Radar data inversion method, marine meteorological detection method and device

Abstract

The present invention provides a radar data inversion method, a maritime meteorological detection method and a device. The inversion method includes: obtaining multi-channel meteorological satellite observation data of a target area; inputting the multi-channel meteorological satellite observation data into a pre-trained radar echo inversion model to obtain the predicted radar reflectivity data of the target area; the radar echo inversion model is trained by adopting a generative adversarial learning strategy on a pre-collected training sample set; the training sample set includes a plurality of training samples, the training samples are multi-channel meteorological satellite observation data of the observation area at the previous moment, and the label of the training sample is the radar reflectivity network observation data of the same observation area at the next moment. This inversion method can accurately invert radar reflectivity data based on meteorological satellite observation data, making up for the deficiency of maritime meteorological radar data caused by the difficulty of building radar base stations at sea.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological observation, and in particular, to a radar data inversion method, a marine meteorological detection method and a device. Background Art

[0002] In current marine meteorological observations, due to the difficulty of building radar base stations at sea, there is a shortage of marine meteorological radar data. To solve this problem, the industry usually uses meteorological observation elements and satellite observation data to establish statistical correlation relationships, and then inversely calculates radar observation data. However, this method has low accuracy. Summary of the Invention

[0003] One or more embodiments of this specification describe a radar data inversion method, a marine meteorological detection method and a device, which can effectively solve the above problems existing in the prior art.

[0004] In a first aspect, a radar data inversion method is provided. The method includes:

[0005] Obtaining multi-channel meteorological satellite observation data of a target area;

[0006] Inputting the multi-channel meteorological satellite observation data into a pre-trained radar echo inversion model to obtain the predicted radar reflectivity data of the target area;

[0007] The radar echo inversion model is trained by using a generative adversarial learning strategy on a pre-collected training sample set. The training sample set includes a plurality of training samples, and the training sample is the multi-channel meteorological satellite observation data of the previous moment of the observation area. The label of the training sample is the radar reflectivity network observation data of the same observation area at the next moment.

[0008] As an optional implementation manner of the method described in the first aspect, in the above method, constructing the training sample specifically includes:

[0009] Collecting the observation data of multi-channel meteorological satellites and radar networks deployed in the same observation area;

[0010] Performing equatorial and meridian projection conversion on the observation data of the multi-channel meteorological satellites, and interpolating the missing values of the converted projection data to obtain the multi-channel meteorological satellite observation data;

[0011] Performing mosaicking on the observation data of the radar network to obtain radar network mosaic data;

[0012] Performing spatio-temporal matching on the multi-channel meteorological satellite observation data and the radar network mosaic data to obtain the radar reflectivity network observation data of the same observation area at the next moment.

[0013] As an alternative implementation of the method described in the first aspect, in the above method, training the radar echo inversion model specifically includes:

[0014] Input the training samples into the generator composed of the radar echo inversion model to obtain the predicted radar reflectivity data;

[0015] Based on the predicted radar reflectivity data and the labels of the training samples, determine the generation loss function;

[0016] Input the predicted radar reflectivity data and the real radar reflectivity data into the discriminator to obtain the discrimination result of the discriminator for the input data;

[0017] According to the discrimination result of the discriminator for the input data and the true label of the input data, determine the discrimination loss function; the true label is used to characterize whether the input data is real radar reflectivity data;

[0018] Based on the generation loss function and the discrimination loss function, update the parameters of the radar echo inversion model.

[0019] Specifically, training the radar echo inversion model further includes:

[0020] For each training sample, determine the first feature image according to the radar reflectivity data predicted by the generator for this training sample;

[0021] Determine the second feature image according to the label of the training sample;

[0022] Determine the style loss function according to the difference in image style between the first feature image and the second feature image;

[0023] Update the parameters of the radar echo inversion model according to the style loss function, the generation loss function, and the discrimination loss function.

[0024] More specifically, determining the style loss function according to the difference in image style between the first feature image and the second feature image specifically includes:

[0025] Use the first Gram matrix to represent the style features of the first feature image;

[0026] Use the second Gram matrix to represent the style features of the second feature image;

[0027] Construct the style loss function according to the first Gram matrix and the second Gram matrix:

[0028]

[0029] Among them, L style represents the style loss, C represents the number of channels of the first feature image and the second feature image, H represents the height of the first feature image and the second feature image, and W represents the width of the first feature image and the second feature image. represents the element value of the i-th row and j-th column of the first Gram matrix. represents the element value of the i-th row and j-th column of the second Gram matrix.

[0030] In a second aspect, a radar data inversion device is provided. The device includes:

[0031] A first data acquisition module configured to acquire multi-channel meteorological satellite observation data of a target area;

[0032] A first inversion module configured to input the multi-channel meteorological satellite observation data into a pre-trained radar echo inversion model to obtain the predicted radar reflectivity data of the target area; the radar echo inversion model is trained using a generative adversarial learning strategy on a pre-collected training sample set; the training sample set includes a plurality of training samples, and the training sample is the multi-channel meteorological satellite observation data of the observation area at the previous moment, and the label of the training sample is the radar reflectivity network observation data of the same observation area at the next moment.

[0033] In a third aspect, a marine meteorological detection method is provided. The method includes:

[0034] Acquire multi-channel meteorological satellite observation data of the current moment of the area to be detected at sea;

[0035] According to the multi-channel meteorological satellite observation data of the current moment of the area to be detected, use the above radar data inversion method to invert the radar reflectivity data of the area to be detected at the next moment;

[0036] Perform meteorological detection according to the radar reflectivity data of the area to be detected at the next moment.

[0037] In a fourth aspect, a marine meteorological detection device is provided. The device includes:

[0038] A second data acquisition module configured to acquire multi-channel meteorological satellite observation data of the current moment of the area to be detected at sea;

[0039] A second inversion module configured to invert the radar reflectivity data of the area to be detected at the next moment according to the multi-channel meteorological satellite observation data of the current moment of the area to be detected by using the above radar data inversion method;

[0040] A meteorological detection module configured to perform meteorological detection based on the radar reflectivity data of the next moment in the area to be detected.

[0041] In a fifth aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program. When the computer program runs on an electronic device, the electronic device is caused to execute the above radar data inversion method, or execute the above marine meteorological detection method.

[0042] In a sixth aspect, an electronic device is provided, including:

[0043] At least one memory for storing programs;

[0044] At least one processor for executing the programs stored in the memory. When the programs stored in the memory are executed, the processor is used to execute the above radar data inversion method, or execute the above marine meteorological detection method.

[0045] Advantageous effects: The present invention uses the strategy of generative adversarial learning to train a radar echo inversion model. Through this radar echo inversion model, radar reflectivity data is inverted based on multi-channel meteorological satellite observation data, solving the problem of lack of marine meteorological radar data caused by the difficulty of building radar base stations at sea. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present specification or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present specification. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0047] Figure 1 It is a flowchart of a radar data inversion method involved in the embodiments of the present specification.

[0048] Figure 2 It is a flowchart of the radar data inversion method in an implementation scenario involved in the embodiments of the present specification.

[0049] Figure 3 It is a flowchart of a training sample construction method involved in the embodiments of the present specification.

[0050] Figure 4 It is an observation image of the Fengyun-4 satellite involved in the embodiments of the present specification.

[0051] Figure 5 It is a schematic diagram of the training process of a generative adversarial model involved in the embodiments of the present specification.

[0052] Figure 6 It is a schematic structural diagram of a radar data inversion device involved in the embodiments of this specification.

[0053] Figure 7 It is a schematic flow diagram of a maritime meteorological detection method involved in the embodiments of this specification.

[0054] Figure 8 It is a schematic structural diagram of a maritime meteorological detection device involved in the embodiments of this specification.

[0055] Figure 9 It is a schematic structural diagram of an electronic device involved in the embodiments of this specification. Detailed implementation manners

[0056] First of all, it should be noted that the terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms "a", "the" and "said" used in the embodiments of the present invention and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise.

[0057] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described here without departing from the scope and spirit of the present invention. Similarly, for the sake of clarity and conciseness, the descriptions of well-known functions and structures are omitted below.

[0058] It should be noted that: in other embodiments, the steps of the corresponding methods are not necessarily executed in the order shown and described in this specification. In some other embodiments, the steps included in the method may be more or less than those described in this specification. In addition, a single step described in this specification may be decomposed into multiple steps for description in other embodiments; and multiple steps described in this specification may also be combined into a single step for description in other embodiments.

[0059] This specification aims to propose a solution to the problem of the lack of maritime meteorological radar data caused by the difficulty of building radar stations at sea currently. One or more embodiments of this specification propose a radar data inversion method, a maritime meteorological detection method and device, using the strategy of generative adversarial learning to train a radar echo inversion model, and realizing the inversion of radar reflectivity data based on multi-channel meteorological satellite observation data through this radar echo inversion model to make up for the deficiency of the current lack of maritime meteorological radar data.

[0060] The radar data inversion method, the marine meteorological detection method and device described in one or more embodiments of this specification will be further described in detail below in conjunction with the accompanying drawings of the specification and specific embodiments, but this detailed description does not constitute a limitation on the embodiments of this specification.

[0061] Please refer to Figure 1 , Figure 1 which schematically shows a flowchart of a radar data inversion method. The method includes steps S100 to S102:

[0062] S100: Obtain multi-channel meteorological satellite observation data of the target area.

[0063] The above multi-channel meteorological satellite observation data can be collected from the observation data of a single meteorological satellite or from the observation data of a satellite network composed of multiple meteorological satellites. The number and positions of the multi-channel meteorological satellites can be adaptively selected according to specific requirements, and this embodiment does not limit this.

[0064] S102: Input the multi-channel meteorological satellite observation data into a pre-trained radar echo inversion model to obtain the predicted radar reflectivity data of the target area.

[0065] The above radar echo inversion model is trained by adopting a generative adversarial learning strategy on a pre-collected training sample set. The training sample set here includes multiple training samples. Each training sample is the multi-channel meteorological satellite observation data of the observation area at the previous moment, and the label of the training sample is the radar reflectivity network observation data of the same observation area at the next moment.

[0066] Considering that it is difficult to obtain marine meteorological radar observation data, in this embodiment, ground multi-channel meteorological satellite observation data is used to train the radar echo inversion model, and then the trained radar echo inversion model is migrated to the marine meteorological observation scenario. Therefore, when constructing the above training samples, the following method can be adopted:

[0067] Collect the observation data of the multi-channel meteorological satellite and radar network deployed in the same ground observation area.

[0068] Perform equatorial and meridian projection conversion on the observation data of the multi-channel meteorological satellite, and perform interpolation processing on the missing values of the converted projection data to obtain the multi-channel meteorological satellite observation data.

[0069] Perform mosaicking on the observation data of the radar network to obtain radar network mosaic data.

[0070] Perform spatio-temporal matching on the multi-channel meteorological satellite observation data and the radar network mosaic data to obtain the radar reflectivity network observation data of the same observation area at the next moment.

[0071] Based on the above training samples, the above radar echo inversion model can be trained in the following manner, specifically including:

[0072] Input the training samples into a generator composed of the radar echo inversion model to obtain the predicted radar reflectivity data.

[0073] Based on the predicted radar reflectivity data and the labels of the training samples, determine the generator loss function.

[0074] Input the predicted radar reflectivity data and the true radar reflectivity data into a discriminator to obtain the discrimination result of the discriminator for the input data.

[0075] According to the discrimination result of the discriminator for the input data and the true value label of the input data, determine the discriminator loss function; the true value label here is used to characterize whether the input data is the true radar reflectivity data.

[0076] Update the parameters of the radar echo inversion model based on the generator loss function and the discriminator loss function.

[0077] In the above solution, a generative adversarial learning strategy is adopted to train the radar echo inversion model. Therefore, when constructing the loss function, the generator loss and the discriminator loss are considered. In addition, a style transfer loss can be introduced during the training process to optimize the above radar echo inversion model. Style transfer is a deep learning technique mainly applied to image processing. It can transform the content of one image into the style of another image to generate a new image. Style transfer can, to a certain extent, make the inverted radar echo similar to the true echo in the horizontal structure, and the inversion result is more similar to the real weather system. Based on this, in some embodiments, a style transfer loss function can be constructed to enable the model to focus on the style features in the training samples. Specifically, the style transfer loss function can be constructed in the following manner:

[0078] For each training sample, determine a first feature image according to the radar reflectivity data predicted by the generator for the training sample.

[0079] Determine a second feature image according to the label of the training sample.

[0080] Determine the style loss function according to the difference in image style between the first feature image and the second feature image.

[0081] The above style loss function is a loss function used to measure the style similarity between the generated image and the specified style image. In this embodiment, the Gram matrix can be used to represent the style features of the image. The calculation method of the Gram matrix is to regard the number of channels of the feature map as one dimension, reshape the feature map into a two-dimensional matrix, and then calculate the product of the transposed matrix of the matrix and itself. The result obtained is the Gram matrix.

[0082] The above style loss function is constructed using the Gram matrix. The specific method is as follows:

[0083] Use the first Gram matrix to represent the style features of the first feature image;

[0084] Use the second Gram matrix to represent the style features of the second feature image;

[0085] Construct the style loss function according to the first Gram matrix and the second Gram matrix:

[0086]

[0087] where L style represents the style loss, C represents the number of channels of the first feature image and the second feature image, H represents the height of the first feature image and the second feature image, W represents the width of the first feature image and the second feature image, represents the element value of the i-th row and j-th column of the first Gram matrix, represents the element value of the i-th row and j-th column of the second Gram matrix.

[0088] After introducing the style loss function, the parameters of the radar echo inversion model can be updated according to the style loss function, the generation loss function, and the discriminant loss function. Specifically, the style loss function, the generation loss function, and the discriminant loss function can be weighted and summed to obtain a total loss function, and then this total loss function is used to perform gradient update on the parameters of the radar echo inversion model to achieve end-to-end training of the above radar echo inversion model.

[0089] The following combines a specific application scenario to further illustrate the above radar data inversion method.

[0090] Please refer to Figure 2 , Figure 2 which shows the flowchart of the above radar data inversion method in an implementation scenario.

[0091] In this scenario, this embodiment selects the FY-4 satellite data in Jiangsu region in the past five years and the radar reflectivity network observation data of 9 existing dual-polarization meteorological radars in Jiangsu region at corresponding times, and then performs spatio-temporal matching on the satellite observation data and the radar observation data to construct a training sample set. Please refer to Figure 3 , for the construction of the training samples, the Figure 3 shown process can be adopted to achieve.

[0092] As Figure 4 shown, the FY-4 satellite has 14 observation channels with different wavelengths. Among them, 2 visible channels have no data at night and cannot be used. Here, the observation data of 2 water vapor window channels and 4 atmospheric transparency window channels (B09 to B14) are selected as the training samples.

[0093] Before inputting the above training samples into the radar echo inversion model, it is also necessary to preprocess the observation data of channels B09 to B14 and convert these observation data into a data form that meets the requirements of the input layer format of the radar echo inversion model. Specifically, the full-disk data of the brightness temperature of different bands of channels B09 to B14 can be converted by equal longitude and latitude projection. For the missing values in the projection data, they can be filled by interpolation. For example, the projection data can be uniformly interpolated to the grid points with an equal longitude and latitude spacing of 0.04°×0.04° to form the input data of the radar echo inversion model.

[0094] Correspondingly, it is also necessary to perform quality control such as deblurring and clutter elimination on the observation data of the above 9 dual-polarization meteorological radars and then network them across the province to obtain the radar network reflectivity data.

[0095] Finally, perform time downscaling matching on the networked mosaic reflectivity data of the above 9 dual-polarization meteorological radars and the preprocessed observation data of the FY-4 satellite. Here, mainly find the networked mosaic reflectivity data that is closest in time according to the satellite observation data for time matching, and use the matched networked mosaic reflectivity data as the label of the corresponding satellite observation data.

[0096] After obtaining the training sample set, use the above training sample set to train the radar echo inversion model. Please refer to Figure 5 , Figure 5The training process of the generative adversarial model is shown. In generative adversarial training, there are two models. One is the generator composed of the radar echo inversion model, and the other is the discriminator. The generator generates predicted radar reflectivity data based on the input training samples. The predicted radar reflectivity data and the real radar reflectivity data are input into the discriminator together, and the discriminator is used to distinguish whether the input data is the generated radar reflectivity data or the real radar reflectivity data. The task of the generator is to generate as realistic radar reflectivity data as possible to deceive the discriminator, and the task of the discriminator is to identify the authenticity of the input data as accurately as possible. The two are in an adversarial relationship. Through generative adversarial training, the generator can pay more attention to the feature information related to the real radar reflectivity data during the learning process.

[0097] Please continue to refer to Figure 2 , in Figure 2 , SEResUnet is used as the above-mentioned radar echo inversion model. The channel attention mechanism and residual mechanism of SEResUnet can better adapt to multi-channel service scenarios, have good feature extraction ability for multi-channel satellite data, and can better distinguish the features related to the inverted radar echo data between different channels.

[0098] The radar data inversion method provided in this embodiment can accurately and effectively implement radar data inversion, can be widely applied to areas without meteorological radars at sea, and can provide assistance for meteorological guidance and weather emergency warning in fields such as ocean navigation, fishery fishing, and maritime combat drills.

[0099] Corresponding to the above radar data inversion method, this embodiment also proposes a radar data inversion device. It should be noted that Figure 1 the radar data inversion shown can be performed by this device, but is not limited to this device. Please refer to Figure 6 , the device includes:

[0100] The first data acquisition module 601 is configured to acquire multi-channel meteorological satellite observation data of the target area.

[0101] The first inversion module 602 is configured to input the multi-channel meteorological satellite observation data into a pre-trained radar echo inversion model to obtain the predicted radar reflectivity data of the target area.

[0102] For the first inversion module 602, the radar echo inversion model here is trained using a generative adversarial learning strategy on a pre-collected training sample set. The training sample set includes multiple training samples. The training sample is the multi-channel meteorological satellite observation data of the previous moment in the observation area, and the label of the training sample is the radar reflectivity network observation data of the next moment in the same observation area.

[0103] The above radar echo inversion model can be directly trained by the first inversion module 602. It can also be pre-trained and saved in the storage module, and loaded by the first inversion module 602 from the storage module.

[0104] For the training process of the radar echo inversion model, the corresponding scheme in the above radar data inversion method can be adopted, which will not be elaborated here.

[0105] For the above radar data inversion device, taking the module as an example of a software functional unit, the first data acquisition module 601 may include code running on a computing instance. Among them, the computing instance may include at least one of a physical host (computing device), a virtual machine, and a container. Further, the above computing instance may be one or more. For example, the first data acquisition module 601 may include code running on multiple hosts / virtual machines / containers. The multiple hosts / virtual machines / containers for running this code may be distributed in the same region, or may be distributed in different regions. Further, the multiple hosts / virtual machines / containers for running this code may be distributed in the same availability zone (AZ), or may be distributed in different AZs, and each AZ includes one data center or multiple geographically proximate data centers. Among them, generally one region may include multiple AZs.

[0106] Similarly, the multiple hosts / virtual machines / containers for running this code may be distributed in the same virtual private cloud (VPC), or may be distributed in multiple VPCs. Among them, generally one VPC is set within one region. For cross-region communication between two VPCs within the same region and between VPCs in different regions, a communication gateway needs to be set in each VPC, and the interconnection between VPCs is realized through the communication gateway.

[0107] As an example of a hardware functional unit, the first data acquisition module 601 includes at least one computing device, such as a server, etc. Alternatively, the first data acquisition module 601 can also be a device implemented using an application-specific integrated circuit (ASIC) or a programmable logic device (PLD). Among them, the above PLD can be implemented by a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0108] The multiple computing devices included in the first data acquisition module 601 can be distributed in the same region or in different regions. The multiple computing devices included in the first data acquisition module 601 can be distributed in the same availability zone (AZ) or in different AZs. Similarly, the multiple computing devices included in the first data acquisition module 301 can be distributed in the same virtual private cloud (VPC) or in multiple VPCs. Among them, the multiple computing devices can be any combination of computing devices such as servers, ASICs, PLDs, CPLDs, FPGAs, and GALs.

[0109] In other embodiments, the first data acquisition module 601 can be used to execute any step in the above radar data inversion method, and the first inversion module 602 can be used to execute any step in the above radar data inversion method. The steps to be implemented by the first data acquisition module 601 and the first inversion module 602 can be specified as needed, and all functions of the above radar data inversion device can be implemented by respectively implementing different steps in the above radar data inversion method through the first data acquisition module 601 and the first inversion module 602.

[0110] In this implementation manner, the radar data inversion device can also be applied to computing devices such as computers and servers, or to a computing device cluster including at least one computing device to implement the radar data inversion function.

[0111] This embodiment also proposes a marine weather detection method. Please refer to Figure 7 , and this method includes steps S700 to S704:

[0112] S700: Obtain multi-channel meteorological satellite observation data of the sea area to be detected at the current moment.

[0113] S702: Based on the multi-channel meteorological satellite observation data of the area to be detected at the current moment, use the radar data inversion method to invert the radar reflectivity data of the area to be detected at the next moment.

[0114] S704: Conduct meteorological detection based on the radar reflectivity data of the area to be detected at the next moment. For example, in the identification and detection of severe convection, the strong radar reflectivity echoes obtained by inversion can be processed into a gray-level co-occurrence matrix. By extracting the morphological and intensity information of the strong radar reflectivity echoes, calculate the longitude and latitude of the centroid of each strong echo, the maximum value of the strong echo, the average value of the strong echo, the area of the strong echo, the radius of the strong echo, and the eccentricity of the strong echo, so as to identify the severe convection area.

[0115] In addition, considering the limitation of the fixed input resolution of the radar echo inversion model, when directly applying the radar echo inversion model trained based on land observation data to invert the radar echo in a larger area at sea, Gaussian smoothing needs to be performed on the edges of the spliced multiple inversion areas.

[0116] Corresponding to the above-mentioned marine meteorological detection method, this embodiment also proposes a marine meteorological detection device. Please refer to Figure 8 and this device includes:

[0117] The second data acquisition module 801 is configured to acquire the multi-channel meteorological satellite observation data of the area to be detected at the current moment at sea.

[0118] The second inversion module 802 is configured to, based on the multi-channel meteorological satellite observation data of the area to be detected at the current moment, use the radar data inversion method to invert the radar reflectivity data of the area to be detected at the next moment.

[0119] The meteorological detection module 803 is configured to conduct meteorological detection based on the radar reflectivity data of the area to be detected at the next moment.

[0120] For the above-mentioned marine meteorological detection device, taking the module as an example of a software functional unit, the second data acquisition module 801 may include code running on a computing instance. The computing instance may include at least one of a physical host (computing device), a virtual machine, and a container. Further, the above-mentioned computing instance may be one or more. For example, the second data acquisition module 801 may include code running on multiple hosts / virtual machines / containers. The multiple hosts / virtual machines / containers for running the code may be distributed in the same region or in different regions. Further, the multiple hosts / virtual machines / containers for running the code may be distributed in the same availability zone (AZ) or in different AZs, and each AZ includes one data center or multiple geographically proximate data centers. Usually, one region may include multiple AZs.

[0121] Similarly, the multiple hosts / virtual machines / containers for running the code may be distributed in the same virtual private cloud (VPC) or in multiple VPCs. Usually, one VPC is set within one region. For cross-region communication between two VPCs within the same region and between VPCs in different regions, a communication gateway needs to be set in each VPC, and the interconnection between VPCs is achieved through the communication gateway.

[0122] Taking the module as an example of a hardware functional unit, the second data acquisition module 801 includes at least one computing device, such as a server, etc. Alternatively, the second data acquisition module 801 may also be a device implemented by an application-specific integrated circuit (ASIC) or a programmable logic device (PLD). Among them, the above-mentioned PLD may be implemented by a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0123] The multiple computing devices included in the second data acquisition module 801 can be distributed in the same region or in different regions. The multiple computing devices included in the second data acquisition module 801 can be distributed in the same availability zone (AZ) or in different AZs. Similarly, the multiple computing devices included in the second data acquisition module 801 can be distributed in the same virtual private cloud (VPC) or in multiple VPCs. Among them, the multiple computing devices can be any combination of computing devices such as servers, application-specific integrated circuits (ASICs), programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), and generic array logic (GALs).

[0124] In other embodiments, the second data acquisition module 801 can be used to perform any step in the above-mentioned marine weather detection method, the second inversion module 802 can be used to perform any step in the above-mentioned marine weather detection method, and the weather detection module 803 can be used to perform any step in the above-mentioned marine weather detection method. The steps to be implemented by the second data acquisition module 801, the second inversion module 802, and the weather detection module 803 can be specified as needed. The above-mentioned entire function of the marine weather detection device is realized by respectively implementing different steps in the above-mentioned marine weather detection method through the second data acquisition module 801, the second inversion module 802, and the weather detection module 803.

[0125] In this implementation, the marine weather detection device can also be applied to computing devices such as computers and servers, or to a computing device cluster including at least one computing device to implement the marine weather detection function.

[0126] In some embodiments, an electronic device is also provided. Please refer to Figure 9 This electronic device includes: a bus 901, a processor 902, a memory 903, and a communication interface 904. The processor 902, the memory 903, and the communication interface 904 communicate with each other through the bus 901. It should be understood that the present application does not limit the number of processors and memories in the electronic device.

[0127] The bus 901 can be a peripheral component interconnect (PCI) bus, an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, Figure 9 only one line is used to represent it, but it does not mean that there is only one bus or one type of bus. The bus 901 can include a path for transmitting information between various components of the electronic device (for example, the processor 902, the memory 903, and the communication interface 904).

[0128] The processor 902 may include any one or more of processors such as a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP).

[0129] The memory 903 may include volatile memory, such as random access memory (RAM). The memory 903 may also include non-volatile memory, such as read-only memory (ROM), flash memory, a hard disk drive (HDD), or a solid state drive (SSD).

[0130] The memory 903 stores executable program code, and the processor 902 executes the executable program code to respectively implement the functions of the foregoing first data acquisition module 601 and first inversion module 602, that is, to implement the functions of the foregoing radar data inversion device, thereby implementing the foregoing radar data inversion method. Alternatively, the processor 902 executes the executable program code to respectively implement the functions of the foregoing second data acquisition module 801, second inversion module 802, and meteorological detection module 803, that is, to implement the functions of the foregoing marine meteorological detection device, thereby implementing the foregoing marine meteorological detection method.

[0131] The communication interface 904 uses a transceiver module such as, but not limited to, a network interface card or a transceiver to implement communication between the electronic device and other devices or a communication network.

[0132] In some embodiments, a computer-readable storage medium is also provided, on which a computer program is stored. When the computer program is executed by a processor, the foregoing radar data inversion method is implemented, or the foregoing marine meteorological detection method is implemented.

[0133] The computer-readable storage medium may be any available medium that the electronic device can store or a data storage device such as a data center that includes one or more available media. The available medium may be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state drive), etc. The computer-readable storage medium includes instructions that direct the electronic device to execute the foregoing radar data inversion method, or to execute the foregoing marine meteorological detection method.

[0134] It is to be understood that the structure illustrated in the embodiments of this specification does not constitute a specific limitation on the system of the embodiments of this specification. In other embodiments of the specification, the above system may include more or fewer components than shown in the figure, or combine some components, or split some components, or arrange the components differently. The components shown in the figure may be implemented in hardware, software, or a combination of software and hardware.

[0135] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0136] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0137] It should be noted that the above examples are only specific embodiments of the present invention, and the present invention is obviously not limited to the above examples, and there are many similar variations. All variations directly derived or associated from the contents disclosed by the technicians in this field should fall within the protection scope of the present invention.

Claims

1. A radar data inversion method, characterized in that: include: Acquire multi-channel meteorological satellite observation data of the target area; Inputting the multi-channel meteorological satellite observation data into a pre-trained radar echo inversion model to obtain predicted radar reflectivity data of the target area; The training method of the radar echo inversion model specifically includes: Inputting training samples into a generator composed of the radar echo inversion model to obtain predicted radar reflectivity data; the training samples are multi-channel meteorological satellite observation data of the observation area at the previous moment, and the labels of the training samples are radar reflectivity network observation data of the same observation area at the next moment; Determining a loss function based on the predicted radar reflectivity data and the labels of the training samples; Inputting the predicted radar reflectivity data and the real radar reflectivity data into a discriminator to obtain a discriminant result of the discriminator for the input data; Determining a discrimination loss function according to a discrimination result of the discriminator on the input data and a true value label of the input data; the true value label is used to characterize whether the input data is true radar reflectivity data; For each training sample, determining a first feature image according to radar reflectivity data predicted by the generator for the training sample; Determining a second feature image according to the label of the training sample; Using a first Gram matrix to represent the style features of the first feature image; Using a second Gram matrix to represent the style features of the second feature image; According to the first Gram matrix and the second Gram matrix, a style loss function is constructed: Among them, L style represents the style loss, C represents the number of channels of the first feature image and the second feature image, H represents the height of the first feature image and the second feature image, W represents the width of the first feature image and the second feature image, represents the element value of the i-th row and j-th column of the first Gram matrix, represents the element value of the i-th row and j-th column of the second Gram matrix; The parameters of the radar echo inversion model are updated according to the style loss function, the generation loss function and the discrimination loss function.

2. The method according to claim 1, characterized in that Constructing the training sample specifically includes: Collect observation data from multi-channel meteorological satellites and radar networks deployed in the same observation area; Performing equal longitude and latitude projection conversion on the observation data of the multi-channel meteorological satellite, and interpolating the missing values ​​of the converted projection data to obtain the multi-channel meteorological satellite observation data; Piecing together the observation data of the radar network to obtain radar network puzzle data; The multi-channel meteorological satellite observation data and the radar network mosaic data are temporally and spatially matched to obtain the radar reflectivity network observation data of the same observation area at the next moment.

3. A radar data inversion device, used to implement the method according to any one of claims 1 to 2, characterized in that: include: A first data acquisition module is configured to acquire multi-channel meteorological satellite observation data of a target area; A first inversion module is configured to input the multi-channel meteorological satellite observation data into a pre-trained radar echo inversion model to obtain predicted radar reflectivity data of the target area; The radar echo inversion model is trained using a generative adversarial learning strategy on a pre-collected training sample set; the training sample set includes multiple training samples, the training samples are multi-channel meteorological satellite observation data of the observation area at the previous moment, and the labels of the training samples are radar reflectivity network observation data of the same observation area at the next moment.

4. A method for detecting marine weather, characterized in that: include: Obtain multi-channel meteorological satellite observation data of the current time of the marine area to be detected; According to the multi-channel meteorological satellite observation data of the area to be detected at the current moment, the radar reflectivity data of the area to be detected at the next moment is inverted using the radar data inversion method described in any one of claims 1 to 2; Meteorological detection is performed based on the radar reflectivity data of the area to be detected at the next moment.

5. A marine weather detection device, characterized in that: include: The second data acquisition module is configured to acquire multi-channel meteorological satellite observation data of the marine area to be detected at the current moment; A second inversion module is configured to invert the radar reflectivity data of the area to be detected at a next moment using the radar data inversion method according to any one of claims 1 to 2 based on the multi-channel meteorological satellite observation data of the area to be detected at a current moment; The meteorological detection module is configured to perform meteorological detection according to the radar reflectivity data of the area to be detected at the next moment.

6. A computer-readable storage medium storing a computer program, wherein when the computer program is executed on an electronic device, the electronic device executes the method according to any one of claims 1 to 2, or executes the method according to claim 4.

7. An electronic device comprising: at least one memory for storing a program; At least one processor is used to execute the program stored in the memory. When the program stored in the memory is executed, the processor is used to execute the method according to any one of claims 1 to 2, or execute the method according to claim 4.

Images