Method and system for determining a historical similar typhoon based on large-scale environmental field similarity

Abstract

The application discloses a kind of based on large-scale environment field similarity historical similar typhoon determination method and system, the method includes: obtaining the state information of current target typhoon and corresponding large-scale environment field data, processing and generating target subtropical high binary image with typhoon as center;Construct historical subtropical high binary image library;Based on bayesian network model, subtropical high image pixel block is node, the effective area between target and historical subtropical high image, coincidence effective area and similarity coefficient are calculated by learning its conditional probability dependency relationship;Further, the similarity degree result is sequentially subjected to space, time and three physical constraints of moving to joint screening, and the optimal similar historical subtropical high list is obtained;Finally, according to the list, determine corresponding historical typhoon and its subsequent path, as the forecast reference of current typhoon, can make similarity discrimination more in line with meteorological mechanism, significantly improve the accuracy and business reference value of historical similar typhoon retrieval.

Description

Technical Field

[0001] This application relates to the field of typhoon retrieval technology, specifically to a method and system for determining historically similar typhoons based on large-scale environmental field similarity. Background Technology

[0002] Tropical cyclones (TCs, commonly known as typhoons in the Northwest Pacific, hereinafter referred to as typhoons) are highly destructive weather systems, ranked first among the world's ten major natural disasters. Wind movement is influenced by a variety of complex factors, but practice shows that steering flow is the dominant mechanism. Therefore, the distribution of large-scale pressure fields and their corresponding extensive and deep environmental basic airflows constitute important bases for typhoon track forecasting. The subtropical high (hereinafter referred to as "subtropical high"), as a key circulation system influencing typhoon tracks, largely determines the future movement trend of typhoons and their accompanying wind and rain impacts through its periodic changes, morphological characteristics, and interactions with systems such as westerly troughs. Therefore, conducting historical similar typhoon searches based on the similarity of subtropical high morphology and circulation background is expected to provide more accurate reference information for current typhoon track and impact forecasts.

[0003] Existing research has attempted to employ various similarity metrics for meteorological gridded fields, such as Euclidean distance, Hamming distance, similarity distance, similarity coefficient, correlation coefficient, and optimal similarity coefficient. However, these methods based on "distance" or "correlation analysis" generally have certain limitations, making it difficult to accurately characterize the location and intensity features of key systems (such as the subtropical high) in the environmental field. In recent years, machine learning methods have been gradually introduced into subtropical high similarity discrimination. For example, some studies have proposed a subtropical high similarity retrieval method based on an improved version of the Visual Geometry Group (VGG16) model, and use Learned Perceptual Image Patch Similarity (LPIPS) as the similarity metric. However, LPIPS itself suffers from domain adaptability bias, and its accuracy is highly dependent on the model pre-trained on natural images (ImageNet), thus limiting its applicability in non-natural image scenarios such as meteorology. Of particular concern is that LPIPS focuses on subjective "perceptual similarity," which may ignore physical differences with clear meteorological significance, such as the morphology and intensity of the subtropical high, thereby affecting its actual effectiveness in operational discrimination. Summary of the Invention

[0004] This application provides a method and system for determining historically similar typhoons based on large-scale environmental field similarity. It can overcome the limitations of traditional distance metrics or general image similarity models in accurately depicting key physical characteristics such as the position, intensity, and shape of the subtropical high, making similarity discrimination more consistent with meteorological mechanisms.

[0005] A first aspect of this application provides a method for determining historically similar typhoons based on large-scale environmental field similarity, the method comprising: Obtain real-time status information of the current target typhoon and large-scale environmental field data at the corresponding time. The large-scale environmental field data is processed to extract and generate a binarized image of the target subtropical high pressure centered on the current typhoon; Construct a historical image library containing binarized images of the subtropical high at various times during historical typhoons; Based on a Bayesian network model, the similarity between the target subtropical high image and each historical subtropical high image in the historical image database is calculated to obtain a similarity set. The results in the similarity set are then filtered by applying spatial constraints, temporal constraints, and directional constraints in sequence to obtain the final list of similar historical subtropical high-altitude images. Based on the list of similar historical subtropical high images, the historical typhoon information and subsequent paths corresponding to the list of similar historical subtropical high images are determined as a reference for the current target typhoon forecast.

[0006] In one possible implementation, processing the large-scale environmental field data to extract and generate a binarized image of the target subtropical high centered on the current typhoon includes: The potential height grid field of the predetermined isobaric surface is extracted from the large-scale environmental field data. The potential height grid field is binarized, and grid points with potential height values ​​greater than or equal to a first threshold are marked as first values, and grid points with potential height values ​​less than the first threshold are marked as second values, thereby generating a sub-highland binarized field. Using the center latitude and longitude of the current target typhoon as a reference, a rectangular region with a predetermined latitude and longitude span is cropped from the binarized field of the subtropical high as the image of the target subtropical high.

[0007] In one possible implementation, constructing a historical image library containing binarized images of the subtropical high at various times during historical typhoons includes: Determine the central location information of each historical typhoon at various moments in its life cycle from the historical typhoon database; Acquire large-scale historical environmental field data that matches each historical typhoon at each moment; For each historical typhoon-time combination, a binarized image of the corresponding historical subtropical high is generated, and the images are compiled to obtain the historical image library.

[0008] In one possible implementation, the similarity between the target subtropical high image and each historical subtropical high image in the historical image database is calculated based on a Bayesian network model to obtain a similarity set, including: Each valid pixel unit in the target subalpine image is used as a random variable node; Using the data from the historical image database, a directed acyclic graph describing the dependencies between pixel units is constructed through a structure learning algorithm, and a corresponding conditional probability table is generated through parameter learning. Based on the directed acyclic graph and the conditional probability table, the effective area of ​​the target subtropical high image, the effective area of ​​each historical subtropical high image, and the effective overlapping area between the target subtropical high image and each historical subtropical high image are calculated respectively. The similarity coefficient of each pair of images is calculated based on the effective area of ​​the target subtropical high image, the effective area of ​​each historical subtropical high image, and the effective overlapping area between the target subtropical high image and each historical subtropical high image.

[0009] In one possible implementation, the formulas for calculating the overlapping effective area and the similarity coefficient are shown below, respectively: , In the formula, For the i-th pixel block in the target sub-height map A, For the i-th pixel block in the historical subtropical high image B, , They are , The parent pixel block, , They are , The conditional probability of its parent pixel block. For indicator functions; The formula for calculating the similarity coefficient is as follows: , In the formula, Let be the similarity coefficient between subtropical highs A and B. The effective overlapping area of ​​the two is [area]. Each represents its own effective area.

[0010] In one possible implementation, the results in the similarity set are sequentially filtered by applying spatial constraints, temporal constraints, and directional constraints to obtain a final list of similar historical subtropical high-altitude images. The application of spatial constraints includes: Obtain the current typhoon center location corresponding to the target subtropical high image, and the historical typhoon center location corresponding to each historical subtropical high image to be screened; Calculate the geographical distance between the current typhoon center location and the center location of each historical typhoon; Historical subtropical high images corresponding to geographical distances less than or equal to a preset distance threshold are selected to form a primary filtering list after spatial constraints.

[0011] In one possible implementation, the results in the similarity set are sequentially filtered by applying spatial constraints, temporal constraints, and directional constraints to obtain a final list of similar historical subtropical high-altitude images. The application of temporal constraints includes: Extract the calendar month corresponding to each historical subtropical high image from the primary filtering list; Obtain the calendar month corresponding to the target subtropical high image; Calculate the absolute difference between the target month and each historical month, and filter out the historical subtropical high images corresponding to the absolute difference being less than or equal to a preset monthly threshold, forming a secondary filtering list after spatiotemporal constraints.

[0012] In one possible implementation, the results in the similarity set are sequentially filtered by applying spatial constraints, temporal constraints, and directional constraints to obtain a final list of similar historical subtropical high images. The application of directional constraints includes: For each historical subtropical high image in the secondary filtering list, obtain the movement direction of the corresponding historical typhoon during a preset period before the time the image was generated; Obtain the movement direction of the current target typhoon during the same preset time period before the current moment; Calculate the angle between the current typhoon's direction of movement and the direction of movement of each historical typhoon; From the secondary filtering list, the historical subtropical high image with the smallest angle to the current typhoon's movement direction is selected as the optimal similarity matching result, forming the final filtered list.

[0013] In one possible implementation, determining the historical typhoon information and subsequent paths corresponding to the list of similar historical subtropical high images, as a reference for forecasting the current target typhoon, includes: Based on the final filtered list of similar historical subtropical high images, the corresponding historical typhoon number and the time corresponding to the image are determined. From the historical typhoon database, retrieve the path and location data of the historical typhoon at multiple future forecast times after the corresponding time. The retrieved path location data will be output as reference information for the path forecast of the current target typhoon at the corresponding future forecast time.

[0014] This example provides a method for determining historically similar typhoons based on large-scale environmental field similarity. First, the real-time status of the current target typhoon and its corresponding large-scale environmental field data are acquired. The environmental field data is then processed to extract and generate a binarized image of the target subtropical high centered on the typhoon. Simultaneously, a historical image library containing binarized images of the subtropical high at various times of historical typhoons is constructed. Then, based on a Bayesian network model, by using subtropical high image pixel blocks as nodes and learning their dependencies and conditional probabilities, the effective area, overlapping effective area, and similarity coefficient between the target and historical subtropical highs are calculated, forming a similarity set. Subsequently, spatial, temporal, and directional constraints are sequentially applied to the results in this set for layer-by-layer filtering, resulting in the final optimized list of similar historical subtropical high images. Finally, based on this, the corresponding historical typhoons and their subsequent paths were determined as a reference for forecasting the current target typhoon. By using Bayesian networks to probabilistically model and calculate the similarity of the subtropical high's morphology and structure, the limitations of traditional distance metrics or general image similarity models in accurately depicting the key physical characteristics of the subtropical high's position, intensity, and morphology were overcome. This made the similarity judgment more consistent with meteorological mechanisms. At the same time, by introducing multiple physical constraints of space, time, and direction of movement for joint screening, the high consistency between the retrieved historical typhoons and the current typhoons in key environmental fields, seasonal backgrounds, and movement trends was effectively ensured. This provided more accurate, reliable, and meteorologically significant similar historical cases for typhoon path forecasting, significantly improving the operational reference value and forecast support capability of the search results.

[0015] A second aspect of this application provides a historical similarity typhoon system based on large-scale environmental field similarity, the system comprising: The acquisition unit is used to acquire the real-time status information of the current target typhoon and the corresponding large-scale environmental field data; The first processing unit is used to process the large-scale environmental field data, extract and generate a binarized image of the target subtropical high pressure centered on the current typhoon; The second processing unit is used to construct a historical image library containing binarized images of the subtropical high at various times during historical typhoons. The third processing unit is used to calculate the similarity between the target subtropical high image and each historical subtropical high image in the historical image database based on a Bayesian network model, and obtain a similarity set. The filtering unit is used to apply spatial constraints, temporal constraints and directional constraints to the results in the similarity set in sequence to filter them, and obtain the final filtered list of similar historical subtropical high images. The reference unit is used to determine the historical typhoon information and subsequent path corresponding to the similar historical subtropical high image list, based on the similar historical subtropical high image list, as a reference for the current target typhoon forecast.

[0016] A third aspect of this application provides a terminal including a processor, an input device, an output device, and a memory, wherein the processor, input device, output device, and memory are interconnected, wherein the memory is used to store a computer program, the computer program including program instructions, and the processor is configured to invoke the program instructions to execute the steps of the method for determining historical similar typhoons based on large-scale environmental field similarity in the first aspect of this application.

[0017] A fourth aspect of this application provides a computer-readable storage medium storing a computer program for electronic data interchange, wherein the computer program causes a computer to perform some or all of the steps described in the method for determining historical similar typhoons based on large-scale environmental field similarity in the first aspect of this application.

[0018] A fifth aspect of this application provides a computer program product, comprising a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps described in the method for determining historical similar typhoons based on large-scale environmental field similarity in the first aspect of this application. The computer program product may be a software installation package. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This application provides a schematic diagram of the overall process for determining historically similar typhoons based on large-scale environmental field similarity in its embodiments. Figure 2 This application provides a schematic diagram of the subtropical high similarity discrimination process for determining historical similar typhoons based on large-scale environmental field similarity in an embodiment of the present application. Figure 3 This application provides a comparative diagram of pixel topology and corresponding directed acyclic graph for an embodiment of a historical similar typhoon determination method based on large-scale environmental field similarity. Figure 4This application provides a schematic diagram of the top 12 historical subtropical highs with SC values ​​based on a method for determining historical similar typhoons using large-scale environmental field similarity. Figure 5 This application provides a schematic diagram of the subtropical high region within a 20*20 degree range outward from the typhoon center, based on a historical similarity typhoon determination method using large-scale environmental field similarity. Figure 6 This application provides a schematic diagram comparing the monthly differences between the target subtropical high map and the historical subtropical high map using a method for determining historical similar typhoons based on large-scale environmental field similarity. Figure 7 This application provides a schematic diagram comparing the current typhoon movement with the historical typhoon movement using a method for determining historically similar typhoons based on large-scale environmental field similarity. Figure 8 This application provides a schematic diagram of the overall structure of a historically similar typhoon system based on large-scale environmental field similarity as an embodiment of the present application; Figure 9 This application provides a schematic diagram of the structure of a terminal. Figure label: Acquisition unit-1, first processing unit-2, second processing unit-3, third processing unit-4, filtering unit-5, reference unit-6. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0023] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application can be combined with other embodiments.

[0024] The method for determining historical similar typhoons based on large-scale environmental field similarity is applied to the system for determining historical similar typhoons based on large-scale environmental field similarity. Figure 1 A schematic diagram of the overall process for determining historically similar typhoons based on large-scale environmental field similarity is shown. Figure 1 As shown, it includes: S1. Obtain the real-time status information of the current target typhoon and the corresponding large-scale environmental field data.

[0025] The real-time status information of the target typhoon includes the real-time time, center latitude, and center longitude of the target typhoon; the corresponding large-scale environmental field data includes the center longitude of the target typhoon 6 hours ago, the center latitude of the target typhoon 6 hours ago, the center longitude of the target typhoon 12 hours ago, and the center latitude of the target typhoon 12 hours ago.

[0026] S2. Process the large-scale environmental field data to extract and generate a binarized image of the target subtropical high pressure centered on the current typhoon.

[0027] Step S2 includes: S201. Extract the potential height grid field of the predetermined isobaric surface from the large-scale environmental field data.

[0028] The process involves obtaining a 500hPa isobaric surface geopotential height distribution data file (H500) of the target typhoon at the current real-time moment from the business center, parsing the file to obtain geopotential height grid field data covering the target area, with each grid point corresponding to a latitude and longitude position and its geopotential height value.

[0029] S202. The potential height grid field is binarized, and grid points with potential height values ​​greater than or equal to a first threshold are marked as first values, and grid points with potential height values ​​less than the first threshold are marked as second values, thereby generating a sub-highland binarized field.

[0030] In this method, grid points with a geopotential height value ≥ 588 on the 500hPa isobaric surface are marked as the first value (e.g., "1"), representing the subtropical high region; grid points with a geopotential height value < 588 are marked as the second value (e.g., "0"), representing the non-subtropical high region, thereby converting the continuous geopotential height field into a binary subtropical high distribution field.

[0031] S203. Using the center latitude and longitude of the current target typhoon as a reference, a rectangular region with a predetermined latitude and longitude span is cropped from the binarized field of the subtropical high as the target subtropical high image.

[0032] Specifically, based on the current typhoon center location, a 20°×20° rectangular region is cropped from the subtropical high binarized field. This region covers the typhoon center and the surrounding large-scale environmental field, serving as the target subtropical high binarized image corresponding to the current target typhoon, for subsequent similarity analysis.

[0033] In this example, the above steps were used to extract and construct a binarized image centered on the current typhoon and highlighting the distribution of the subtropical high region from the original large-scale environmental field data. This provides structured input data for subsequent similarity retrieval based on the subtropical high morphology. At the same time, the cropping operation focused on the environmental region most closely related to the typhoon, improving the efficiency and relevance of subsequent calculations.

[0034] S3. Construct a historical image library containing binarized images of the subtropical high at various times during historical typhoons.

[0035] Step S3 includes: S301. Determine the central location information of each historical typhoon at various moments in its life cycle from the historical typhoon database.

[0036] This involves reading historical typhoon datasets (usually stored by year or typhoon number), parsing records for each typhoon at 6-hour or hourly intervals, extracting the corresponding time (time stamp) and the center latitude and longitude of the typhoon at that time, and forming a list of "typhoon-time-location" correspondences.

[0037] S302. Obtain large-scale historical environmental field data that matches each historical typhoon at each time.

[0038] Specifically, based on each "typhoon-time" combination determined in step S301, the 500hPa isobaric geopotential height field data (H500) for the corresponding time is retrieved from the historical reanalysis database or the business archive database to ensure that the environmental field data strictly matches the typhoon location in time and covers a sufficient spatial range.

[0039] S303. For each historical typhoon-time combination, generate a binarized image of the historical subtropical high corresponding to each historical typhoon-time combination, and summarize them to obtain the historical image library.

[0040] In this process, each environmental field data obtained in step S302 is processed in the same way as in steps S201 to S203: First, the geopotential height grid field in the field data is parsed, and then binarized with a threshold of 588 geopotential shm (≥588 is 1, <588 is 0) to generate the global subtropical high binarized field at that moment; then, based on the typhoon center position at that moment, a 20°×20° rectangular area is cropped from the binarized field to obtain the historical subtropical high binarized image corresponding to the "typhoon-moment"; finally, all such images generated by historical typhoons at all times are summarized, numbered and stored to construct a unified format historical subtropical high binarized image library.

[0041] In this example, a subtropical high morphology image library covering all historical typhoons and all key moments was systematically constructed, providing a comprehensive, standardized, and comparable data foundation for subsequent Bayesian network-based similarity retrieval, and ensuring the coverage of historical retrieval and the consistency of calculation.

[0042] S4. Based on the Bayesian network model, calculate the similarity between the target subtropical high image and each historical subtropical high image in the historical image database to obtain a similarity set.

[0043] S401. Each valid pixel unit in the target sub-altitude image is used as a random variable node.

[0044] In this approach, each pixel unit (corresponding to a 1°×1° geographic grid point) with a value of "1" (representing the subtropical high region) in the binarized image of the target subtropical high is regarded as a random variable, and all variables constitute the node set of the Bayesian network.

[0045] S402. Using the data from the historical image library, construct a directed acyclic graph describing the dependencies between pixel units through a structure learning algorithm, and generate a corresponding conditional probability table through parameter learning.

[0046] Based on historical subtropical high-latitude map data, pixel blocks at different latitudes and longitudes are used as random variables. The log-likelihood fitting function is used as the weighting coefficient to measure the dependency between pixel blocks. A heuristic search is employed to determine the parent-child structure layer by layer, constructing a maximum-weighted directed acyclic graph (DAG) from order 1 to K. The log-likelihood fitting function is as follows:

[0047] (1) In the formula, Bayesian network log-likelihood function Defined as joint distribution With a given topology Lower joint conditional probability The likelihood is the product of the logarithms of the variables. This likelihood can be decomposed into two parts: the first part involves only the variables... marginal probability The product of its logarithm; the second part is for each variable in the network. Consider its parent variable set With variables Conditional probability under common conditions It is the weighted sum of its logarithms, where the weights are the joint probabilities. Then, by learning parameters from historical subtropical high-pressure system data, a conditional probability table is constructed.

[0048] S403. Based on the directed acyclic graph and the conditional probability table, calculate the effective area of ​​the target subtropical high image, the effective area of ​​each historical subtropical high image, and the overlapping effective area between the target subtropical high image and each historical subtropical high image.

[0049] Among them, the target sub-highland map is calculated. effective area (2) (3) In the formula, The effective area of ​​target subtropical high A. For pixel blocks in the sub-high elevation map, For pixel blocks The parent pixel block, Let be the conditional probabilities of the two. For indicating functions, when pixel blocks The value is 1 and its conditional probability is greater than the threshold. If the value is 1, then the value is 0; otherwise, the value is 0. This is a preset conditional probability threshold. Historical subtropical high map. effective area The calculation method is the same as above.

[0050] Furthermore, Figure 3 A comparative diagram of the pixel topology graph and its corresponding directed acyclic graph is shown below. Please refer to [link / reference]. Figure 3 The chain-like calculation method for the joint probability between pixel blocks is as follows: (4) In the formula, For pixel blocks to The joint probability, for Take the probability of its corresponding value in the subtropical high-pressure map. for , The conditional probability.

[0051] Furthermore, the target subtropical high map Compared with historical subtropical high map overlapping effective area The calculation is as follows: (5) (6) In the formula, For the target subtropical high in Figure A, the first Pixel blocks, The first one in the historical subtropical high in Figure B Pixel blocks, , They are , The parent pixel block, , They are , The conditional probability of its parent pixel block.

[0052] S404. Calculate the similarity coefficient of each pair of images based on the effective area of ​​the target subtropical high image, the effective area of ​​each historical subtropical high image, and the effective overlapping area between the target subtropical high image and each historical subtropical high image.

[0053] The formula for calculating SC is as follows: (7) In the formula, Let be the similarity coefficient between subtropical highs A and B. The effective overlapping area of ​​the two is [area]. Each represents its own effective area. For example... Figure 4 As shown, Figure 4 A schematic diagram of the historical subtropical high with the top 12 SC values ​​is shown.

[0054] S5. Apply spatial constraints, temporal constraints and directional constraints to the results in the similarity set in sequence to filter them, and obtain the final filtered list of similar historical subtropical high images.

[0055] Step S5 includes three parts: applying spatial constraints, temporal constraints, and directional constraints for screening. First, spatial constraints are applied, such as... Figure 5 As shown, Figure 5 This diagram shows a subtropical high-pressure zone extending 20*20 degrees outward from the typhoon center, including: A1. Obtain the current typhoon center position corresponding to the target subtropical high image, and the historical typhoon center position corresponding to each historical subtropical high image to be screened.

[0056] Specifically, the center latitude and longitude of the target typhoon are extracted from the real-time status information of the current target typhoon obtained in step S1 and used as the current typhoon center location; for each historical subtropical high image to be screened in the historical image library, the center latitude and longitude of the historical typhoon at that moment are parsed from the metadata of the corresponding "historical typhoon-time" combination and used as the historical typhoon center location.

[0057] A2. Calculate the geographical distance between the current typhoon center location and the center location of each historical typhoon.

[0058] Based on latitude and longitude coordinates, the spatial distance between the current typhoon center and each historical typhoon center is calculated using the spherical geographic distance formula or a simplified version of the absolute value of the latitude and longitude difference (unit: degrees). To simplify the calculation and ensure operational efficiency, this method uses the absolute value of the latitude and longitude difference as the distance metric, focusing on the relative positional relationship between the typhoon and the subtropical high.

[0059] A3. Filter out historical subtropical high images whose geographical distance is less than or equal to a preset distance threshold to form a primary filtering list after spatial constraints.

[0060] Among them, the distance between the current typhoon center and the subtropical high is no more than 20 degrees, that is, the preset distance threshold is set to 20 degrees. All historical subtropical high images that meet the condition of "the current distance from the center of the historical typhoon ≤ 20 degrees" are retained and included in the primary screening list. Subsequent constraints will be further processed based on this.

[0061] In this example, by imposing spatial constraints, it is ensured that the retrieved historical similar typhoons are within a similar scale range to the current target typhoon in terms of their core geographical location (typhoon center). This is consistent with the physical influence radius of the subtropical high as a large-scale environmental field, avoiding the decrease in the comparability of the environmental field due to the initial positions of the typhoons being too far apart, thereby improving the meteorological rationality of the similarity search results.

[0062] Then apply time constraints, such as Figure 6 As shown, Figure 6 This diagram illustrates the monthly differences between the target subtropical high and historical subtropical high maps, including: B1. Extract the calendar month corresponding to each historical subtropical high image from the primary filtering list.

[0063] Specifically, for each historical subtropical high image in the primary screening list, the calendar month (January to December) of that moment is parsed based on the timestamp information in the corresponding "historical typhoon-time" combination, and used as the month label for that historical subtropical high image.

[0064] B2. Obtain the calendar month corresponding to the target subtropical high image.

[0065] In this process, based on the real-time time information of the current target typhoon obtained in step S1, the calendar month of the current moment is parsed out and used as the month label for the target subtropical high image.

[0066] B3. Calculate the absolute difference between the target month and each historical month, and filter out the historical subtropical high images corresponding to the absolute difference being less than or equal to the preset monthly threshold, forming a secondary filtering list after spatiotemporal constraints.

[0067] Specifically, the difference between the target subtropical high image and the historical subtropical high image should not exceed one month, meaning the preset month threshold is set to one month. For each historical month label in the primary filtering list, the absolute value of the difference between it and the target month label is calculated. All historical subtropical high images that meet the condition of "absolute difference of month ≤ 1" are retained and included in the secondary filtering list for further processing under subsequent directional constraints.

[0068] In this example, by imposing a time constraint, it is ensured that the retrieved historically similar typhoons are consistent with or highly similar to the current target typhoon in terms of seasonal background. Since the subtropical high system exhibits significant seasonal variations, this constraint effectively eliminates systematic deviations in subtropical high morphology, intensity, and circulation background caused by seasonal differences, thereby enhancing the comparability and forecasting reference value of similar subtropical highs in terms of meteorological evolution patterns.

[0069] Finally, apply the shift constraint, such as Figure 7 The diagram shown compares the current typhoon's movement with historical typhoon movements, including: C1. For each historical subtropical high image in the secondary filtering list, obtain the movement direction of the corresponding historical typhoon during a preset period before the time the image was generated.

[0070] For each historical subtropical high image in the secondary filtering list, the center position (latitude and longitude) of the historical typhoon in the previous 12 hours is extracted from its corresponding "historical typhoon-time" combination. Based on these two positions, the direction of movement of the historical typhoon in the corresponding 12-hour period is calculated (expressed in angles, e.g., 0° is due north, 90° is due east).

[0071] C2. Obtain the movement direction of the current target typhoon during the same preset time period before the current moment.

[0072] In this process, based on the real-time location of the current target typhoon and its location information for the previous 12 hours (i.e., the center longitude and latitude of the target typhoon in the previous 12 hours) obtained in step S1, the movement direction of the current typhoon in the past 12 hours is calculated, and the calculation standard is consistent with that of historical typhoons.

[0073] C3. Calculate the angle between the current typhoon's direction of movement and the direction of movement of each historical typhoon.

[0074] For each historical typhoon in the secondary filtering list, the minimum angle between its movement direction and the current typhoon's movement direction is calculated (ranging from 0° to 180°). This angle directly reflects the degree of consistency between their movement trends; the smaller the angle, the closer their movement directions are.

[0075] C4. From the secondary filtering list, select the historical subtropical high image with the smallest angle to the current typhoon's movement direction as the optimal similarity matching result, and form the final filtered list.

[0076] The process involves comparing the angles between the movement directions of all historical typhoons and the current typhoon, and selecting the historical subtropical high image with the smallest angle as the optimal match. If multiple historical images have the same angle and are all the smallest, they can be retained together. Finally, the optimal match result (or result set) is output as a final filtered list of similar historical subtropical high images.

[0077] In this example, by applying a directional constraint, the spatial and temporal consistency is further ensured, guaranteeing a high degree of agreement between the retrieved historical similar typhoons and the current target typhoon in terms of short-term motion trends. Since the future path of a typhoon is largely influenced by the inertia of its existing directional movement, this constraint strengthens the physical consistency of the retrieval results from a kinematic perspective. This ensures that the ultimately selected historical similar typhoons not only match the environmental field configuration but also possess the most valuable motion characteristics, thus providing a more reliable analogy for subsequent path forecasts.

[0078] S6. Based on the list of similar historical subtropical high images, determine the historical typhoon information and subsequent paths corresponding to the list of similar historical subtropical high images, and use them as a reference for forecasting the current target typhoon.

[0079] Step S6 includes: S601. Based on the final filtered list of similar historical subtropical high images, determine the corresponding historical typhoon number and the time corresponding to the image.

[0080] This process involves parsing the metadata associated with each similar historical subtropical high image in the final filter list, extracting its corresponding historical typhoon number (such as international number or internal business identifier) ​​and the specific time (year, month, day, and hour) represented by the subtropical high image. This information has already been established and stored with each image during the image library construction phase (S3).

[0081] S602. Retrieve the path and location data of the historical typhoon from the historical typhoon database for multiple future forecast times after the corresponding time.

[0082] Specifically, based on each "historical typhoon-time" combination determined in step S601, the database of best paths for historical typhoons or the reanalysis data set are queried to retrieve the predicted or actual location data (including latitude and longitude coordinates) of the historical typhoon for the next 6, 12, 18, 24, 48, and 72 hours from that time. These data represent the subsequent actual movement trajectory of the historical typhoon under similar environmental field configurations.

[0083] S603. The retrieved path location data is output as path forecast reference information for the current target typhoon in the corresponding future forecast time.

[0084] In this process, each set of historical typhoon follow-up path location data obtained in step S602 is mapped to a path forecast reference point for the current target typhoon in the next 6, 12, 18, 24, 48, and 72 hours. Typically, the geometric center of the historical path or the most representative trajectory can be used as a reference, or multiple similar historical paths can be provided in a set as a probabilistic forecast reference. Finally, the data is output in the form of tables, graphs, or data interfaces for forecasters to use in integrated typhoon path forecasting and decision analysis.

[0085] In this example, the steps described above complete the crucial transformation from similar subtropical high images to actionable forecast reference information. This method not only provides similar historical weather maps but also the actual subsequent movement trajectory of the typhoon under these similar backgrounds. This provides an intuitive and physically meaningful analogy based on historical similarities for forecasting the current typhoon's path, significantly enhancing forecasters' decision-making confidence and the interpretability of forecast products.

[0086] This example provides a method for determining historically similar typhoons based on large-scale environmental field similarity. First, the real-time status of the current target typhoon and its corresponding large-scale environmental field data are acquired. The environmental field data is then processed to extract and generate a binarized image of the target subtropical high centered on the typhoon. Simultaneously, a historical image library containing binarized images of the subtropical high at various times of historical typhoons is constructed. Then, based on a Bayesian network model, by using subtropical high image pixel blocks as nodes and learning their dependencies and conditional probabilities, the effective area, overlapping effective area, and similarity coefficient between the target and historical subtropical highs are calculated, forming a similarity set. Subsequently, spatial, temporal, and directional constraints are sequentially applied to the results in this set for layer-by-layer filtering, resulting in the final optimized list of similar historical subtropical high images. Finally, based on this, the corresponding historical typhoons and their subsequent paths were determined as a reference for forecasting the current target typhoon. By using Bayesian networks to probabilistically model and calculate the similarity of the subtropical high's morphology and structure, the limitations of traditional distance metrics or general image similarity models in accurately depicting the key physical characteristics of the subtropical high's position, intensity, and morphology were overcome. This made the similarity judgment more consistent with meteorological mechanisms. At the same time, by introducing multiple physical constraints of space, time, and direction of movement for joint screening, the high consistency between the retrieved historical typhoons and the current typhoons in key environmental fields, seasonal backgrounds, and movement trends was effectively ensured. This provided more accurate, reliable, and meteorologically significant similar historical cases for typhoon path forecasting, significantly improving the operational reference value and forecast support capability of the search results.

[0087] For those consistent with the above, please refer to Figure 8 , Figure 8 This application provides a schematic diagram of the structure of a historically similar typhoon system based on large-scale environmental field similarity as an embodiment of the present application. For example... Figure 8 As shown, the system includes: Acquisition Unit 1 is used to acquire the real-time status information of the current target typhoon and the corresponding large-scale environmental field data; The first processing unit 2 is used to process the large-scale environmental field data, extract and generate a binarized image of the target subtropical high pressure centered on the current typhoon; The second processing unit 3 is used to construct a historical image library containing binarized images of the subtropical high at various times of historical typhoons; The third processing unit 4 is used to calculate the similarity between the target subtropical high image and each historical subtropical high image in the historical image database based on a Bayesian network model, and obtain a similarity set. The filtering unit 5 is used to apply spatial constraints, temporal constraints and directional constraints to the results in the similarity set in sequence to filter them, so as to obtain the final filtered list of similar historical subtropical high images. Reference unit 6 is used to determine the historical typhoon information and subsequent path corresponding to the similar historical subtropical high image list based on the similar historical subtropical high image list, as a reference for the current target typhoon forecast.

[0088] For examples consistent with the above embodiments, please refer to... Figure 9 , Figure 9 A schematic diagram of a terminal structure provided in an embodiment of this application is shown in the figure. It includes a processor, an input device, an output device, and a memory. The processor, input device, output device, and memory are interconnected. The memory is used to store a computer program, which includes program instructions. The processor is configured to call the program instructions. The program includes instructions for performing the following steps. Obtain real-time status information of the current target typhoon and large-scale environmental field data at the corresponding time. The large-scale environmental field data is processed to extract and generate a binarized image of the target subtropical high pressure centered on the current typhoon; Construct a historical image library containing binarized images of the subtropical high at various times during historical typhoons; Based on a Bayesian network model, the similarity between the target subtropical high image and each historical subtropical high image in the historical image database is calculated to obtain a similarity set. The results in the similarity set are then filtered by applying spatial constraints, temporal constraints, and directional constraints in sequence to obtain the final list of similar historical subtropical high-altitude images. Based on the list of similar historical subtropical high images, the historical typhoon information and subsequent paths corresponding to the list of similar historical subtropical high images are determined as a reference for the current target typhoon forecast.

[0089] The above mainly describes the solutions of the embodiments of this application from the perspective of the method execution process. It is understood that, in order to achieve the above functions, the terminal includes the corresponding hardware structure and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments provided herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0090] This application embodiment can divide the terminal into functional units according to the above method example. For example, each function can be divided into a separate functional unit, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this application embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0091] This application also provides a computer storage medium storing a computer program for electronic data interchange, which causes a computer to perform some or all of the steps of any of the historical similar typhoon determination methods based on large-scale environmental field similarity as described in the above method embodiments.

[0092] This application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program that causes a computer to perform some or all of the steps of any of the historical similar typhoon determination methods based on large-scale environmental field similarity as described in the above method embodiments.

[0093] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0094] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0095] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0096] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0097] Furthermore, the functional units in the various embodiments of the application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software program module.

[0098] If the integrated unit is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0099] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage device, which may include: a flash drive, a read-only memory, a random access memory, a magnetic disk, or an optical disk, etc.

[0100] The embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for determining historically similar typhoons based on large-scale environmental field similarity, characterized in that, include: Obtain real-time status information of the current target typhoon and large-scale environmental field data at the corresponding time. The large-scale environmental field data is processed to extract and generate a binarized image of the target subtropical high pressure centered on the current target typhoon; Construct a historical image library containing binarized images of the subtropical high at various times during historical typhoons; Based on a Bayesian network model, the similarity between the binarized image of the target subtropical high and the binarized image of the subtropical high at each time point of each historical typhoon in the historical image database is calculated, resulting in a similarity set, including: Each valid pixel unit in the binarized image of the target subtropical high pressure is used as a random variable node; Using the data from the historical image database, a directed acyclic graph describing the dependencies between pixel units is constructed through a structure learning algorithm, and a corresponding conditional probability table is generated through parameter learning. Based on the directed acyclic graph and conditional probability table, the effective area of ​​the target subtropical high binarized image, the effective area of ​​the subtropical high binarized image of each historical typhoon at each time, and the effective overlapping area between the target subtropical high binarized image and the subtropical high binarized image of each historical typhoon at each time are calculated respectively. Based on the effective area of ​​the target subtropical high binarized image, the effective area of ​​each historical typhoon binarized image of the subtropical high at each time, and the effective overlapping area between the target subtropical high binarized image and each historical typhoon binarized image of the subtropical high at each time, the similarity coefficient of each pair of images is calculated; The formula for calculating the overlapping effective area is as follows: ， In the formula, For the i-th pixel block in the binarized image A of the target subtropical high pressure, The i-th pixel block in the binarized image B of the subtropical high at various historical typhoon times. , They are , The parent pixel block, , They are , The conditional probability of its parent pixel block. For indicator functions; The formula for calculating the effective area of ​​the target subtropical high-pressure binarized image A is as follows: In the formula, The effective area of ​​the binarized image A of the target subtropical high pressure system. For the pixel blocks in the binarized image of the target subtropical high pressure, For pixel blocks The parent pixel block, Let be the conditional probabilities of the two; The calculation method for the binarized image of the subtropical high at various times in the history of typhoons is the same as the calculation method for the binarized image of the target subtropical high. The formula for calculating the similarity coefficient is as follows: ， In the formula, The similarity coefficients are used to represent the binarized image A of the target subtropical high and the binarized image B of the subtropical high at each time point for each historical typhoon. The effective overlap area between the binarized image of the target subtropical high and the binarized image of the subtropical high at each time point of each historical typhoon is calculated. These represent the effective area of ​​the target subtropical high-pressure binarized image and the effective area of ​​the subtropical high-pressure binarized image at each moment of each historical typhoon. Indicator Function The calculation formula is as follows: In the formula, The preset conditional probability threshold is used; spatial constraints, temporal constraints and directional constraints are applied sequentially to the results in the similarity set to filter them, and a final list of binarized images of the subtropical high at each moment of similar historical typhoons is obtained. Based on the list of binarized images of the subtropical high at various times of similar historical typhoons, the historical typhoon information and subsequent paths corresponding to the list of binarized images of the subtropical high at various times of similar historical typhoons are determined, serving as a reference for the forecast of the current target typhoon.

2. The method for determining historically similar typhoons based on large-scale environmental field similarity according to claim 1, characterized in that, The process of processing the large-scale environmental field data to extract and generate a binarized image of the target subtropical high centered on the current target typhoon includes: The potential height grid field of the predetermined isobaric surface is extracted from the large-scale environmental field data. The potential height grid field is binarized, and grid points with potential height values ​​greater than or equal to a first threshold are marked as first values, and grid points with potential height values ​​less than the first threshold are marked as second values, thereby generating a sub-highland binarized field. Using the center latitude and longitude of the current target typhoon as a reference, a rectangular region with a predetermined latitude and longitude span is cropped from the binarized field of the subtropical high to serve as the binarized image of the target subtropical high.

3. The method for determining historically similar typhoons based on large-scale environmental field similarity according to claim 1, characterized in that, The construction of the historical image library, which includes binarized images of the subtropical high at various times during historical typhoons, includes: Determine the central location information of each historical typhoon at various points in its life cycle from the historical typhoon database; Acquire large-scale historical environmental field data that matches each historical typhoon at each moment; For each historical typhoon-time combination, a binarized image of the corresponding historical subtropical high is generated, and the images are compiled to obtain the historical image library.

4. The method for determining historically similar typhoons based on large-scale environmental field similarity according to claim 1, characterized in that, The results in the similarity set are filtered by sequentially applying spatial constraints, temporal constraints, and directional constraints to obtain a final list of binarized subtropical high images of similar historical typhoons at various times. The spatial constraints include: Obtain the current target typhoon center position corresponding to the binarized image of the target subtropical high, and the historical typhoon center position corresponding to the binarized image of the subtropical high at each moment of each historical typhoon to be screened; Calculate the geographical distance between the current target typhoon center location and the center location of each historical typhoon; The subtropical high binarized images of historical typhoons at various times corresponding to geographical distances less than or equal to a preset distance threshold are filtered out to form a spatially constrained primary filtering list.

5. The method for determining historically similar typhoons based on large-scale environmental field similarity according to claim 4, characterized in that, The results in the similarity set are filtered by sequentially applying spatial constraints, temporal constraints, and directional constraints to obtain the final list of binarized subtropical high images of similar historical typhoons at various times. The temporal constraints include: From the primary filtering list, extract the calendar month corresponding to each binarized image of the subtropical high at each moment of a historical typhoon; Obtain the calendar month corresponding to the binarized image of the target subtropical high pressure; Calculate the absolute difference between the target month and each historical month, and filter out the binarized images of the subtropical high at each moment of the historical typhoon corresponding to the absolute difference being less than or equal to a preset monthly threshold, forming a secondary filtering list after spatiotemporal constraints.

6. The method for determining historically similar typhoons based on large-scale environmental field similarity according to claim 5, characterized in that, The results in the similarity set are then filtered by sequentially applying spatial constraints, temporal constraints, and directional constraints to obtain the final list of binarized subtropical high images of similar historical typhoons at various times. The directional constraints include: For each binarized image of the subtropical high at each moment of a historical typhoon in the secondary filtering list, obtain the movement direction of the corresponding historical typhoon during a preset period before the moment the image was generated; Obtain the movement direction of the current target typhoon during the same preset time period before the current moment; Calculate the angle between the current target typhoon's movement direction and the movement direction of each historical typhoon; From the secondary filtering list, the binarized images of the subtropical high at each moment of the historical typhoon with the smallest angle to the current target typhoon's movement direction are selected as the optimal similarity matching results, forming the final filtered list of binarized images of the subtropical high at each moment of the similar historical typhoon.

7. The method for determining historically similar typhoons based on large-scale environmental field similarity according to claim 1, characterized in that, The step of determining the historical typhoon information and subsequent path corresponding to the binarized subtropical high images of similar historical typhoons at various times, based on the list of binarized subtropical high images of similar historical typhoons at various times, as a reference for forecasting the current target typhoon, includes: Based on the final list of binarized images of similar historical typhoons at various times, the corresponding historical typhoon number and the time corresponding to the image are determined. From the historical typhoon database, retrieve the path and location data of the historical typhoon at multiple future forecast times after the corresponding time. The retrieved path location data will be output as reference information for the path forecast of the current target typhoon at the corresponding future forecast time.

8. A historical similarity typhoon system based on large-scale environmental field similarity, characterized in that, include: The acquisition unit is used to acquire the real-time status information of the current target typhoon and the corresponding large-scale environmental field data; The first processing unit is used to process the large-scale environmental field data, extract and generate a binarized image of the target subtropical high pressure centered on the current target typhoon; The second processing unit is used to construct a historical image library containing binarized images of the subtropical high at various times during historical typhoons. The third processing unit is used to calculate the similarity between the binarized image of the target subtropical high and the binarized image of the subtropical high at each time of each historical typhoon in the historical image database, based on a Bayesian network model, to obtain a similarity set, including: Each valid pixel unit in the binarized image of the target subtropical high pressure is used as a random variable node; Using the data from the historical image database, a directed acyclic graph describing the dependencies between pixel units is constructed through a structure learning algorithm, and a corresponding conditional probability table is generated through parameter learning. Based on the directed acyclic graph and conditional probability table, the effective area of ​​the target subtropical high binarized image, the effective area of ​​the subtropical high binarized image of each historical typhoon at each time, and the effective overlapping area between the target subtropical high binarized image and the subtropical high binarized image of each historical typhoon at each time are calculated respectively. Based on the effective area of ​​the target subtropical high binarized image, the effective area of ​​each historical typhoon binarized image of the subtropical high at each time, and the effective overlapping area between the target subtropical high binarized image and each historical typhoon binarized image of the subtropical high at each time, the similarity coefficient of each pair of images is calculated; The formula for calculating the overlapping effective area is as follows: ， In the formula, For the i-th pixel block in the binarized image A of the target subtropical high pressure, The i-th pixel block in the binarized image B of the subtropical high at various historical typhoon times. , They are , The parent pixel block, , They are , The conditional probability of its parent pixel block. For indicator functions; The formula for calculating the effective area of ​​the target subtropical high-pressure binarized image A is as follows: In the formula, The effective area of ​​the binarized image A of the target subtropical high pressure system. For the pixel blocks in the binarized image of the target subtropical high pressure, For pixel blocks The parent pixel block, Let be the conditional probabilities of the two; The calculation method for the binarized image of the subtropical high at various times in the history of typhoons is the same as the calculation method for the binarized image of the target subtropical high. The formula for calculating the similarity coefficient is as follows: ， In the formula, The similarity coefficients are used to represent the binarized image A of the target subtropical high and the binarized image B of the subtropical high at each time point for each historical typhoon. The effective overlap area between the binarized image of the target subtropical high and the binarized image of the subtropical high at each time point of each historical typhoon is calculated. These represent the effective area of ​​the target subtropical high-pressure binarized image and the effective area of ​​the subtropical high-pressure binarized image at each moment of each historical typhoon. Indicator Function The calculation formula is as follows: In the formula, This is a preset conditional probability threshold; The filtering unit is used to apply spatial constraints, temporal constraints and directional constraints to the results in the similarity set in sequence to filter them, and obtain the final list of binarized images of the subtropical high at each moment of similar historical typhoons after filtering. The reference unit is used to determine the historical typhoon information and subsequent path corresponding to the binarized images of the subtropical high at each moment of the similar historical typhoons, based on the list of binarized images of the subtropical high at each moment of the similar historical typhoons, as a reference for the forecast of the current target typhoon.

Images