61 is used to store the program, and the processor 60 executes the program after receiving the execution instruction to implement a multi-phase cloud image prediction method based on multi-channel satellite data in this embodiment. The processor 60 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit in the processor 60 or the instruction in the form of software. The above processor 60 can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or product-based logic devices, discrete hardware components, which can implement or execute the disclosed methods, steps and logic block diagrams in this embodiment. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiment of the present invention can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 61, and the processor 60 reads the information in the memory 61 and completes the steps of the above method in combination with its hardware.

[0073] The computer program product of a multi-channel satellite cloud image prediction method and device based on attention provided in this embodiment includes a computer-readable storage medium storing a non-volatile program code executable by a processor, and the instructions included in the program code can be used to execute the method described in the previous method embodiment. The specific implementation can be referred to the method embodiment, which will not be repeated here. In addition, each functional unit in each embodiment of the present invention can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in this embodiment. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or CD and other media that can store program code.

[0074] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0075] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations.

Claims

1. A multi-temporal cloud image prediction method based on multi-channel satellite data, characterized in that: The steps include: Step 1: Download satellite cloud images of different bands; Step 2: Splice satellite cloud images of different bands to form a multi-channel satellite cloud image; Step 3: Calculate the brightness temperature value of each pixel in each satellite cloud image, thereby mapping the satellite cloud image into the form of brightness temperature value; Step 4: Logarithmically normalize the brightness temperature value in each satellite cloud image; Step 5: construct a multi-channel multi-temporal cloud image prediction network, and use the multi-channel satellite cloud image normalized in step 4 as a training set to train the multi-channel multi-temporal cloud image prediction network to obtain a trained multi-channel multi-temporal cloud image prediction network; Step 6: Use the trained multi-channel multi-phase cloud image prediction network for prediction.

2. The multi-temporal cloud image prediction method based on multi-channel satellite data according to claim 1, characterized in that: In step 3, the brightness temperature value of each pixel in each satellite cloud image is calculated using the following formula: Among them, T b Represents the brightness temperature value of the bth pixel in the satellite cloud image, Image b represents the pixel value of the bth pixel in the satellite cloud image, f1 and f2 represent the coefficients of the Planck function, b1 and b2 are the spectral response function offset and scale correction terms.

3. The multi-temporal cloud image prediction method based on multi-channel satellite data according to claim 1, characterized in that: In step 4, a logarithmic function is used for normalization, and the expression is as follows: in, represents the normalized brightness temperature value of the bth pixel in the satellite cloud image, I min Indicates the minimum brightness temperature, T b It represents the brightness temperature of the bth pixel in the satellite cloud image, and s' is the scale size.

4. The multi-temporal cloud image prediction method based on multi-channel satellite data according to claim 1, characterized in that: The multi-channel multi-phase cloud map prediction network includes a selective attention module, an iterative downsampling module, an iterative upsampling module and a channel separation output module; The selective attention module includes a first stage and a second stage, wherein the first stage is used to extract multi-scale deep features of the original input image, and the multi-scale deep features are input into the second stage to establish correlations between features of different scales to obtain channel enhancement features, and then the channel enhancement features are convolved to the same size as the input feature size to obtain spatial channel enhancement features, and the spatial channel enhancement features are input into an iterative downsampling module to gradually extract long-term and short-term time information contained in the spatial channel enhancement features, while learning the complex structure and motion pattern of the cloud clusters implied in the image; then the output features of the iterative downsampling module are input into an iterative upsampling module to restore the detailed features of the image, and the output features of the iterative upsampling module are input into a channel separation module, and the channel separation module uses C groups of independent filters to predict a single-channel satellite cloud image, where C is the number of channels of the original input image input into the multi-channel multi-phase cloud image prediction network, and then the predicted single-channel cloud image is spliced ​​in the channel dimension to obtain a predicted multi-channel satellite cloud image.

5. The multi-temporal cloud image prediction method based on multi-channel satellite data according to claim 4, characterized in that: The first stage of the selective attention module includes n pooling layers, n upsampling modules, n element-by-element multiplication modules and a splicing module. The i-th pooling layer is connected to the i-th upsampling module. The output of the i-th upsampling module is multiplied by the original image input to the selective attention module through the i-th element-by-element multiplication module. The outputs of all element-by-element multiplication modules are spliced ​​through the splicing module to obtain multi-scale deep features; the second stage includes the first to fourth reshaping modules, the first to second matrix multiplication modules, the softmax module, the addition module and the first convolution module. The first to third reshaping modules reshape the features input to the second stage into different dimensional features, multiply the features output by the second and third reshaping modules through the first matrix multiplication module and then input them into the softmax module. The softmax module multiplies the output features with the output features of the first reshaping module through the second matrix multiplication module and then inputs them into the fourth reshaping module. The features output by the fourth reshaping module are added to the features input to the second stage to obtain channel enhancement features. The obtained channel enhancement features are input into the first convolution module to obtain spatial channel enhancement features.

6. The multi-temporal cloud image prediction method based on multi-channel satellite data according to claim 4, characterized in that: The structures of the C groups of filters in the channel separation output module are the same, and all include a second convolution module, a first deconvolution module, a third convolution module and a fourth convolution module which are connected in sequence.

7. The multi-temporal cloud image prediction method based on multi-channel satellite data according to claim 4, characterized in that: The loss function of the multi-channel multi-phase cloud image prediction network As shown below: Among them, Y represents the real multi-channel satellite cloud image, represents the predicted value of Y, i represents the i-th sample, j represents the j-th frame, N represents the batch size, and S represents the total number of frames of the cloud map in a sample.

8. A prediction model applied to a multi-temporal cloud image prediction method based on multi-channel satellite data, characterized in that: include: Data acquisition module, data processing module, model building module and model training module; The data acquisition module is used to download satellite cloud images of different bands; The data processing module is used to execute steps 2 to 4; The model building module is used to build a multi-channel and multi-phase cloud image prediction network; The model training module is used to train the constructed multi-channel and multi-phase cloud image prediction network.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the multi-temporal cloud image prediction method based on multi-channel satellite data in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the multi-temporal cloud image prediction method based on multi-channel satellite data according to any one of claims 1 to 7 is executed.

Citation Information

Patent Citations

  • Cloud type calculation method and device, computer equipment and storage medium

    CN111382698A

  • AI-CFD-based hypersonic aerodynamic thermal environment intelligent rapid prediction method

    CN115375012A

  • Satellite cloud picture prediction method based on improved U-Net network

    CN117115621A

  • Foundation cloud image classification method based on improved combined convolutional network

    CN117671325A

Cited By

  • Tropical cyclone satellite cloud picture prediction method based on certainty-probabilistic combination

    CN120853042A