Typhoon intensity estimation method and system fusing time sequence and space dynamic characteristics

Abstract

The invention relates to the field of meteorological monitoring, in particular to a typhoon intensity estimation method and system fusing time sequence and space dynamic characteristics, and the method comprises the following steps: obtaining the basic information data of a typhoon to be estimated; obtaining a satellite cloud picture of the typhoon to be estimated based on the basic information data; obtaining a time sequence three-dimensional cloud picture of the typhoon to be estimated according to the satellite cloud picture; constructing and optimizing a nonlinear regression model for estimating typhoon intensity; and importing the time sequence three-dimensional cloud picture into the nonlinear regression model to obtain estimated typhoon intensity. According to the method, the non-linear regression model is constructed to estimate the typhoon intensity, information loss is effectively reduced, typhoon space feature information at multiple moments is imported into the non-linear regression model, time sequence change features of the typhoon are effectively captured, an attention mechanism is introduced to dynamically adjust time sequence three-dimensional cloud picture feature information, irrelevant information interference is inhibited, and the typhoon intensity is estimated. The key feature information of the image is enhanced, and the accuracy of typhoon intensity estimation is improved.

Description

Technical Field

[0001] The present invention relates to the field of meteorological monitoring, and in particular to a typhoon intensity estimation method and system integrating time series and spatial dynamic characteristics. Background Art

[0002] Typhoons are a type of disastrous weather that originates from tropical oceans. When a typhoon approaches the sea or makes landfall, it often causes huge losses to people in the surrounding areas. Strengthening typhoon monitoring and forecasting is an important means to mitigate typhoon disasters, and the accuracy of typhoon location and intensity estimation is the key to improving the accuracy of typhoon forecasts. Since typhoons are mainly active on tropical oceans, there is a lack of routine monitoring, so typhoon intensity estimation based on satellite observations has become the best choice for typhoon monitoring.

[0003] The basic physical basis for using satellite infrared band observation data to estimate typhoon intensity is that the typhoon cloud top information detected by satellite remote sensing has a strong correlation with its intensity. At present, the typhoon intensity estimation methods based on satellite infrared band observation data mainly include four categories: Dvorak technology, directional gradient angle technology, convective kernel analysis method and deep learning method. Dvorak technology is a method for estimating the intensity of tropical cyclones by analyzing the cloud structure in satellite images. It was originally based on user experience, subjective and time-consuming. Subsequent derivative algorithms have gradually achieved automation, reduced subjectivity and improved the accuracy of estimation, but are still limited by image resolution and dependence on cloud structure. Directional gradient angle technology calculates typhoon intensity by analyzing the directional gradient of satellite infrared image brightness and fitting nonlinear equations. It uses the axial symmetry of cyclone clouds to estimate wind speed. Although it effectively combines statistical parameters, it still relies on expert experience to optimize parameters and has subjective bias. The convective core analysis method can help estimate the intensity of tropical cyclones. A regression equation is established based on the number, location, brightness temperature and other characteristics of the convective cores. However, when the cyclone is weak, the dispersion of the convective cores cannot accurately reflect the intensity, while when the cyclone is strong, insufficient coverage of the retrieval area may lead to underestimation of the typhoon intensity.

[0004] In recent years, deep learning methods have been widely used in the field of meteorology and have achieved remarkable results. Taking convolutional neural networks as an example, their highly nonlinear characteristics give the model powerful expressive power, and can automatically extract features from data, eliminating the tedious feature extraction process. This makes typhoon intensity estimation not only more automated and objective, but also improves the accuracy and efficiency of estimation. However, there are still some problems: ignoring time series information; existing research on estimating typhoon intensity using convolutional neural networks is often based on satellite data of multiple bands at a single moment. However, the development of typhoons is a continuous process, and images at a single moment cannot fully reflect the actual state of the typhoon. In addition, the existing convolutional neural network models fail to fully utilize the dynamic information of the input data in typhoon intensity estimation, resulting in limited estimation accuracy. These models process the regions and channels in the feature map in a static manner, and cannot flexibly adjust the weights, which limits the adaptability to complex typhoon characteristics and fails to effectively mine the deep relationships in the feature map. In addition, most previous studies used convolutional neural network classification models to estimate typhoon intensity. Because the intensity was divided into discrete categories, there was information loss, which made it impossible to accurately reflect continuous intensity changes. In addition, the classification model also had the problem of error margin that was not taken into account, that is, the error of misclassifying a super typhoon category as a strong typhoon category and a tropical depression category in the classification task was the same. Summary of the invention

[0005] In view of the defects in the prior art, the present invention provides a typhoon intensity estimation method and system integrating time series and spatial dynamic characteristics.

[0006] In order to achieve the above-mentioned purpose, in the first aspect, the present invention provides a typhoon intensity estimation method integrating time series and spatial dynamic characteristics, the method comprising the following steps: obtaining basic information data of the typhoon to be estimated; obtaining a satellite cloud map of the typhoon to be estimated based on the basic information data; obtaining a time series three-dimensional cloud map of the typhoon to be estimated based on the satellite cloud map; constructing and optimizing a nonlinear regression model for estimating typhoon intensity; importing the time series three-dimensional cloud map into the nonlinear regression model to obtain the estimated typhoon intensity. The present invention introduces typhoon characteristic information at multiple moments into the nonlinear regression model, which can effectively capture the time series characteristics of typhoon development, and introduces an attention mechanism to dynamically adjust the typhoon cloud map characteristic information, effectively suppress the interference of irrelevant information, and directly estimate the typhoon intensity value through the nonlinear regression model, effectively reducing information loss and improving the accuracy of estimation.

[0007] Optionally, the basic information data of the typhoon to be estimated is obtained, including: obtaining the best path data set of the typhoon to be estimated according to the number and name of the typhoon to be estimated; obtaining the real typhoon intensity of the typhoon to be estimated at the time to be estimated based on the best path data set; obtaining the central longitude and latitude data of the typhoon to be estimated according to the best path data set. The present invention matches satellite data based on the best path of the typhoon to construct a continuous data set in which the actual value of the intensity of the typhoon to be estimated matches the satellite observation, thereby ensuring the accuracy and comprehensiveness of the data set, being able to accurately capture the development and change trend of the typhoon, and obtaining the real typhoon intensity and the center position of the typhoon at the time to be estimated in combination with the best path data set, thereby ensuring the rigor of the source of typhoon data information, providing strong data support for the subsequent steps of the present invention, and improving the rigor and reliability of the present invention.

[0008] Optionally, the satellite cloud image of the typhoon to be estimated is obtained based on the basic information data, including: obtaining the satellite cloud image of the typhoon to be estimated by using satellite observation according to the central longitude and latitude data of the typhoon to be estimated. The present invention obtains a time-series three-dimensional satellite cloud image of the typhoon at the time to be estimated based on the central longitude and latitude information of the typhoon to be estimated, which ensures the accuracy and reliability of the image information, and provides image data support for further processing of typhoon data, effectively displays the brightness temperature information about the typhoon cloud top, can obtain the convection strength indication in the vertical direction of the typhoon, can more intuitively observe the convection state of the typhoon, and improves the scientificity, intuitiveness and feasibility of the method of the present invention.

[0009] Optionally, the method of obtaining the time-series three-dimensional cloud map of the typhoon to be estimated based on the satellite cloud map includes: obtaining the position coordinates of the center of the typhoon to be estimated in the satellite cloud map; clipping the satellite cloud map with the position coordinates as the center to obtain the typhoon cloud map of the typhoon to be estimated; and obtaining the time-series three-dimensional cloud map of the typhoon to be estimated based on the typhoon cloud map. The present invention clips the satellite cloud map with the position coordinates of the typhoon center as the center to obtain a typhoon cloud map that can effectively reflect the characteristics of the typhoon, performs image processing on complex typhoon conditions, simplifies the steps of data processing, and greatly improves the effectiveness of data processing. A time-series three-dimensional cloud map containing typhoon information is obtained based on the typhoon cloud map, which can effectively capture the time-series characteristics of typhoon development, reflect the changing trend of the typhoon, ensure the accuracy and reliability of the data, and effectively improve the accuracy of typhoon estimation.

[0010] Optionally, the construction and optimization of the nonlinear regression model for estimating typhoon intensity includes: constructing a neural network model as a nonlinear regression model for estimating typhoon intensity; establishing a feature label data set for the typhoon to be estimated; and training and optimizing the nonlinear regression model based on the feature label data set. The present invention constructs a nonlinear regression model based on a convolutional neural network, and its powerful nonlinear fitting ability can efficiently and accurately model complex and changeable typhoon characteristics, and can more accurately estimate typhoon intensity. At the same time, a feature label data set containing typhoon feature information is constructed, and the nonlinear regression model is trained and optimized based on the feature label data set, thereby improving the accuracy and robustness of the constructed nonlinear regression model in estimating typhoon intensity.

[0011] Optionally, the method of importing the time series three-dimensional cloud image into the nonlinear regression model to obtain the estimated typhoon intensity includes: importing the time series three-dimensional cloud image into the nonlinear regression model; introducing an attention mechanism to strengthen the characteristic information of the time series three-dimensional cloud image; processing the imported time series three-dimensional cloud image based on the nonlinear regression model and obtaining an output result; and using the output result obtained by the nonlinear regression model as the estimated typhoon intensity. The present invention imports the time series three-dimensional cloud image into the nonlinear regression model, and at the same time strengthens the key image features in the time series three-dimensional cloud image by introducing an attention mechanism, thereby improving the expression ability of the characteristic information of the typhoon cloud image, dynamically adjusting the spatial area and channel of the characteristic information of the typhoon cloud image, and effectively suppressing the interference of irrelevant information. The loss function in the nonlinear regression model can flexibly evaluate the error amplitude, significantly improving the accuracy and robustness of the nonlinear regression model in estimating the typhoon intensity, providing sufficient theoretical and technical support for practical applications, and enhancing the practical value of the method of the present invention.

[0012] Optionally, the introduction of the attention mechanism to strengthen the characteristic information of the time series three-dimensional cloud map includes: the attention mechanism includes a channel attention mechanism and a spatial attention mechanism; the channel attention mechanism is used to process the typhoon cloud map characteristic information of the typhoon to be estimated to obtain weighted typhoon cloud map characteristic information; the spatial attention mechanism is used to weight each position in the weighted typhoon cloud map characteristic information to obtain the final typhoon cloud map characteristic information; based on the final typhoon cloud map characteristic information, the time series three-dimensional cloud map characteristic information after strengthening the key characteristic information is formed. The present invention dynamically adjusts the channels and spatial regions in the typhoon cloud map characteristic information by introducing the channel attention mechanism and the spatial attention mechanism, strengthens the channels and spatial regions in the typhoon cloud map characteristic information that are most important for estimating the typhoon intensity, suppresses the interference of irrelevant information, improves the expression ability of the typhoon cloud map characteristic information, enhances the simplicity and reliability of the data, provides strong data support for the subsequent estimation of the typhoon intensity, and significantly improves the accuracy of the nonlinear regression model in estimating the typhoon intensity.

[0013] Optionally, the using the channel attention mechanism to process the typhoon cloud image feature information of the typhoon to be estimated to obtain weighted typhoon cloud image feature information includes:

[0014] M C =σ(FC2(RELU(FC1(Global Average Pooling(F)+Global Max Pooling(F)))))

[0015] F C =F·M C

[0016] Among them, M C is the channel attention weight, σ is the activation function, FC1 and FC2 are fully connected layers, RELU is the nonlinear activation function, Global Average Pooling is the global average pooling result, F is the typhoon cloud image feature information, GlobalMax Pooling is the global maximum pooling result, F C The present invention introduces a channel attention mechanism to weight the typhoon cloud map feature information of each channel, thereby enhancing the expression ability of important feature channels. The channel attention weight is used to adjust the channel importance of the typhoon cloud map feature information, thereby enhancing the ability of the nonlinear regression model to adapt to complex typhoon conditions and improving the accuracy of estimating typhoon intensity.

[0017] Optionally, the using the spatial attention mechanism to weight each position in the weighted typhoon cloud image feature information to obtain the final typhoon cloud image feature information includes:

[0018] M S =σ(Conv7(RELU(Conv1([Average Pooling(F C ),Max Pooling(F C )]))))

[0019] F final =F C ·M S

[0020] Among them, M S is the spatial attention weight, σ is the activation function, Conv7 is a 7×7 convolution operation, RELU is a nonlinear activation function, Conv1 is a 1×1 convolution operation, Average Pooling is the average pooling result, and F C is the weighted typhoon cloud image feature information, Max Pooling is the maximum pooling result, F finalThe present invention introduces a spatial attention mechanism to weight each position of the typhoon cloud map feature information, thereby enhancing the feature expression of important spatial positions. The spatial attention weight is used to adjust the spatial position importance of the typhoon cloud map feature information, thereby enhancing the reliability and rigor of the nonlinear regression model and improving the accuracy of estimating the typhoon intensity.

[0021] In the second aspect, the present invention provides a typhoon intensity estimation system integrating time series and spatial dynamic features, the system executes the typhoon intensity estimation method integrating time series and spatial dynamic features provided by the present invention, the system includes an input device, an output device, a processor and a memory, and its gain is: the hardware facilities integrated by the present invention have excellent performance, the input device, the output device, the processor and the memory are interconnected, the information transmission between the various components is smooth, and an efficient information processing system is constructed through the interaction of multiple hardware facilities. The system of the present invention introduces the spatial feature information of the typhoon at multiple moments into the nonlinear regression model, comprehensively considers the historical state and change trend of the typhoon, effectively captures the time series change characteristics of the typhoon development, and introduces the attention mechanism to dynamically adjust the time series three-dimensional cloud map feature information of the typhoon, effectively suppresses the interference of irrelevant information, and directly estimates the typhoon intensity value through the nonlinear regression model, which can effectively reduce the loss of information and improve the accuracy of the estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flow chart of a typhoon intensity estimation method integrating time series and spatial dynamic characteristics according to an embodiment of the present invention;

[0023] Figure 2 A schematic diagram of key process steps and nodes of an embodiment of the present invention;

[0024] Figure 3 It is a schematic diagram of optimizing the time-series three-dimensional cloud image features according to an embodiment of the present invention;

[0025] Figure 4 Schematic diagram of a channel attention mechanism according to an embodiment of the present invention;

[0026] Figure 5 Schematic diagram of the spatial attention mechanism of an embodiment of the present invention;

[0027] Figure 6 The present invention is a framework diagram of a typhoon intensity estimation system that integrates temporal and spatial dynamic characteristics. DETAILED DESCRIPTION

[0028] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are only for illustration and are not intended to limit the present invention. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details do not need to be adopted to implement the present invention. In other examples, in order to avoid confusing the present invention, known circuits, software or methods are not specifically described.

[0029] Throughout the specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment of the present invention. Therefore, the phrases "in one embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily all refer to the same embodiment or example. In addition, particular features, structures, or characteristics may be combined in one or more embodiments or examples in any suitable combination and / or subcombination. In addition, it should be understood by those of ordinary skill in the art that the figures provided herein are for illustrative purposes and that the figures are not necessarily drawn to scale.

[0030] See also Figure 1 An embodiment of the present invention provides a typhoon intensity estimation method integrating temporal and spatial dynamic characteristics, the method comprising the following steps:

[0031] S1. Obtain basic information data of the typhoon to be estimated.

[0032] Among them, S1 specifically includes the following steps:

[0033] S11. Acquire an optimal path data set of the typhoon to be estimated according to the number and name of the typhoon to be estimated.

[0034] Specifically, according to the number and name of the typhoon to be estimated, a processor is used to match the satellite infrared band observation data of the corresponding time to obtain a satellite infrared band observation data set, and based on the infrared band observation data set, the best path data set of the typhoon to be estimated is obtained and stored in a memory.

[0035] S12. Obtaining the actual typhoon intensity of the typhoon to be estimated at the time to be estimated based on the optimal path data set.

[0036] Specifically, the processor is used to extract the real typhoon intensity of the typhoon at the time to be estimated from the optimal path data set, and the real typhoon intensity is stored in the memory.

[0037] S13. Acquire the central longitude and latitude data of the typhoon to be estimated according to the optimal path data set.

[0038] Specifically, a processor is used to extract the central longitude and latitude data of the typhoon at the time to be estimated, 6 hours before the time and 12 hours before the time from the optimal path data set, and save it to the memory.

[0039] S2. Obtaining a satellite cloud image of the typhoon to be estimated based on the basic information data.

[0040] Specifically, based on the central longitude and latitude data, a processor is used to download satellite cloud images of the satellite at the time to be estimated, 6 hours before the time and 12 hours before the time, and store them in a memory.

[0041] It should be noted that the satellite's operating band is the sensor's infrared band, which can effectively provide brightness temperature information on typhoon cloud tops and provide indications on the strength of convection in the vertical direction of the typhoon.

[0042] S3. Obtaining a time-series three-dimensional cloud map of the typhoon to be estimated based on the satellite cloud map.

[0043] Among them, S3 specifically includes the following steps:

[0044] S31, obtaining the position coordinates of the typhoon center to be estimated in the satellite cloud map.

[0045] Specifically, a processor is used to convert the central longitude and latitude data of the typhoon into position coordinates in a satellite cloud map to obtain the central position of the typhoon in the satellite cloud map.

[0046] S32. Crop the satellite cloud image with the position coordinates as the center to obtain a typhoon cloud image of the typhoon to be estimated.

[0047] Specifically, a processor is used to expand n grid points around the position coordinate as the center to crop the satellite cloud image, and the obtained 2n+1×2n+1 square typhoon cloud image is stored in the memory.

[0048] More specifically, when n=85, the clipping operation includes:

[0049] Crop(Image t ,c t ,r t )=Image t (r t -85:r t +85, c t -85:c t +85)

[0050] Among them, Crop is the cropping operation, Image t For satellite cloud images, c t is the column index in the image, r t is the row index in the image.

[0051] Furthermore, the satellite cloud image is cropped by using a processor to obtain the typhoon cloud image at the time to be estimated, 6 hours before the time and 12 hours before the time, and the typhoon cloud image is stored in a memory.

[0052] S33. Obtaining a time-series three-dimensional cloud map of the typhoon to be estimated according to the typhoon cloud map.

[0053] Specifically, the processor is used to stack the typhoon cloud images at the time to be estimated, 6 hours before, and 12 hours before in chronological order to obtain a time-series three-dimensional cloud image, and save it in a memory.

[0054] Furthermore, the time series three-dimensional cloud map includes:

[0055] x i =[C t-12 ,C t-6 ,C t ]

[0056] Among them, x i is a time series three-dimensional cloud map, C t-12 The typhoon cloud map for the previous 12 hours, C t-6 The typhoon cloud map for the previous 6 hours, C t This is the typhoon cloud map at the time of estimation.

[0057] S4. Construct and optimize a nonlinear regression model for estimating typhoon intensity.

[0058] Wherein, S4 specifically includes the following steps:

[0059] S41. Construct a neural network model as a nonlinear regression model for estimating typhoon intensity.

[0060] Specifically, a convolutional neural network model is constructed using a processor as a nonlinear regression model, and a convolutional neural network architecture is designed, including a convolutional layer, a pooling layer, and a fully connected layer, which are stored in a memory, as shown in Table 1:

[0061] Table 1 Model structure

[0062]

[0063] It should be noted that the convolution layer is used to extract the features of the typhoon cloud image, the pooling layer is used to reduce the spatial size of the typhoon cloud image, the processed typhoon cloud image is passed to the fully connected layer, and the output dimension of the last fully connected layer is set to 1.

[0064] S42: Establish a feature label data set of the typhoon to be estimated.

[0065] Specifically, a processor is used to construct and save a feature label data set using the typhoon intensity at the time to be estimated in the typhoon best path data set as label data, including:

[0066] D={ <x i ,label i >|i=1,2,3…,N}

[0067] Among them, D is the feature label dataset, x i It is a time series three-dimensional cloud map, label i is the corresponding typhoon intensity, and N is the sample size.

[0068] Furthermore, the feature label dataset is divided into training set, validation set and test set, with the proportions of 70%, 15% and 15% respectively. The training set is used to train the model, the validation set is used to verify the effect of the optimized parameters, and the test set is used to finally evaluate the performance of the model.

[0069] S43: training and optimizing the nonlinear regression model according to the feature label data set.

[0070] Specifically, the mean square error is selected as the loss function to calculate the error between the estimated value of the regression model and the actual label value, including:

[0071]

[0072] Among them, MSE is the mean square error, N is the total number of samples, is the predicted value of the i-th sample, y i is the actual value.

[0073] Furthermore, the parameters of the nonlinear regression model are optimized by back propagation algorithm and stochastic gradient descent optimizer, including:

[0074]

[0075] Among them, θ t+1 is the parameter of the t+1th iteration, θ t is the parameter of the tth iteration, η is the learning rate, is the gradient of the loss function with respect to the parameters.

[0076] S5. Importing the time series three-dimensional cloud image into the nonlinear regression model to obtain an estimated typhoon intensity.

[0077] See also Figure 2, the figure is a schematic diagram of the key process steps of the method of the present invention. First, based on the satellite cloud images of the time to be estimated, the 6 hours before and the 12 hours before the time, the time series 3D cloud image is obtained, and then the time series 3D cloud image is imported into the nonlinear regression model. After the convolution layer and pooling layer of the nonlinear regression model, a feature map with multi-channel information is obtained, and then the attention mechanism is introduced to further enhance the features of the multi-channel information feature map, and then the feature information processed by the attention mechanism is passed to the fully connected layer of the nonlinear regression model, and finally the estimated typhoon intensity is obtained.

[0078] For details, see Figure 3 , the figure shows a schematic diagram of introducing the attention mechanism to optimize the input features of the time series three-dimensional cloud map. First, the features with multi-channel information obtained after the convolution layer and pooling layer of the nonlinear regression model are input into the channel attention mechanism to obtain the output, and then the output is used as the input of the spatial attention mechanism. Finally, the optimized time series three-dimensional cloud map feature information is obtained through the spatial attention mechanism.

[0079] It should be noted that the typhoon cloud images at different times in the time series three-dimensional cloud image contain different feature information, and different development stages of the typhoon will show different cloud image features; although the convolution layer and pooling layer have extracted the features of this information, the channel attention mechanism further automatically identifies the feature channels or moments that are most important to the final estimation results in the feature map, thereby highlighting the core key features; the spatial attention mechanism enables the model to automatically identify the areas in the typhoon cloud image that have a greater impact on the estimation results. Key areas such as the typhoon's eye wall and cloud distribution will be given higher weights, so that the nonlinear regression model can learn the features of key areas more accurately.

[0080] For further information, see Figure 4 The attention mechanism includes a channel attention mechanism. The figure is a schematic diagram of the channel attention mechanism. The channel attention mechanism weights the typhoon cloud image feature information of each channel to enhance the expression ability of important feature channels, including:

[0081] M C =σ(FC2(RELU(FC1(Global Average Pooling(F)+Global Max Pooling(F)))))

[0082] F C =F·M C

[0083] Among them, M Cis the channel attention weight, σ is the activation function, FC1 and FC2 are fully connected layers, RELU is the nonlinear activation function, Global Average Pooling is the global average pooling result, F is the typhoon cloud image feature information, GlobalMax Pooling is the global maximum pooling result, F C It is the weighted typhoon cloud image characteristic information.

[0084] For further information, see Figure 5 The attention mechanism includes a spatial attention mechanism. The figure is a schematic diagram of the spatial attention mechanism. The spatial attention mechanism weights each position of the typhoon cloud image feature information to enhance the features of important spatial positions, including:

[0085] M S =σ(Conv7(RELU(Conv1([Average Pooling(F C ),Max Pooling(F C )]))))

[0086] F final =F C ·M S

[0087] Among them, M S is the spatial attention weight, σ is the activation function, Conv7 is a 7×7 convolution operation, RELU is a nonlinear activation function, Conv1 is a 1×1 convolution operation, Average Pooling is the average pooling result, and F C is the weighted typhoon cloud image feature information, Max Pooling is the maximum pooling result, F final This is the final typhoon cloud map feature information.

[0088] See also Figure 6 In an optional embodiment, in order to efficiently execute the typhoon intensity estimation method integrating time series and spatial dynamic features provided by the present invention, the present invention provides a typhoon intensity estimation system integrating time series and spatial dynamic features, the system comprising an input device, an output device, a processor and a memory, the hardware facilities being interconnected, wherein the memory is used to store a computer program, the computer program comprising program instructions, the processor being configured to call the program instructions and execute the specific steps of the embodiment related to the typhoon intensity estimation method integrating time series and spatial dynamic features provided by the present invention. The typhoon intensity estimation system integrating time series and spatial dynamic features provided by the present invention has a complete structure, is objective and stable, and can efficiently execute the typhoon intensity estimation method integrating time series and spatial dynamic features provided by the present invention, thereby improving the overall applicability and practical application capability of the present invention.

[0089] In summary, the method of the present invention provides a typhoon intensity estimation method and system that integrates time series and spatial dynamic characteristics, introduces image data of typhoons at multiple times into a nonlinear regression model, comprehensively considers the historical state and change trend of the typhoon, effectively captures the time series characteristics of the typhoon development, and introduces an attention mechanism to dynamically adjust the time series three-dimensional cloud map of the typhoon, effectively suppresses the interference of irrelevant information, and directly estimates the typhoon intensity value through a nonlinear regression model, which can effectively reduce the loss of information and improve the accuracy of the estimation. The method of the present invention is easy to understand, simple to calculate, has a small workload, is easy to apply in engineering, and provides a theoretical basis and technical support for the further development of the typhoon monitoring field.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A typhoon intensity estimation method integrating temporal and spatial dynamic characteristics, characterized in that: The steps include: Obtain basic information data of the typhoon to be estimated; Obtaining a satellite cloud image of the typhoon to be estimated based on the basic information data; Obtaining a time-series three-dimensional cloud map of the typhoon to be estimated based on the satellite cloud map; Construct and optimize a nonlinear regression model for estimating typhoon intensity; The time series three-dimensional cloud map is imported into the nonlinear regression model to obtain the estimated typhoon intensity.

2. The typhoon intensity estimation method integrating temporal and spatial dynamic characteristics according to claim 1 is characterized in that: The obtaining of basic information data of the typhoon to be estimated includes: Acquire the best path data set of the typhoon to be estimated according to the number and name of the typhoon to be estimated; Obtaining the true typhoon intensity of the typhoon to be estimated at the time to be estimated based on the optimal path data set; The central longitude and latitude data of the typhoon to be estimated are obtained according to the optimal path data set.

3. The typhoon intensity estimation method integrating temporal and spatial dynamic characteristics according to claim 2 is characterized in that: The step of obtaining the satellite cloud image of the typhoon to be estimated based on the basic information data includes: A satellite cloud image of the typhoon to be estimated is obtained by using satellite observation based on the central longitude and latitude data of the typhoon to be estimated.

4. The typhoon intensity estimation method integrating temporal and spatial dynamic characteristics according to claim 3 is characterized in that: The step of obtaining the time-series three-dimensional cloud image of the typhoon to be estimated based on the satellite cloud image comprises: Obtain the position coordinates of the typhoon center to be estimated in the satellite cloud map; Cutting the satellite cloud image with the position coordinates as the center to obtain a typhoon cloud image of the typhoon to be estimated; A time-series three-dimensional cloud map of the typhoon to be estimated is obtained according to the typhoon cloud map.

5. The typhoon intensity estimation method integrating temporal and spatial dynamic characteristics according to claim 4 is characterized in that: The nonlinear regression model for estimating typhoon intensity is constructed and optimized, including: A neural network model was constructed as a nonlinear regression model for estimating typhoon intensity; Establishing a feature label data set of the typhoon to be estimated; The nonlinear regression model is trained and optimized according to the feature label data set.

6. The typhoon intensity estimation method integrating temporal and spatial dynamic characteristics according to claim 5 is characterized in that: The step of importing the time series three-dimensional cloud image into the nonlinear regression model to obtain the estimated typhoon intensity includes: Importing the time series three-dimensional cloud map into the nonlinear regression model; An attention mechanism is introduced to strengthen the feature information of the time-series three-dimensional cloud image; Processing the imported time series three-dimensional cloud image based on the nonlinear regression model and obtaining an output result; The output result obtained by the nonlinear regression model is used as the estimated typhoon intensity.

7. The typhoon intensity estimation method integrating temporal and spatial dynamic characteristics according to claim 6 is characterized in that: The introduction of the attention mechanism to strengthen the feature information of the time-series three-dimensional cloud image includes: The attention mechanism includes a channel attention mechanism and a spatial attention mechanism; Using the channel attention mechanism to process the typhoon cloud image feature information of the typhoon to be estimated to obtain weighted typhoon cloud image feature information; Using the spatial attention mechanism to weight each position in the weighted typhoon cloud image feature information to obtain the final typhoon cloud image feature information; Based on the final typhoon cloud image characteristic information, the time-series three-dimensional cloud image characteristic information after strengthening the key characteristic information is formed.

8. The typhoon intensity estimation method integrating temporal and spatial dynamic characteristics according to claim 7 is characterized in that: The step of using the channel attention mechanism to process the typhoon cloud image feature information of the typhoon to be estimated to obtain weighted typhoon cloud image feature information includes: M C =σ(FC2(RELU(FC1(Global Average Pooling(F)+Global Max Pooling(F))))) F C =F·M C Among them, M C is the channel attention weight, σ is the activation function, FC1 and FC2 are fully connected layers, RELU is the nonlinear activation function, Global Auerage Pooling is the global average pooling result, F is the typhoon cloud image feature information, Global MaxPooling is the global maximum pooling result, F C It is the weighted typhoon cloud image characteristic information.

9. The typhoon intensity estimation method integrating temporal and spatial dynamic characteristics according to claim 8, characterized in that: The step of using the spatial attention mechanism to weight each position in the weighted typhoon cloud image feature information to obtain the final typhoon cloud image feature information includes: M S =σ(Conv7(RELU(Conv1([Average Pooling(F C ),Max Pooling(F C )])))) F final =F C ·M S Among them, M S is the spatial attention weight, σ is the activation function, Conv7 is a 7×7 convolution operation, RELU is a nonlinear activation function, Conv1 is a 1×1 convolution operation, Average Pooling is the average pooling result, and F C is the weighted typhoon cloud image feature information, Max Pooling is the maximum pooling result, F final This is the final typhoon cloud map feature information.

10. A typhoon intensity estimation system integrating temporal and spatial dynamic characteristics, characterized in that: The system includes an input device, an output device, a processor and a memory, wherein the input device, the output device, the processor and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the typhoon intensity estimation method integrating temporal and spatial dynamic characteristics as described in any one of claims 1 to 9.

Images