CN111505738A - Method and equipment for predicting meteorological factors in numerical weather forecast - Google Patents
Method and equipment for predicting meteorological factors in numerical weather forecast Download PDFInfo
- Publication number
- CN111505738A CN111505738A CN202010188600.0A CN202010188600A CN111505738A CN 111505738 A CN111505738 A CN 111505738A CN 202010188600 A CN202010188600 A CN 202010188600A CN 111505738 A CN111505738 A CN 111505738A
- Authority
- CN
- China
- Prior art keywords
- resolution
- image
- predicted
- channel
- meteorological
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
Landscapes
- Environmental & Geological Engineering (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Atmospheric Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Ecology (AREA)
- Environmental Sciences (AREA)
- Image Processing (AREA)
Abstract
The application provides a method and equipment for predicting meteorological factors in numerical weather forecast, which can construct an image resolution improvement model corresponding to the meteorological factors to be predicted, inputting the multi-channel image with low resolution into the image resolution improvement model to obtain the multi-channel image with high resolution, finally determining the channel data corresponding to the meteorological factors to be predicted in the multi-channel image with high resolution as the high resolution data corresponding to the meteorological factors to be predicted, realizing the fitting of the computational dynamics framework through the image resolution improvement network, simplifying the complex operation in numerical weather forecast into the training of a deep learning network, therefore, the calculation time of the numerical weather forecast is reduced, the prediction efficiency of the single weather factor is improved, the influence of the relevant weather factors on the weather factors to be predicted in the prediction process is also considered, and the prediction accuracy of the single weather factor is improved.
Description
Technical Field
The application relates to the field of weather forecast, in particular to a method and equipment for predicting meteorological factors in numerical weather forecast.
Background
The Numerical Weather forecast (Numerical Weather Prediction) refers to a Numerical calculation method for predicting the atmospheric motion state and the Weather phenomenon in a certain period of time by solving a fluid mechanics and thermodynamic equation set describing the Weather evolution process on a large computer according to the actual atmospheric conditions under certain initial values and boundary values. Numerical weather forecasting is essentially a complex set of equations solved consisting of the continuous equation, thermodynamic equation, water vapor equation, state equation, and multiple equations of motion, as well as many different parameterization schemes. The parameters involved in the numerical weather forecast are very many, the system of correlation equations is very complex, and the calculation amount is very large.
In present-day numerical weather forecasts, the weather data often has only low-resolution data, such as 25 km resolution, etc., where adjacent data represents data collected by two weather stations that are 25 km apart, and high-resolution data, such as 3 km resolution or 9 km resolution, etc., is very little. If it is desired to refine weather forecast, such as forecast for a xu hui region or a Min-go region, based on low resolution weather data, such as weather data in the Shanghai region, the existing methods for numerical weather forecast have problems of long calculation time and inaccurate prediction result.
Disclosure of Invention
An object of the present application is to provide a method and an apparatus for predicting meteorological factors in a numerical weather forecast, which are used to solve the problems in the prior art that the calculation time of the numerical weather forecast is long and the prediction of a single meteorological factor is not accurate enough.
In order to achieve the above object, the present application provides a method for predicting meteorological factors in a numerical weather forecast, wherein the method comprises:
constructing an image resolution improvement model corresponding to meteorological factors to be predicted, wherein a feature extraction network in the image resolution improvement model combines the output feature map of the convolution layer of the channel corresponding to the meteorological factors to be predicted and the output feature map of the convolution layer of the channel corresponding to the related meteorological factors according to preset weight, and then takes the combined feature map as an input feature map of the next convolution layer;
respectively taking the low-resolution data corresponding to the meteorological factors to be predicted and the low-resolution data corresponding to the relevant meteorological factors as image channel data to synthesize a low-resolution multi-channel image;
inputting the low-resolution multi-channel image into the image resolution improvement model to obtain a high-resolution multi-channel image;
and determining channel data corresponding to the meteorological factors to be predicted in the high-resolution multi-channel image as high-resolution data corresponding to the meteorological factors to be predicted.
Further, constructing an image resolution improvement model corresponding to the meteorological factors to be predicted, which comprises the following steps:
acquiring low-resolution sample data corresponding to meteorological factors to be predicted and low-resolution sample data corresponding to relevant meteorological factors;
respectively taking the low-resolution sample data corresponding to the meteorological factors to be predicted and the low-resolution sample data corresponding to the relevant meteorological factors as image channel data to synthesize a low-resolution multi-channel sample image;
acquiring high-resolution sample data corresponding to meteorological factors to be predicted and high-resolution sample data corresponding to relevant meteorological factors;
respectively taking the high-resolution sample data corresponding to the meteorological factors to be predicted and the high-resolution sample data corresponding to the relevant meteorological factors as image channel data to synthesize a high-resolution multi-channel sample image;
and inputting the low-resolution multi-channel sample image as training data and the high-resolution multi-channel sample image as supervised learning data into a super-resolution neural network, and determining the trained super-resolution neural network model as an image resolution improvement model corresponding to the meteorological factor to be predicted.
Further, after obtaining the low resolution sample data corresponding to the meteorological factor to be predicted and the low resolution sample data corresponding to the relevant meteorological factor, the method further comprises the following steps: converting the low-resolution sample data of the meteorological factors to be predicted into first conversion data with values within a preset value interval through linear mapping, and converting the low-resolution sample data corresponding to the relevant meteorological factors into second conversion data with values within the preset value interval through linear mapping;
after obtaining the high resolution sample data corresponding to the meteorological factors to be predicted and the high resolution sample data corresponding to the relevant meteorological factors, the method further comprises the following steps: and converting the high-resolution sample data of the meteorological factors to be predicted into third conversion data with values within a preset numerical value interval through linear mapping, and converting the high-resolution sample data corresponding to the relevant meteorological factors into fourth conversion data with values within the preset numerical value interval through linear mapping.
Further, the preset value interval is 0-255.
Further, synthesizing the low-resolution sample data corresponding to the meteorological factor to be predicted and the low-resolution sample data corresponding to the relevant meteorological factor as image channel data respectively into a low-resolution multi-channel sample image, including:
synthesizing the first conversion data and the second conversion data as image channel data respectively into a low-resolution multichannel sample image;
and respectively taking the high-resolution sample data corresponding to the meteorological factors to be predicted and the high-resolution sample data corresponding to the relevant meteorological factors as image channel data to synthesize a high-resolution multi-channel sample image, wherein the method comprises the following steps:
and respectively taking the third conversion data and the fourth conversion data as image channel data to synthesize the image into a high-resolution multi-channel sample image.
Further, the method further comprises:
and determining other meteorological factors influencing the prediction of the meteorological factors to be predicted as related meteorological factors.
Further, determining the channel data corresponding to the meteorological factor to be predicted in the high-resolution multi-channel image as the high-resolution data corresponding to the meteorological factor to be predicted, including:
converting channel data corresponding to the meteorological factors to be predicted in the high-resolution multi-channel image into data in the value range of the meteorological factors to be predicted through linear mapping;
and determining the converted data as high-resolution data corresponding to the meteorological factors to be predicted.
Further, the merging, by the feature extraction network in the image resolution enhancement model, the convolution layer output feature map of the channel corresponding to the meteorological factor to be predicted and the convolution layer output feature map of the channel corresponding to the relevant meteorological factor according to the preset weight includes:
and combining the convolution layer output characteristic diagram of the channel corresponding to the meteorological factor to be predicted and the convolution layer output characteristic diagram of the channel corresponding to the related meteorological factor by a characteristic extraction network in the image resolution improvement model in a weighting matrix addition mode according to preset weights.
Based on another aspect of the present application, there is also provided an apparatus comprising a memory for storing computer program instructions and a processor for executing the program instructions, wherein the computer program instructions, when executed by the processor, cause the apparatus to perform the method of predicting meteorological factors in a numerical weather forecast as described above.
The present application also provides a computer readable medium having stored thereon computer readable instructions executable by a processor to implement the aforementioned method of predicting weather factors in a numerical weather forecast.
Compared with the prior art, the scheme provided by the application can construct an image resolution improvement model corresponding to the meteorological factors to be predicted, low resolution data corresponding to the meteorological factors to be predicted and low resolution data corresponding to the related meteorological factors are used as image channel data to be synthesized into a low resolution multi-channel image, the low resolution multi-channel image is input into the image resolution improvement model to obtain a high resolution multi-channel image, and finally the channel data corresponding to the meteorological factors to be predicted in the high resolution multi-channel image is determined as the high resolution data corresponding to the meteorological factors to be predicted, so that fitting of a computational dynamics framework through an image resolution improvement network is realized, complex operation in numerical weather forecast is simplified into training of a depth learning network, the computation time of the numerical weather forecast is reduced, and the prediction efficiency of a single meteorological factor is improved, the influence of relevant meteorological factors on the meteorological factors to be predicted in the prediction process is also considered, and the prediction accuracy of a single meteorological factor is improved.
Drawings
Other features, objects and advantages of the present application will become more apparent upon reading of the following detailed description of non-limiting embodiments thereof, made with reference to the accompanying drawings in which:
fig. 1 is a flowchart illustrating a method for predicting meteorological factors in a numerical weather forecast according to some embodiments of the present disclosure.
The same or similar reference numbers in the drawings identify the same or similar elements.
Detailed Description
The present application is described in further detail below with reference to the attached figures.
In a typical configuration of the present application, the terminal and the network device each include one or more processors (CPUs), input/output interfaces, network interfaces, and memories.
The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.
Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Disks (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include non-transitory computer readable media (transient media), such as modulated data signals and carrier waves.
Fig. 1 illustrates a method for predicting meteorological factors in a numerical weather forecast according to some embodiments of the present application, which may specifically include the following steps:
step S101, constructing an image resolution improvement model corresponding to meteorological factors to be predicted, wherein a feature extraction network in the image resolution improvement model combines a convolutional layer output feature map of a channel corresponding to the meteorological factors to be predicted and a convolutional layer output feature map of a channel corresponding to related meteorological factors according to preset weight, and then uses the combined feature map as an input feature map of a next convolutional layer;
step S102, the low-resolution data corresponding to the meteorological factors to be predicted and the low-resolution data corresponding to the relevant meteorological factors are respectively used as image channel data to be synthesized into a low-resolution multi-channel image;
step S103, inputting the low-resolution multi-channel image into the image resolution improvement model to obtain a high-resolution multi-channel image;
and step S104, determining the channel data corresponding to the meteorological factor to be predicted in the high-resolution multi-channel image as the high-resolution data corresponding to the meteorological factor to be predicted.
The scheme is particularly suitable for scenes in which meteorological factors are expected to be predicted in numerical weather forecast, an image resolution improvement model corresponding to the meteorological factors to be predicted can be built, low-resolution data corresponding to the meteorological factors and relevant meteorological factors for prediction are used as image channel data to be synthesized into a low-resolution multi-channel image, the low-resolution multi-channel image is input into the built image resolution improvement model to obtain an output high-resolution multi-channel image, channel data corresponding to the meteorological factors to be predicted in the high-resolution multi-channel image are acquired, and the channel data are high-resolution data obtained by predicting the meteorological factors.
In step S101, an image resolution enhancement model corresponding to a meteorological factor to be predicted is first constructed. And combining the output characteristic diagram of the convolution layer of the channel corresponding to the meteorological factor to be predicted and the output characteristic diagram of the convolution layer of the channel corresponding to the relevant meteorological factor by the characteristic extraction network in the image resolution improvement model according to the preset weight, and taking the combined characteristic diagram as the input characteristic diagram of the next convolution layer.
The meteorological factors to be predicted are certain meteorological factors which are selected according to the needs of the user and are expected to be predicted, such as precipitation, near-ground humidity, air pressure and the like. And the corresponding model parameters in the prediction process of each meteorological factor are different, so that a corresponding prediction model is established for each meteorological factor, and the prediction model is used for improving the meteorological factor data with low resolution into the meteorological factor data with high resolution. Here, in order to solve the problems of long calculation time and great consumption of calculation resources caused by predicting meteorological factors by solving a large number of equation sets in the conventional numerical weather forecast, some embodiments of the present application merge corresponding data used for meteorological factor prediction as image channel data into an image, obtain an image with high resolution by using an image resolution enhancement model, and use corresponding channel data in the obtained image with high resolution as high resolution data of the meteorological factors to be predicted, so that the calculation time can be greatly shortened, and can be shortened from several hours to several seconds in the conventional method.
In some embodiments of the present application, constructing an image resolution improvement model corresponding to a meteorological factor to be predicted may specifically include the following steps:
(1) acquiring low-resolution sample data corresponding to meteorological factors to be predicted and low-resolution sample data corresponding to relevant meteorological factors, wherein the low-resolution sample data are training data used for constructing an image resolution improvement model, and are usually meteorological data acquired by meteorological stations with a long distance, such as sample data with a resolution of 25 kilometers acquired by two meteorological stations with a distance of 25 kilometers;
(2) respectively taking low-resolution sample data corresponding to meteorological factors to be predicted and low-resolution sample data corresponding to relevant meteorological factors as image channel data to synthesize a low-resolution multi-channel sample image, wherein the meteorological factors to be predicted correspond to one image channel, the relevant meteorological factors correspond to several image channels, the low-resolution data corresponding to the meteorological factors to be predicted are taken as corresponding image channel data, the low-resolution data corresponding to the relevant meteorological factors are similarly taken as data of the corresponding image channels, and finally, the image channel data are combined into one image to obtain the low-resolution multi-channel sample image;
(3) acquiring high-resolution sample data corresponding to meteorological factors to be predicted and high-resolution sample data corresponding to relevant meteorological factors, wherein the high-resolution sample data is usually meteorological data acquired by meteorological stations with a short distance, such as sample data with a resolution of 9 kilometers acquired by two meteorological stations with a distance of 9 kilometers, or sample data with a resolution of 3 kilometers acquired by two meteorological stations with a distance of 3 kilometers, and the like;
(4) respectively taking high-resolution sample data corresponding to meteorological factors to be predicted and high-resolution sample data corresponding to relevant meteorological factors as image channel data to synthesize a high-resolution multi-channel sample image, and obtaining a high-resolution multi-channel image by adopting a similar method for obtaining a low-resolution multi-channel image;
(5) and inputting the low-resolution multi-channel sample image as training data and the high-resolution multi-channel sample image as supervised learning data into a super-resolution neural network, and determining the trained super-resolution neural network model as an image resolution improvement model corresponding to the meteorological factor to be predicted.
In some embodiments of the present application, a super-resolution neural network is used as a base network for constructing an image resolution enhancement model. A Super-Resolution neural network (SRCNN) is a neural network for obtaining a high-Resolution image from a low-Resolution image, and mainly includes three layers of neural networks, including: the method comprises the steps of feature extraction, nonlinear mapping and image reconstruction, wherein a feature extraction network is used for obtaining a fuzzy image by carrying out binomial interpolation on a low-resolution image and extracting image features from the fuzzy image, a nonlinear mapping network is used for mapping the low-resolution image features to high resolution, and an image reconstruction network is used for restoring image details to obtain a clear high-resolution image. In some preferred embodiments, RCAN (Residual channel attention Networks) is used as a super-resolution neural network for constructing the image resolution enhancement model.
In some embodiments of the present application, in the feature extraction network of the super-resolution neural network, the convolutional layer output feature map of the channel corresponding to the meteorological factor to be predicted and the convolutional layer output feature map of the channel corresponding to the relevant meteorological factor may be merged according to the preset weight, and then the merged feature map is used as the input feature map of the next convolutional layer. The feature extraction network performs feature extraction on the image through a plurality of connected convolution layers, each convolution layer receives an input feature map, and outputs a corresponding feature map after convolution processing. In order to improve the prediction accuracy of the meteorological factors to be predicted, output data, namely an output characteristic diagram, obtained after image data of a channel corresponding to the meteorological factors to be predicted pass through a convolutional layer, and output characteristic diagrams obtained after the image data of the channel corresponding to the relevant meteorological factors pass through the convolutional layer are weighted and then combined. Preferably, the weighted output characteristic maps are combined, and the combination can be performed in a weighted matrix addition manner. For example, the preset weight of the channel data corresponding to the meteorological factors to be predicted is set to be 0.8, 2 related meteorological factors are provided, and the corresponding preset weights are respectively 0.1, then the channel data (namely, data in a matrix form) corresponding to the meteorological factors to be predicted are multiplied by the corresponding weights of 0.8, the channel data corresponding to the related meteorological factors are respectively multiplied by the corresponding weights of 0.1, then the weighted channel data are subjected to matrix addition, and finally the data obtained by the matrix addition are used as the input characteristic diagram of the next convolutional layer.
In some embodiments of the present application, the weather-related factors refer to other weather factors affecting the prediction of the weather factor to be predicted, and different weather factors to be predicted have different weather-related factors, so after the weather factor to be predicted is determined, one or more other weather factors affecting the prediction are determined as the weather factor according to the specific situation of the weather factor to be predicted, for example, the weather factor to be predicted is precipitation, and the weather factor affecting the precipitation prediction has near-ground humidity and atmospheric pressure, so that the near-ground humidity and atmospheric pressure are determined as the weather factors.
In some embodiments of the application, after the low-resolution sample data corresponding to the meteorological factor to be predicted and the low-resolution sample data corresponding to the relevant meteorological factor are obtained, the low-resolution sample data of the meteorological factor to be predicted can be converted into first conversion data with a value within a preset value interval through linear mapping, and the low-resolution sample data corresponding to the relevant meteorological factor is converted into second conversion data with a value within the preset value interval through linear mapping. Similarly, after the high-resolution sample data corresponding to the meteorological factor to be predicted and the high-resolution sample data corresponding to the relevant meteorological factor are obtained, the high-resolution sample data of the meteorological factor to be predicted can be converted into third conversion data with the value within a preset numerical value interval through linear mapping, and the high-resolution sample data corresponding to the relevant meteorological factor is converted into fourth conversion data with the value within the preset numerical value interval through linear mapping.
Here, the preset value interval may be preferably set to be between 0 and 255. Because the value ranges of the data corresponding to the meteorological factors to be predicted are different from those of the data corresponding to the relevant meteorological factors and are consistent with the value range of the data in the image RGB channel, the corresponding image processing in the super-resolution neural network is facilitated, and the value range corresponding to the meteorological factor data can be mapped to the value range of the image RGB channel data. For example, the value range of the meteorological factor data to be predicted is 0-10, and the value range can be expanded to be 0-255 through linear mapping.
In some embodiments of the present application, after linear mapping is performed on value ranges corresponding to data of meteorological factors and related meteorological factors to be predicted, a corresponding low-resolution multichannel sample image and a corresponding high-resolution multichannel sample image are synthesized according to the data obtained after mapping, specifically, the first conversion data and the second conversion data are respectively used as image channel data to be synthesized into a low-resolution multichannel sample image; and then the third conversion data and the fourth conversion data are respectively taken as image channel data to be synthesized into a high-resolution multi-channel sample image. After a low-resolution multichannel sample image and a high-resolution multichannel sample image are obtained, the two sample images are input into a super-resolution neural network for training, and a corresponding image resolution improvement model is obtained.
In step S102, the low-resolution data corresponding to the weather factor to be predicted and the low-resolution data corresponding to the relevant weather factor are respectively used as image channel data to be synthesized into a low-resolution multi-channel image. Here, the low resolution data corresponding to the meteorological factor to be predicted and the low resolution data corresponding to the relevant meteorological factor may be the low resolution sample data used for training the image resolution improvement model, or may be other low resolution data that is desired to be predicted. Similarly, the low-resolution data can also be converted into data with values within a preset value interval through linear mapping, and then the converted data is synthesized into a low-resolution multi-channel image.
In step S103, the low-resolution multi-channel image is input to the image resolution enhancement model, and a high-resolution multi-channel image is acquired.
In step S104, the channel data corresponding to the weather factor to be predicted in the obtained high-resolution multi-channel image is determined as the high-resolution data corresponding to the weather factor to be predicted.
In some embodiments of the present application, if the low-resolution multichannel image corresponding to the obtained high-resolution multichannel image is synthesized from the converted data, the obtained high-resolution multichannel image also needs to implement data restoration by linear mapping, specifically, the channel data corresponding to the meteorological factor to be predicted in the high-resolution multichannel image is converted into data within the value range of the meteorological factor to be predicted by linear mapping, and the obtained converted data is determined as the high-resolution data corresponding to the meteorological factor to be predicted.
Table 1 below shows the accuracy and the calculation time of the weather factor high resolution data predicted in some preferred embodiments of the present application and the actually corresponding weather factor high resolution data, in the case of behavior 1 of the table, the weather factor to be predicted is the earth surface pressure, the low resolution data used for prediction is 25 km resolution, the calculation time is 1.16 seconds when the predicted high resolution data is 3 km resolution, the accuracy is 82%, the calculation time is 0.98 seconds when the predicted high resolution data is 9 km resolution, and the accuracy is 91%.
TABLE 1
Some embodiments of the present application also provide an apparatus comprising a memory for storing computer program instructions and a processor for executing the program instructions, wherein the computer program instructions, when executed by the processor, cause the apparatus to perform the method of predicting meteorological factors in a numerical weather forecast as previously described.
Some embodiments of the present application also provide a computer readable medium having stored thereon computer readable instructions executable by a processor to implement the aforementioned method of predicting weather factors in a numerical weather forecast.
In summary, the scheme provided by the application can construct an image resolution improvement model corresponding to the meteorological factors to be predicted, low resolution data corresponding to the meteorological factors to be predicted and low resolution data corresponding to the relevant meteorological factors are used as image channel data to be synthesized into a low resolution multi-channel image, the low resolution multi-channel image is input into the image resolution improvement model to obtain a high resolution multi-channel image, and finally the channel data corresponding to the meteorological factors to be predicted in the high resolution multi-channel image is determined as the high resolution data corresponding to the meteorological factors to be predicted, so that fitting of a computational dynamics framework through an image resolution improvement network is realized, complex operations in numerical weather forecast are simplified into training of a depth learning network, the computation time of the numerical weather forecast is reduced, and the prediction efficiency of a single meteorological factor is improved, the influence of relevant meteorological factors on the meteorological factors to be predicted in the prediction process is also considered, and the prediction accuracy of a single meteorological factor is improved.
It should be noted that the present application may be implemented in software and/or a combination of software and hardware, for example, implemented using Application Specific Integrated Circuits (ASICs), general purpose computers or any other similar hardware devices. In one embodiment, the software programs of the present application may be executed by a processor to implement the steps or functions described above. Likewise, the software programs (including associated data structures) of the present application may be stored in a computer readable recording medium, such as RAM memory, magnetic or optical drive or diskette and the like. Additionally, some of the steps or functions of the present application may be implemented in hardware, for example, as circuitry that cooperates with the processor to perform various steps or functions.
In addition, some of the present application may be implemented as a computer program product, such as computer program instructions, which when executed by a computer, may invoke or provide methods and/or techniques in accordance with the present application through the operation of the computer. Program instructions which invoke the methods of the present application may be stored on a fixed or removable recording medium and/or transmitted via a data stream on a broadcast or other signal-bearing medium and/or stored within a working memory of a computer device operating in accordance with the program instructions. An embodiment according to the present application comprises a device comprising a memory for storing computer program instructions and a processor for executing the program instructions, wherein the computer program instructions, when executed by the processor, trigger the device to perform a method and/or a solution according to the aforementioned embodiments of the present application.
It will be evident to those skilled in the art that the present application is not limited to the details of the foregoing illustrative embodiments, and that the present application may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned. Furthermore, it is obvious that the word "comprising" does not exclude other elements or steps, and the singular does not exclude the plural. A plurality of units or means recited in the apparatus claims may also be implemented by one unit or means in software or hardware. The terms first, second, etc. are used to denote names, but not any particular order.
Claims (10)
1. A method for predicting meteorological factors in a numerical weather forecast, wherein the method comprises:
constructing an image resolution improvement model corresponding to meteorological factors to be predicted, wherein a feature extraction network in the image resolution improvement model combines the output feature map of the convolution layer of the channel corresponding to the meteorological factors to be predicted and the output feature map of the convolution layer of the channel corresponding to the related meteorological factors according to preset weight, and then takes the combined feature map as an input feature map of the next convolution layer;
respectively taking the low-resolution data corresponding to the meteorological factors to be predicted and the low-resolution data corresponding to the relevant meteorological factors as image channel data to synthesize a low-resolution multi-channel image;
inputting the low-resolution multi-channel image into the image resolution improvement model to obtain a high-resolution multi-channel image;
and determining channel data corresponding to the meteorological factors to be predicted in the high-resolution multi-channel image as high-resolution data corresponding to the meteorological factors to be predicted.
2. The method of claim 1, wherein constructing an image resolution enhancement model corresponding to the meteorological factor to be predicted comprises:
acquiring low-resolution sample data corresponding to meteorological factors to be predicted and low-resolution sample data corresponding to relevant meteorological factors;
respectively taking the low-resolution sample data corresponding to the meteorological factors to be predicted and the low-resolution sample data corresponding to the relevant meteorological factors as image channel data to synthesize a low-resolution multi-channel sample image;
acquiring high-resolution sample data corresponding to meteorological factors to be predicted and high-resolution sample data corresponding to relevant meteorological factors;
respectively taking the high-resolution sample data corresponding to the meteorological factors to be predicted and the high-resolution sample data corresponding to the relevant meteorological factors as image channel data to synthesize a high-resolution multi-channel sample image;
and inputting the low-resolution multi-channel sample image as training data and the high-resolution multi-channel sample image as supervised learning data into a super-resolution neural network, and determining the trained super-resolution neural network model as an image resolution improvement model corresponding to the meteorological factor to be predicted.
3. The method according to claim 2, wherein after obtaining the low resolution sample data corresponding to the meteorological factor to be predicted and the low resolution sample data corresponding to the relevant meteorological factor, further comprising: converting the low-resolution sample data of the meteorological factors to be predicted into first conversion data with values within a preset value interval through linear mapping, and converting the low-resolution sample data corresponding to the relevant meteorological factors into second conversion data with values within the preset value interval through linear mapping;
after obtaining the high resolution sample data corresponding to the meteorological factors to be predicted and the high resolution sample data corresponding to the relevant meteorological factors, the method further comprises the following steps: and converting the high-resolution sample data of the meteorological factors to be predicted into third conversion data with values within a preset numerical value interval through linear mapping, and converting the high-resolution sample data corresponding to the relevant meteorological factors into fourth conversion data with values within the preset numerical value interval through linear mapping.
4. The method according to claim 3, wherein the preset value interval is 0-255.
5. The method according to claim 3, wherein synthesizing low resolution sample data corresponding to the meteorological factor to be predicted and low resolution sample data corresponding to the relevant meteorological factor as image channel data respectively into a low resolution multi-channel sample image comprises:
synthesizing the first conversion data and the second conversion data as image channel data respectively into a low-resolution multichannel sample image;
and respectively taking the high-resolution sample data corresponding to the meteorological factors to be predicted and the high-resolution sample data corresponding to the relevant meteorological factors as image channel data to synthesize a high-resolution multi-channel sample image, wherein the method comprises the following steps:
and respectively taking the third conversion data and the fourth conversion data as image channel data to synthesize the image into a high-resolution multi-channel sample image.
6. The method of claim 1, wherein the method further comprises:
and determining other meteorological factors influencing the prediction of the meteorological factors to be predicted as related meteorological factors.
7. The method of claim 1, wherein determining the channel data corresponding to the meteorological factor to be predicted in the high-resolution multi-channel image as the high-resolution data corresponding to the meteorological factor to be predicted comprises:
converting channel data corresponding to the meteorological factors to be predicted in the high-resolution multi-channel image into data in the value range of the meteorological factors to be predicted through linear mapping;
and determining the converted data as high-resolution data corresponding to the meteorological factors to be predicted.
8. The method of claim 1, wherein the merging, according to the preset weight, the convolutional layer output feature map of the channel corresponding to the meteorological factor to be predicted and the convolutional layer output feature map of the channel corresponding to the relevant meteorological factor by the feature extraction network in the image resolution enhancement model comprises:
and combining the convolution layer output characteristic diagram of the channel corresponding to the meteorological factor to be predicted and the convolution layer output characteristic diagram of the channel corresponding to the related meteorological factor by a characteristic extraction network in the image resolution improvement model in a weighting matrix addition mode according to preset weights.
9. An apparatus comprising a memory for storing computer program instructions and a processor for executing the program instructions, wherein the computer program instructions, when executed by the processor, cause the apparatus to perform the method of any of claims 1 to 8.
10. A computer readable medium having computer readable instructions stored thereon which are executable by a processor to implement the method of any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010188600.0A CN111505738A (en) | 2020-03-17 | 2020-03-17 | Method and equipment for predicting meteorological factors in numerical weather forecast |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010188600.0A CN111505738A (en) | 2020-03-17 | 2020-03-17 | Method and equipment for predicting meteorological factors in numerical weather forecast |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111505738A true CN111505738A (en) | 2020-08-07 |
Family
ID=71867067
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010188600.0A Pending CN111505738A (en) | 2020-03-17 | 2020-03-17 | Method and equipment for predicting meteorological factors in numerical weather forecast |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111505738A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112819218A (en) * | 2021-01-27 | 2021-05-18 | 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) | High-resolution urban mobile source pollution space-time prediction method, system and storage medium |
| CN117233870A (en) * | 2023-11-15 | 2023-12-15 | 南京信息工程大学 | Short-term precipitation set forecasting and downscaling method based on multiple meteorological elements |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108564587A (en) * | 2018-03-07 | 2018-09-21 | 浙江大学 | A kind of a wide range of remote sensing image semantic segmentation method based on full convolutional neural networks |
| KR101933856B1 (en) * | 2017-07-03 | 2018-12-31 | (주)시정 | System for image processing using the convolutional neural network and method using the same |
| CN109146788A (en) * | 2018-08-16 | 2019-01-04 | 广州视源电子科技股份有限公司 | Super-resolution image reconstruction method and device based on deep learning |
| CN110136060A (en) * | 2019-04-24 | 2019-08-16 | 西安电子科技大学 | The image super-resolution rebuilding method of network is intensively connected based on shallow-layer |
| CN110618473A (en) * | 2019-10-09 | 2019-12-27 | 北京北科融智云计算科技有限公司 | Instant meteorological information guarantee method |
| CN110619605A (en) * | 2019-09-17 | 2019-12-27 | 中国气象局公共气象服务中心(国家预警信息发布中心) | Parallel meteorological element downscaling method and device, electronic equipment and storage medium |
| CN110874630A (en) * | 2019-10-23 | 2020-03-10 | 中国人民解放军空军研究院战场环境研究所 | Deep learning-based numerical model product downscaling refinement method |
-
2020
- 2020-03-17 CN CN202010188600.0A patent/CN111505738A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101933856B1 (en) * | 2017-07-03 | 2018-12-31 | (주)시정 | System for image processing using the convolutional neural network and method using the same |
| CN108564587A (en) * | 2018-03-07 | 2018-09-21 | 浙江大学 | A kind of a wide range of remote sensing image semantic segmentation method based on full convolutional neural networks |
| CN109146788A (en) * | 2018-08-16 | 2019-01-04 | 广州视源电子科技股份有限公司 | Super-resolution image reconstruction method and device based on deep learning |
| CN110136060A (en) * | 2019-04-24 | 2019-08-16 | 西安电子科技大学 | The image super-resolution rebuilding method of network is intensively connected based on shallow-layer |
| CN110619605A (en) * | 2019-09-17 | 2019-12-27 | 中国气象局公共气象服务中心(国家预警信息发布中心) | Parallel meteorological element downscaling method and device, electronic equipment and storage medium |
| CN110618473A (en) * | 2019-10-09 | 2019-12-27 | 北京北科融智云计算科技有限公司 | Instant meteorological information guarantee method |
| CN110874630A (en) * | 2019-10-23 | 2020-03-10 | 中国人民解放军空军研究院战场环境研究所 | Deep learning-based numerical model product downscaling refinement method |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112819218A (en) * | 2021-01-27 | 2021-05-18 | 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) | High-resolution urban mobile source pollution space-time prediction method, system and storage medium |
| CN112819218B (en) * | 2021-01-27 | 2022-05-06 | 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) | High-resolution urban mobile source pollution space-time prediction method, system and storage medium |
| CN117233870A (en) * | 2023-11-15 | 2023-12-15 | 南京信息工程大学 | Short-term precipitation set forecasting and downscaling method based on multiple meteorological elements |
| CN117233870B (en) * | 2023-11-15 | 2024-01-23 | 南京信息工程大学 | Short-term precipitation set forecasting and downscaling method based on multiple meteorological elements |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hutchinson et al. | ANUSPLIN version 4.4 user guide | |
| CN113780149B (en) | Remote sensing image building target efficient extraction method based on attention mechanism | |
| US10339707B2 (en) | Automated generation of digital elevation models | |
| CN114254715B (en) | Super-resolution method, system and application of GF-1WFV satellite image | |
| CN113344291B (en) | Urban inland inundation range forecasting method, device, medium and equipment | |
| CN111754404B (en) | Remote sensing image space-time fusion method based on multi-scale mechanism and attention mechanism | |
| CN111860233B (en) | SAR image complex building extraction method and system based on attention network selection | |
| CN110929543B (en) | Meteorological data processing method and device | |
| CN111505738A (en) | Method and equipment for predicting meteorological factors in numerical weather forecast | |
| CN109447897B (en) | Real scene image synthesis method and system | |
| CN115345866B (en) | Building extraction method in remote sensing image, electronic equipment and storage medium | |
| CN110826454B (en) | Remote sensing image change detection method and device | |
| CN116486255A (en) | High-resolution remote sensing image semantic change detection method based on self-attention feature fusion | |
| CN117454116A (en) | Ground carbon emission monitoring method based on multi-source data interaction network | |
| CN113436287B (en) | Tampered image blind evidence obtaining method based on LSTM network and coding and decoding network | |
| CN114625992A (en) | Visualization method and system for hundred-megametric meteorological network data | |
| KR20220102329A (en) | Apparatus and method for forecasting the seasonal forest fire risk | |
| CN113034511A (en) | Rural building identification algorithm based on high-resolution remote sensing image and deep learning | |
| CN111830595A (en) | Meteorological element prediction method and equipment | |
| CN114510875B (en) | Circulation-current-situation red tide forecasting system and method based on multi-source meteorological data | |
| CN115861891A (en) | Video target detection method, device, equipment and medium | |
| CN115496654A (en) | Image super-resolution reconstruction method, device and medium based on self-attention mechanism | |
| CN115797181A (en) | Image super-resolution reconstruction method for mine fuzzy environment | |
| CN115272846A (en) | Improved Orientdrcnn-based rotating target detection method | |
| CN115527113A (en) | Bare land classification method and device for remote sensing image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| AD01 | Patent right deemed abandoned | ||
| AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20230728 |