A typhoon positioning and intensity determination method and device based on remote sensing data

Abstract

The present application provides a typhoon positioning and intensity determination method and device based on remote sensing data, which relates to the technical field of remote sensing image processing. Specifically, it includes: obtaining a remote sensing image, preprocessing the remote sensing image to obtain a typhoon area image; using a pre-trained typhoon prediction model to process the typhoon area image to obtain the typhoon center position and typhoon intensity; the network structure of the typhoon prediction model is a ResNet that integrates the InceptionNet network structure. The present application improves the prediction accuracy of typhoon center position and typhoon intensity information.

Description

Technical Field

[0001] This application relates to the technical field of remote sensing image processing, and in particular to a typhoon positioning and intensity determination method and device based on remote sensing data. Background Art

[0002] In the field of remote sensing, true color images refer to images obtained by performing RGB synthesis on data from three channels with central wavelengths in the red, green, and blue ranges. True color images are closest to the prior information directly observed by the human eye, are most easily understood by image interpreters, and are intuitive and fast for identifying vegetation changes, cloud details, water eutrophication, fire point plumes, burned area identification, etc.

[0003] The AGRI imager sensor carried by the FY4A geostationary satellite includes two visible light channels, namely the blue light channel (central wavelength 470nm) and the red light channel (central wavelength 650nm). This instrument does not have a green light channel, which is equivalent to lacking the middle G band during RGB synthesis, so true color image synthesis cannot be performed. Summary of the Invention

[0004] In view of this, this application provides a typhoon positioning and intensity determination method and device based on remote sensing data to solve the above technical problems.

[0005] In a first aspect, an embodiment of this application provides a typhoon positioning and intensity determination method based on remote sensing data, including:

[0006] Obtain a remote sensing image, and preprocess the remote sensing image to obtain a typhoon area image;

[0007] Use a pre-trained typhoon prediction model to process the typhoon area image to obtain the typhoon center position and typhoon intensity; the network structure of the typhoon prediction model is a ResNet that integrates the InceptionNet network structure.

[0008] Further, preprocessing the remote sensing image to obtain a typhoon area image includes:

[0009] Convert the remote sensing image into a grayscale image;

[0010] Crop an initial typhoon area image of a preset size from the grayscale image;

[0011] Expand the initial typhoon area image into a square image according to the longest side, and fill the expanded area with black;

[0012] Rotate, add noise, displace, and change the brightness of the expanded square image to obtain a typhoon area image.

[0013] Further, the typhoon prediction model includes a first processing module, a second processing module, a third processing module, a fourth processing module, a max pooling layer, and a linear layer connected in sequence;

[0014] The first processing module is used to perform dimensionality reduction processing and feature extraction on the input typhoon area image, and output a first feature map;

[0015] The second processing module is used to process the first feature map and output a second feature map;

[0016] The third processing module is used to process the second feature map and output a third feature map;

[0017] The fourth processing module is used to process the third feature map and output a fourth feature map;

[0018] The max pooling layer is used to perform a max pooling operation on the fourth feature map and output a fifth feature map;

[0019] The linear layer is used to perform a linear transformation on the fifth feature map and output a three-dimensional typhoon information vector, where the three-dimensional typhoon information vector includes: the longitude of the typhoon center position, the latitude of the typhoon center position, and intensity information.

[0020] Further, the first processing module sequentially includes: a first unit, a second unit, a third unit, and a fourth unit; among them, the second unit and the fourth unit have the same structure;

[0021] The first unit includes a 3x3 convolutional layer and a 5x5 convolutional layer connected in series;

[0022] The second unit includes: a 3x3 average pooling layer and a 3x3 convolutional layer connected in parallel, and a filtering connection layer for synthesizing the output results of the two paths;

[0023] The third unit includes three branches and a filtering connection layer for synthesizing the output results of the three branches. Among them, the first branch is a direct output, the second branch includes a 3x3 convolutional layer and a 1x1 convolutional layer connected in series, and the third branch includes a 1x1 convolutional layer, a 5x1 convolutional layer, a 1x5 convolutional layer, and a 5x5 convolutional layer connected in series.

[0024] Further, the second processing module includes: 3 ResNet-A modules connected in series and 1 Reduction-1 module;

[0025] The ResNet-A module includes: a first activation function layer, a fifth unit, and a second activation function layer;

[0026] The fifth unit includes a first branch, a second branch, a third branch, a fourth branch, a fifth branch, a 1x1 convolutional layer, and an adder; the 1x1 convolutional layer processes the outputs of the third branch, the fourth branch, and the fifth branch, and the adder accumulates the outputs of the first branch, the second branch, and the 1x1 convolutional layer and outputs them to the second activation function layer; the first branch includes a 1x7 convolutional layer and a 7x1 convolutional layer connected in series; the second branch is a direct output; the third branch includes a 1x1 convolutional layer; the fourth branch includes a 1x1 convolutional layer and a 3x3 convolutional layer connected in series; the fifth branch includes a 1x1 convolutional layer, a 3x3 convolutional layer, and a 3x3 convolutional layer connected in series;

[0027] The Reduction-1 module includes: a filtering connection layer, a sixth unit, and a filtering connection layer;

[0028] The sixth unit includes three branches, where the first branch includes a 5x5 max pooling layer; the second branch includes a 3x3 convolutional layer; the third branch includes a 1x1 convolutional layer and a 3x3 convolutional layer connected in series.

[0029] Furthermore, the third processing module includes 5 ResNet-B modules connected in series and 1 Reduction-2 module;

[0030] The ResNet-B module includes a third activation function layer, a seventh unit, and a fourth activation function layer;

[0031] The seventh unit includes a first branch, a second branch, a third branch, a fourth branch, a 1x1 convolutional layer, and an adder. The 1x1 convolutional layer processes the outputs of the third branch and the fourth branch; the adder accumulates the outputs of the first branch, the second branch, and the 1x1 convolutional layer and inputs them to the fourth activation function layer; among them, the first branch includes a 1x5 convolutional layer and a 5x1 convolutional layer connected in series; the second branch is an unprocessed direct output; the third branch includes a 1x1 convolutional layer; the fourth branch includes a 1x1 convolutional layer and a 3x3 convolutional layer connected in series;

[0032] Reduction-2 includes: a filtering connection layer, 4 branches, and a filtering connection layer; among them, the first branch includes a 5x5 max pooling layer, the second branch includes a 1x1 convolutional layer and a 3x3 convolutional layer connected in series; the third branch includes a 1x1 convolutional layer and a 3x3 convolutional layer; the fourth branch includes a 1x1 convolutional layer, a 3x3 convolutional layer, and a 3x3 convolutional layer connected in series.

[0033] Furthermore, the ResNet-C module includes a fifth activation function layer, a seventh unit, and a sixth activation function layer;

[0034] The seventh unit includes a first branch, a second branch, a third branch, a 1x1 convolutional layer, and an adder; the 1x1 convolutional layer processes the outputs of the second branch and the third branch; the adder accumulates the outputs of the first branch and the 1x1 convolutional layer and inputs them to the sixth activation function layer; the first branch outputs directly without being processed; the second branch includes a 1x1 convolutional layer; the third branch includes a 1x1 convolutional layer and a 3x3 convolutional layer.

[0035] Furthermore, the training process of the typhoon prediction model includes:

[0036] Obtain the remote sensing data of all typhoon moments in the typhoon best track dataset, convert the remote sensing data into grayscale images, maintain the inherent resolution, and intercept the typhoon area sample images of a preset size from the grayscale images;

[0037] Preprocess all typhoon area sample images and label the typhoon center position and typhoon intensity information;

[0038] Use the typhoon prediction model to process each typhoon area sample image to obtain the predicted typhoon center position and typhoon intensity information;

[0039] Calculate the loss function according to the predicted typhoon center position and typhoon intensity information and the labeled typhoon center position and typhoon intensity information

[0040]

[0041] Among them, the vector y = (y 1 , y 2 , y 3 ), where (y 1 , y 2 ) are the longitude and latitude of the predicted typhoon center position, and y 3 is the predicted typhoon intensity information; the vector Among them are the longitude and latitude of the labeled typhoon center position, is the labeled typhoon intensity information;

[0042] Based on the loss function value and the batch gradient descent method, update the model parameters of the typhoon prediction model.

[0043] In a second aspect, an embodiment of the present application provides a typhoon positioning and intensity determination device based on remote sensing data, including:

[0044] An acquisition unit, configured to acquire a remote sensing image and preprocess the remote sensing image to obtain a typhoon area image;

[0045] A processing unit is configured to process an image of a typhoon area by using a pre-trained typhoon prediction model to obtain the typhoon center position and typhoon intensity; the network structure of the typhoon prediction model is a ResNet that integrates the InceptionNet network structure.

[0046] In a third aspect, an embodiment of the present application provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where when the processor executes the computer program, it implements the typhoon positioning and intensity determination method based on remote sensing data in the embodiments of the present application.

[0047] The present application improves the prediction accuracy of typhoon center position and typhoon intensity information. Description of the Drawings

[0048] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0049] Figure 1 It is a flowchart of the typhoon positioning and intensity determination method based on remote sensing data provided by an embodiment of the present application;

[0050] Figure 2 It is a schematic diagram of the main ResNet-I network structure adopted by the typhoon prediction model provided by an embodiment of the present application;

[0051] Figure 3 It is a schematic diagram of the Stem structure provided by an embodiment of the present application;

[0052] Figure 4 It is a schematic diagram of the ResNet-A structure provided by an embodiment of the present application;

[0053] Figure 5 It is a schematic diagram of the Reduction-1 structure provided by an embodiment of the present application;

[0054] Figure 6 It is a schematic diagram of the ResNet-B structure provided by an embodiment of the present application;

[0055] Figure 7 It is a schematic diagram of the Reduction-2 structure provided by an embodiment of the present application;

[0056] Figure 8 It is a schematic diagram of the ResNet-C structure provided by an embodiment of the present application;

[0057] Figure 9 This is the functional structure diagram of the typhoon positioning and intensity determination device based on remote sensing data provided by the embodiments of the present application;

[0058] Figure 10 This is the structure diagram of the electronic device provided by the embodiments of the present application. Detailed implementation manners

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are some, but not all, of the embodiments of the present application. Usually, the components of the embodiments of the present application described and illustrated herein can be arranged and designed in various different configurations.

[0060] Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but merely represents selected embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0061] First, a brief introduction to the design concept of the embodiments of the present application will be given.

[0062] The landing of typhoon strong weather often brings huge economic losses to people. Identifying and tracking typhoons faster and more accurately and timely forecasting the typhoon intensity and the typhoon center position can effectively reduce people's property losses. Currently, most typhoon positioning and intensity determination methods rely on forecast data, and the timeliness and accuracy are relatively poor.

[0063] To solve the above technical problems, the present application proposes a typhoon positioning and intensity determination method based on remote sensing data. By establishing a typhoon prediction model with a new network structure, the typhoon area image is processed, and the typhoon center position and typhoon intensity information are obtained simultaneously. The present application improves the accuracy of forecasting the typhoon center position and typhoon intensity information.

[0064] After introducing the application scenarios and design concepts of the embodiments of the present application, the technical solutions provided by the embodiments of the present application will be described below.

[0065] As Figure 1 shown, the embodiments of the present application provide a typhoon positioning and intensity determination method based on remote sensing data, including:

[0066] Step 101: Obtain a remote sensing image, and preprocess the remote sensing image to obtain a typhoon image;

[0067] Select FY4A full-disk remote sensing data from 2017 to 2018, combine with historical typhoon message information, select FY4A full-disk data with typhoon message information time, read its 10.8um channel and convert it into a grayscale image; intercept the typhoon area position image;

[0068] Convert the intercepted typhoon area image into an equilateral rectangle according to the longest side, and fill the extended area with black; perform data enhancement on the extended square image by rotating, adding noise, displacement, changing brightness, etc.

[0069] Combine the message information and the enhanced typhoon area image information.

[0070] Step 102: Use the pre-trained typhoon prediction model to process the typhoon images to obtain the eye position information and intensity information of each typhoon image.

[0071] First, this application draws on the characteristics of inceptionNet and ResNet, and proposes a new ResNet-I network by integrating the two network structures; and processes the actual problems of typhoon positioning and intensity determination; the model structure is shown in Figure 2 ;

[0072] The ResNet-I network sequentially includes a Stem, 3 serially connected ResNet-A modules, 1 Reduction-1 module, 5 serially connected ResNet-B modules, 1 Reduction-2 module, 1 ResNet-C module, a Maxpooling layer, and a linear layer; where:

[0073] Stem can be divided into 4 parts in total. The first part is composed of a series connection of 3x3 and 5x5 convolutional layers. The second part has two operations. One is a 3x3 average pooling operation, and the other is a 3x3 convolutional operation. The Filter concat combines the results of the two paths; the third part draws on the resnet idea and divides the network into three paths. One path is directly output without processing, one path is output through a series connection of 3x3 and 1x1 convolutional layers, and one path is output through a series connection of 1x1, 5x1, 1x5, and 5x5 convolutional layers. Then the Filter concat combines the outputs of the three paths; the last part is the same two operations as the second part; as Figure 3 shown;

[0074] ResNet-A includes: an activation function layer (Relu activation), five branches, and an activation function layer (Relu activation); the five branches are: (1) directly output without processing; (2) processed through a 1x1 convolution; (3) processed through a 1x1 convolution and then a 3x3 convolution; (4) processed through a 1x1 convolution and then two 3x3 convolutions; (5) processed through a 1x7 and 7x1 convolution; then the outputs of parts 2, 3, and 4 are combined first and then processed through a 1x1 convolution, and finally combined with the outputs of branches 1 and 5 and output to the activation function layer (Relu activation), as Figure 4 shown.

[0075] Reduction-1 includes: a filtering connection layer, three branches, and a filtering connection layer; the three branches include: (1) processed through a 5x5 max pooling layer; (2) processed through a 3x3 convolution; (3) processed through a 1x1 convolution and then a 3x3 convolution; then the outputs of the three branches are combined and output to the filtering connection layer, as Figure 5 shown.

[0076] ResNet-B includes: an activation function layer (Relu activation), four branches, and an activation function layer (Relu activation); the four branches include: (1) directly output without processing; (2) processed through a 1x1 convolution; (3) processed through a 1x1 convolution and then a 3x3 convolution; (4) processed through a 1x5 and 5x1 convolution; then the outputs of parts 2 and 3 are combined first and then processed through a 1x1 convolution, and finally combined with the outputs of branches 1 and 4 and output to the activation function layer, as Figure 6 shown.

[0077] Reduction-2 includes: a filtering connection layer, four branches, and a filtering connection layer; the four branches include: (1) processed through a 5x5 max pooling layer; (2) processed through a 1x1 and a 3x3 convolution; (3) processed through a 1x1 and a 3x3 convolution; (4) processed through a 1x1 convolution and then two 3x3 convolutions; then the outputs of the four branches are combined and output to the filtering connection layer, as Figure 7 shown.

[0078] ResNet-C includes: an activation function layer (Relu activation), three branches, and an activation function layer (Relu activation); the three branches include: (1) directly output without processing; (2) processed through a 1x1 convolution once; (3) processed through a 1x1 convolution and a 3x3 convolution once; then first combine the outputs of parts 2 and 3, then perform a 1x1 convolution processing, and finally combine with the output of branch 1 and output to the activation function layer, as Figure 8 shown.

[0079] The max pooling layer is used to perform max pooling operation on the fourth feature map and output the fifth feature map;

[0080] The linear layer is used to perform linear transformation on the fifth feature map and output a three-dimensional typhoon information vector, and the three-dimensional typhoon information vector includes: the longitude of the typhoon center position, the latitude of the typhoon center position, and the intensity information.

[0081] After building the typhoon prediction model, train the typhoon prediction model:

[0082] Convert the data of the twelfth channel of FY4A 4KM remote sensing data at all typhoon moments recorded in the typhoon best track dataset from 2017 to 2018 into grayscale images, maintain the original resolution, intercept the typhoon area images, perform image preprocessing, and label the typhoon center position and typhoon intensity information;

[0083] Use the typhoon prediction model to process each typhoon area sample image to obtain the predicted typhoon center position and typhoon intensity information;

[0084] Calculate the loss function according to the predicted typhoon center position and typhoon intensity information and the labeled typhoon center position and typhoon intensity information

[0085]

[0086] Among them, the vector y = (y 1 , y 2 , y 3 ), where (y 1 , y 2 ) are the longitude and latitude of the predicted typhoon center position, and y 3 is the predicted typhoon intensity information; the vector Among them are the longitude and latitude of the labeled typhoon center position, is the labeled typhoon intensity information;

[0087] Based on the loss function value and the Batch Gradient Descent (BGD), update the model parameters of the typhoon prediction model.

[0088] Using the typhoon season data from July to August 2019 for testing, the results show that the model can identify the typhoon center position and intensity information. The average error of typhoon center identification is 1.86 pixels; the R-square of typhoon intensity is 0.94.

[0089] Based on the above embodiments, the embodiments of the present application provide a typhoon positioning and intensity determination device based on remote sensing data. Refer to Figure 9 As shown, the typhoon positioning and intensity determination device 200 based on remote sensing data provided by the embodiments of the present application at least includes:

[0090] An acquisition unit 201, configured to acquire a remote sensing image and preprocess the remote sensing image to obtain a typhoon area image;

[0091] A processing unit 202, configured to process the typhoon area image by using a pre-trained typhoon prediction model to obtain the typhoon center position and typhoon intensity; the network structure of the typhoon prediction model is a ResNet that integrates the InceptionNet network structure.

[0092] It should be noted that the principle of the typhoon positioning and intensity determination device 200 based on remote sensing data provided by the embodiments of the present application to solve technical problems is similar to that of the typhoon positioning and intensity determination method based on remote sensing data provided by the embodiments of the present application. Therefore, for the implementation of the typhoon positioning and intensity determination device 200 based on remote sensing data provided by the embodiments of the present application, reference can be made to the implementation of the typhoon positioning and intensity determination method based on remote sensing data provided by the embodiments of the present application, and the repeated parts will not be elaborated.

[0093] As Figure 10 shown, the electronic device 300 provided by the embodiments of the present application at least includes: a processor 301, a memory 302, and a computer program stored on the memory 302 and executable on the processor 301. When the processor 301 executes the computer program, it implements the typhoon positioning and intensity determination method based on remote sensing data provided by the embodiments of the present application.

[0094] The electronic device 300 provided by the embodiments of the present application may further include a bus 303 connecting different components (including the processor 301 and the memory 302). Among them, the bus 303 represents one or more of several types of bus structures, including a memory bus, a peripheral bus, a local bus, etc.

[0095] The memory 302 may include a readable medium in the form of volatile memory, such as a Random Access Memory (RAM) 3021 and / or a cache memory 3022, and may further include a Read Only Memory (ROM) 3023.

[0096] The memory 302 may also include a program tool 3024 having a set (at least one) of program modules 3025, and the program modules 3025 include, but are not limited to: an operating subsystem, one or more application programs, other program modules, and program data. Implementations of a network environment may be included in each or some combination of these examples.

[0097] The electronic device 300 may also communicate with one or more external devices 304 (such as a keyboard, a remote control, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 300 (such as a mobile phone, a computer, etc.), and / or communicate with any device that enables the electronic device 300 to communicate with one or more other electronic devices 300 (such as a router, a modem, etc.). Such communication may be performed through an Input / Output (I / O) interface 305. And, the electronic device 300 may also communicate with one or more networks (such as a Local Area Network (LAN), a Wide Area Network (WAN), and / or a public network, such as the Internet) through a network adapter 306. As Figure 10 shown, the network adapter 306 communicates with other modules of the electronic device 300 through a bus 303. It should be understood that although Figure 10 not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 300, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, Redundant Arrays of Independent Disks (RAID) subsystems, tape drives, and data backup storage subsystems, etc.

[0098] It should be noted that Figure 10 the illustrated electronic device 300 is merely an example and should not impose any limitation on the functions and usage scope of the embodiments of the present application.

[0099] The embodiments of the present application also provide a computer-readable storage medium storing computer instructions, and when the computer instructions are executed by a processor, the typhoon positioning and intensity determination method based on remote sensing data provided by the embodiments of the present application is implemented.

[0100] In addition, although the operations of the method of the present application are described in a specific order in the drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step for execution, and / or one step may be decomposed into multiple steps for execution.

[0101] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications to these embodiments once they learn the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the present application.

[0102] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for typhoon positioning and intensity determination based on remote sensing data, characterized in that, it includes: Obtain remote sensing images, and preprocess the remote sensing images to obtain typhoon area images; Use the pre-trained typhoon prediction model to process the typhoon area images to obtain the typhoon center position and typhoon intensity; the network structure of the typhoon prediction model is ResNet integrating the InceptionNet network structure; Among them, the typhoon prediction model includes a first processing module, a second processing module, a third processing module, a fourth processing module, a max pooling layer and a linear layer connected in sequence; The first processing module is used to perform dimensionality reduction processing and feature extraction on the input typhoon area images, and output a first feature map; The second processing module is used to process the first feature map and output a second feature map; The third processing module is used to process the second feature map and output a third feature map; The fourth processing module is used to process the third feature map and output a fourth feature map; The max pooling layer is used to perform max pooling operation on the fourth feature map and output a fifth feature map; The linear layer is used to perform linear transformation on the fifth feature map and output a three-dimensional typhoon information vector, and the three-dimensional typhoon information vector includes: the longitude of the typhoon center position, the latitude of the typhoon center position and intensity information; The first processing module sequentially includes: a first unit, a second unit, a third unit and a fourth unit; among them, the second unit and the fourth unit have the same structure; The first unit includes a 3x3 convolutional layer and a 5x5 convolutional layer connected in series; The second unit includes: a 3x3 average pooling layer and a 3x3 convolutional layer connected in parallel, and a filtering connection layer for synthesizing the output results of the two paths; The third unit includes three branches and a filtering connection layer for synthesizing the output results of the three branches. Among them, the first branch is a direct output, the second branch includes a 3x3 convolutional layer and a 1x1 convolutional layer connected in series, and the third branch includes a 1x1 convolutional layer, a 5x1 convolutional layer, a 1x5 convolutional layer and a 5x5 convolutional layer connected in series; The second processing module includes: 3 ResNet-A modules connected in series and 1 Reduction-1 module; The ResNet-A module includes: a first activation function layer, a fifth unit and a second activation function layer; The fifth unit includes a first branch, a second branch, a third branch, a fourth branch, a fifth branch, a 1x1 convolutional layer and an adder; the 1x1 convolutional layer processes the outputs of the third branch, the fourth branch and the fifth branch, and the adder accumulates the outputs of the first branch, the second branch and the 1x1 convolutional layer and outputs to the second activation function layer; the first branch includes a 1x7 convolutional layer and a 7x1 convolutional layer connected in series; the second branch is a direct output; the third branch includes a 1x1 convolutional layer; the fourth branch includes a 1x1 convolutional layer and a 3x3 convolutional layer connected in series; the fifth branch includes a 1x1 convolutional layer, a 3x3 convolutional layer and a 3x3 convolutional layer connected in series; The Reduction-1 module includes: a filtering connection layer, a sixth unit and a filtering connection layer; The sixth unit includes three branches. The first branch includes a 5x5 max pooling layer; the second branch includes a 3x3 convolutional layer; the third branch includes a 1x1 convolutional layer in series and a 3x3 convolutional layer; The third processing module includes 5 ResNet-B modules in series and 1 Reduction-2 module; The ResNet-B module includes a third activation function layer, a seventh unit, and a fourth activation function layer; The seventh unit includes a first branch, a second branch, a third branch, a fourth branch, a 1x1 convolutional layer, and an adder. The 1x1 convolutional layer processes the outputs of the third branch and the fourth branch; the adder accumulates the outputs of the first branch, the second branch, and the 1x1 convolutional layer and inputs them into the fourth activation function layer. Among them, the first branch includes a 1x5 convolutional layer and a 5x1 convolutional layer in series; it is directly output without processing; the third branch includes a 1x1 convolutional layer; the third branch includes a 1x1 convolutional layer and a 3x3 convolutional layer in series; Reduction-2 includes: a filtering connection layer, 4 branches, and a filtering connection layer; among them, the first branch includes a 5x5 max pooling layer, the second branch includes a 1x1 convolutional layer and a 3x3 convolutional layer in series; the third branch includes a 1x1 convolutional layer and a 3x3 convolutional layer; the fourth branch includes a 1x1 convolutional layer, a 3x3 convolutional layer, and a 3x3 convolutional layer in series; The ResNet-C module includes a fifth activation function layer, a seventh unit, and a sixth activation function layer; The seventh unit includes a first branch, a first branch, a third branch, a 1x1 convolutional layer, and an adder; the 1x1 convolutional layer processes the outputs of the second branch and the third branch; the adder accumulates the output of the first branch and the 1x1 convolutional layer and inputs them into the sixth activation function layer; the first branch is directly output without processing; the second branch includes a 1x1 convolutional layer; the third branch includes a 1x1 convolutional layer and a 3x3 convolutional layer.

2. The typhoon positioning and intensity determination method based on remote sensing data according to claim 1, characterized in that, preprocessing the remote sensing image to obtain a typhoon area image, including: converting the remote sensing image into a grayscale image; cropping an initial typhoon area image of a preset size from the grayscale image; expanding the initial typhoon area image into a square image according to the longest side, and filling the expanded area with black; rotating, adding noise, displacing, and changing the brightness of the expanded square image to obtain a typhoon area image.

3. The typhoon positioning and intensity determination method based on remote sensing data according to claim 1, characterized in that, the training process of the typhoon prediction model includes: obtaining the remote sensing data at all typhoon moments in the typhoon best track dataset, converting the remote sensing data into a grayscale image, maintaining the inherent resolution, and cropping a typhoon area sample image of a preset size from the grayscale image; preprocessing all typhoon area sample images and annotating the typhoon center position and typhoon intensity information; using the typhoon prediction model to process each typhoon area sample image to obtain the predicted typhoon center position and typhoon intensity information; Calculate the loss function based on the predicted typhoon center position and typhoon intensity information and the labeled typhoon center position and typhoon intensity information Among them, the vector y = (y 1 , y 2 , y 3 ), where (y 1 , y 2 ) is the longitude and latitude of the predicted typhoon center position, and y 3 is the predicted typhoon intensity information; the vector Among them is the longitude and latitude of the labeled typhoon center position, is the labeled typhoon intensity information; updating the model parameters of the typhoon prediction model based on the loss function value and the batch gradient descent method.

4. A typhoon positioning and intensity determination device based on remote sensing data, characterized in that, it includes: An acquisition unit for acquiring remote sensing images and preprocessing the remote sensing images to obtain typhoon area images; A processing unit for processing the typhoon area images by using a pre-trained typhoon prediction model to obtain the typhoon center position and typhoon intensity; the network structure of the typhoon prediction model is a ResNet that integrates the InceptionNet network structure; Among them, the typhoon prediction model includes a first processing module, a second processing module, a third processing module, a fourth processing module, a max pooling layer, and a linear layer connected in sequence; The first processing module is used for dimensionality reduction processing and feature extraction of the input typhoon area images, and outputs a first feature map; The second processing module is used for processing the first feature map and outputs a second feature map; The third processing module is used for processing the second feature map and outputs a third feature map; The fourth processing module is used for processing the third feature map and outputs a fourth feature map; The max pooling layer is used for performing max pooling operations on the fourth feature map and outputs a fifth feature map; The linear layer is used for linearly transforming the fifth feature map and outputs a three-dimensional typhoon information vector, and the three-dimensional typhoon information vector includes: the longitude of the typhoon center position, the latitude of the typhoon center position, and intensity information; The first processing module sequentially includes: a first unit, a second unit, a third unit, and a fourth unit; among them, the second unit and the fourth unit have the same structure; The first unit includes a 3x3 convolutional layer and a 5x5 convolutional layer connected in series; The second unit includes: a 3x3 average pooling layer and a 3x3 convolutional layer connected in parallel, and a filtering connection layer for synthesizing the output results of the two paths; The third unit includes three branches and a filtering connection layer for synthesizing the output results of the three branches. Among them, the first branch is a direct output, the second branch includes a 3x3 convolutional layer and a 1x1 convolutional layer connected in series, and the third branch includes a 1x1 convolutional layer, a 5x1 convolutional layer, a 1x5 convolutional layer, and a 5x5 convolutional layer connected in series; The second processing module includes: 3 ResNet-A modules connected in series and 1 Reduction-1 module; The ResNet-A module includes: a first activation function layer, a fifth unit, and a second activation function layer; The fifth unit includes a first branch, a second branch, a third branch, a fourth branch, a fifth branch, a 1x1 convolutional layer, and an adder; the 1x1 convolutional layer processes the outputs of the third branch, the fourth branch, and the fifth branch, and the adder accumulates the outputs of the first branch, the second branch, and the 1x1 convolutional layer and outputs them to the second activation function layer; the first branch includes a 1x7 convolutional layer and a 7x1 convolutional layer connected in series; the second branch is a direct output; the third branch includes a 1x1 convolutional layer; the fourth branch includes a 1x1 convolutional layer and a 3x3 convolutional layer connected in series; the fifth branch includes a 1x1 convolutional layer, a 3x3 convolutional layer, and a 3x3 convolutional layer connected in series; The Reduction-1 module includes: a filtering connection layer, a sixth unit, and a filtering connection layer; The sixth unit includes three branches. The first branch includes a 5x5 max pooling layer; the second branch includes a 3x3 convolutional layer; the third branch includes a 1x1 convolutional layer and a 3x3 convolutional layer in series; The third processing module includes 5 ResNet-B modules in series and 1 Reduction-2 module; The ResNet-B module includes a third activation function layer, a seventh unit, and a fourth activation function layer; The seventh unit includes a first branch, a second branch, a third branch, a fourth branch, a 1x1 convolutional layer, and an adder. The 1x1 convolutional layer processes the outputs of the third branch and the fourth branch; the adder accumulates the outputs of the first branch, the second branch, and the 1x1 convolutional layer and inputs them into the fourth activation function layer. Among them, the first branch includes a 1x5 convolutional layer and a 5x1 convolutional layer in series and outputs directly without processing; the third branch includes a 1x1 convolutional layer; the third branch includes a 1x1 convolutional layer and a 3x3 convolutional layer in series; Reduction-2 includes: a filtering connection layer, 4 branches, and a filtering connection layer; among them, the first branch includes a 5x5 max pooling layer, the second branch includes a 1x1 convolutional layer and a 3x3 convolutional layer in series; the third branch includes a 1x1 convolutional layer and a 3x3 convolutional layer; the fourth branch includes a 1x1 convolutional layer, a 3x3 convolutional layer, and a 3x3 convolutional layer in series; The ResNet-C module includes a fifth activation function layer, a seventh unit, and a sixth activation function layer; The seventh unit includes a first branch, a first branch, a third branch, a 1x1 convolutional layer, and an adder; the 1x1 convolutional layer processes the outputs of the second branch and the third branch; the adder accumulates the outputs of the first branch and the 1x1 convolutional layer and inputs them into the sixth activation function layer; the first branch outputs directly without processing; the second branch includes a 1x1 convolutional layer; the third branch includes a 1x1 convolutional layer and a 3x3 convolutional layer.

5. An electronic device, characterized in that, it includes: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the typhoon positioning and intensity determination method based on remote sensing data according to any one of claims 1-3.

Images