A Multi-Source Real-Time Correction Method and Apparatus Based on GFDL-VortexTracker Recognition Results

Abstract

This invention discloses a multi-source real-time correction method and apparatus based on GFDL-VortexTracker identification results, relating to the field of meteorological data processing. The method includes: S1, acquiring model data and multi-source observation data of the output field of a numerical weather prediction model; S2, performing initial identification on the model data of the output field using the GFDL-VortexTracker algorithm to obtain model prior identification results; S3, using the model prior identification results to perform strong constraint correction on authoritative real-time path data to obtain strong constraint correction results; S4, using radar mosaic and satellite cloud images to perform high-frequency observation compensation on the authoritative real-time path data after strong constraint correction to obtain high-frequency observation compensation results; S5, fusing the model prior identification results, strong constraint correction results, and high-frequency observation compensation results to generate real-time correction results. This method significantly improves the accuracy of key parameters such as center position, strong wind radius, and asymmetric structure.

Description

Technical Field

[0001] This invention relates to the field of meteorological data processing, and in particular to a multi-source real-time correction method and apparatus based on GFDL-VortexTracker identification results. Background Technology

[0002] Typhoon identification and parameter extraction are key aspects of numerical weather forecasting and disaster prevention and mitigation operations.

[0003] In the existing technology, the mainstream typhoon identification methods (such as the GFDL-VortexTracker algorithm, hereinafter referred to as "the existing technology") mainly rely on data such as wind field and pressure field output by numerical models to identify vortices, locate the typhoon center through Barnes analysis and iterative search, and calculate its intensity and structural parameters.

[0004] However, those skilled in the art have found in practice that the aforementioned prior art has at least the following technical problems in application, which significantly limit its recognition accuracy and business practicality:

[0005] 1. The identification results heavily rely on models and are disconnected from authoritative real-time data. Existing technologies depend on numerical models. When there are errors in the initial field of the model (such as vortex positioning deviation) or defects in the parameterization of physical processes (such as boundary layer friction coefficient and sea surface temperature forcing errors), the identification results are prone to deviating from the actual typhoon trajectory, and this error accumulates as the forecast lead time increases (for example, the track forecast often systematically deviates eastward by 20-50 km). Crucially, these identification results are disconnected from authoritative real-time bulletins issued by official bodies such as the China Meteorological Administration (CMA), often requiring extensive manual intervention to meet operational needs, which significantly increases the lag in warning issuance.

[0006] 2. Fragmented utilization of observational data and insufficient dynamic response capability. Existing technologies have failed to form a synergistic constraint mechanism of "authoritative real-time data + high-frequency observations". On the one hand, authoritative real-time data (such as CMA tracks) is updated at a low frequency (once every 3 hours), failing to capture the short-term abrupt changes in typhoon behavior (such as rapid intensification and serpentine track shifts) in a timely manner. On the other hand, the application of high-frequency observational data such as satellite and radar is often limited to the correction of single parameters (e.g., using only satellite positioning centers), failing to leverage the complementarity of multi-source data (e.g., satellites have wide coverage but low accuracy near the coast, radar has high accuracy but is limited to land), resulting in large estimation errors (often reaching 30%-50%) for fine-grained structural parameters such as strong wind radius.

[0007] 3. Structural parameter estimates are crude and lack observational verification. The core parameters output by existing technologies (such as center location, strong wind radius, and symmetry indices) rely solely on model dynamic fields for derivation, lacking effective verification and constraints from real-world observations. For example, their estimates of the strong wind radius (areas with Beaufort scale ≥ 6) deviate significantly from actual data, potentially leading to over- or under-warning of secondary disasters such as storm surges. Furthermore, structural parameters such as typhoon eyewall symmetry and maximum wind speed radius are disconnected from actual satellite / radar observations, thus affecting the refined forecasting of downstream wind and rain distribution.

[0008] 4. Parameter adjustment lacks an adaptive mechanism and has poor universality. Most existing technologies rely on fixed thresholds (e.g., manual correction is only performed when the position deviation is >50km) or fixed empirical formulas to adjust parameters. This "one-size-fits-all" approach does not consider the differences between different typhoon types (e.g., ocean-going vs. landfall-bound) and different intensity levels (e.g., tropical storm vs. super typhoon), resulting in unstable correction effects. Furthermore, its data interface lacks compatibility with existing business systems, making automation and seamless integration difficult. Summary of the Invention

[0009] The purpose of this invention is to design a multi-source real-time correction method and apparatus based on GFDL-VortexTracker recognition results in order to solve the above problems.

[0010] The present invention achieves the above objectives through the following technical solutions:

[0011] Multi-source real-time correction methods based on GFDL-VortexTracker recognition results include:

[0012] S1. Obtain model data and multi-source observation data from the numerical weather prediction model output field. The multi-source observation data includes authoritative real-time track data released by the China Meteorological Administration (CMA). Radar mosaics and satellite cloud images provide authoritative real-time path data. Includes live center location Center pressure of intensity level and maximum wind speed ;

[0013] S2. Initially identify the pattern data of the output field using the GFDL-VortexTracker algorithm to obtain the pattern prior identification result;

[0014] S3. Utilize pattern prior recognition results to analyze authoritative real-world path data. Perform strong constraint correction to obtain the strong constraint correction result;

[0015] S4. Use radar mosaic and satellite cloud imagery to perform high-frequency observation compensation on the authoritative real-time path data after strong constraint correction, and obtain high-frequency observation compensation results.

[0016] S5. The model prior identification results, strong constraint correction results and high frequency observation compensation results are fused to generate real-time correction results.

[0017] A multi-source real-time correction device based on GFDL-VortexTracker recognition results includes:

[0018] Storage; storage is used to store computer programs;

[0019] An actuator; the actuator is used to execute a computer program in the storage, which, when executed, implements the multi-source real-time correction method based on the GFDL-VortexTracker identification results as described above.

[0020] The beneficial effects of this invention are as follows: By introducing the prior pattern recognition results identified by the GFDL-VortexTracker algorithm to strongly constrain authoritative real-time path data, the problem of the disconnect between pattern recognition results and official monitoring is fundamentally solved, ensuring the consistency between the output results and the operational benchmark; by utilizing high-frequency radar and satellite observation data, supplemented by high-frequency observation for dynamic compensation and correction, the insufficient update frequency of CMA real-time data is effectively compensated for, enabling it to capture short-term changes in typhoons (such as landfall, rapid intensification), and providing stronger real-time response capabilities; this method significantly improves the accuracy of key parameters such as center location, strong wind radius, and asymmetric structure. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the multi-source real-time correction method based on the GFDL-VortexTracker recognition results of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0023] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0024] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0025] In the description of this invention, it should be understood that the terms "upper," "lower," "inner," "outer," "left," "right," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use, or the orientation or positional relationship commonly understood by those skilled in the art. They are only used to facilitate the description of this invention and to simplify the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0026] Furthermore, the terms "first," "second," etc., are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0027] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, terms such as "set" and "connection" should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0028] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0029] like Figure 1 As shown, the multi-source real-time correction method based on GFDL-VortexTracker recognition results includes:

[0030] S1. Obtain model data and multi-source observation data from the numerical weather prediction model output field. The multi-source observation data includes authoritative real-time track data released by the China Meteorological Administration (CMA). Radar mosaics and satellite cloud images provide authoritative real-time path data. Includes live center location Center pressure of intensity level and maximum wind speed The radar mosaic consists of combined reflectivity and radial wind field data from the CINRAD network. The satellite cloud imagery consists of infrared / visible cloud data from the FY-4 / Sunflower series. Especially the brightness of the cloud top The model data includes zonal winds at the T0 reporting time. , wind direction Sea level pressure and the potential height of different isobaric surfaces ;

[0031] Through spatiotemporal interpolation and coordinate transformation, all model data and observation data are unified to the same spatiotemporal reference (e.g., a unified 0.1°×0.1° grid with a 1-hour interval).

[0032] S2. Initial identification of the output field model data is performed using the GFDL-VortexTracker algorithm to obtain the model prior identification results; specifically, the GFDL-VortexTracker algorithm locates the maximum vortex closure region through Barnes analysis and iterative search, and extracts the point of lowest central pressure as the initial position of the typhoon center. =( Simultaneously, the initial structural parameters are calculated, including the maximum wind speed. strong wind radius and symmetry index Output the initial parameters and initial position as the pattern prior recognition result sequence. .

[0033] S3. Utilize pattern prior recognition results to analyze authoritative real-world path data. Perform strong constraint correction to obtain the strong constraint correction result, ensuring that the prior pattern recognition result is consistent with the official real-world benchmark; specifically:

[0034] Position deviation calculation: Calculation Time-based prior position Authoritative real-time path data spherical distance between , is represented as:

[0035] ;

[0036] Where R is the Earth's radius; , These are the prior positions of the patterns. and authoritative real-time path data Latitude; , These are the differences in latitude and longitude, respectively.

[0037] Position threshold correction: Set the position correction threshold And adjust the threshold according to the location. Perform position correction to obtain the corrected live center position. Specifically: if the spherical distance If the determination mode result is distorted, a position translation correction will be performed, which will correct the background wind center. As the corrected live center position , is represented as: Conversely, the actual center position. As the corrected live center position ;

[0038] Intensity threshold correction: Sets the intensity level deviation threshold. And according to the strength level deviation threshold Maximum wind speed for intensity parameters Make corrections to obtain the corrected maximum wind speed. Specifically: If Then, a linear ratio is used for the maximum wind speed. Adjustments were made to obtain the corrected maximum wind speed. , is represented as: , among which, among which It is an adjustment factor used for smooth transition; conversely, the maximum wind speed is adjusted accordingly. As the corrected maximum wind speed The "Grade" indicates the intensity level. Typhoon intensity levels are determined based on the maximum average wind speed near the center, ranging from 6 to 16, representing the corresponding wind force. This threshold value... It's also based on this value, which is a positive integer. for The absolute value of the difference between the pattern prior identification result at a given time and the authoritative CMA real-time data on typhoon intensity level;

[0039] S4. Utilize radar mosaics and satellite cloud imagery to perform high-frequency observation compensation on the authoritative real-time track data after strong constraint correction, obtaining high-frequency observation compensation results to improve dynamic response capabilities. Specifically, this includes: within the update interval of the authoritative real-time track data released by the China Meteorological Administration (CMA) (e.g., CMA data is updated at T0+24H, while the current time is T0+25H), using radar mosaics to perform high-frequency observation compensation on the strong wind radius and maximum wind speed radius to obtain the compensated strong wind radius. and the radius of maximum wind speed Specifically, this involves using radar mosaics to perform high-frequency observation compensation for the strong wind radius and the maximum wind speed radius; specifically, it involves acquiring... CINRAD radar radial wind field at any given time ,in For radius, For azimuth; set the wind speed threshold for the preset typhoon level. And based on the wind speed threshold Extracting the radius of strong winds and the radius of maximum wind speed , is represented as: , ;

[0040] Using satellite cloud imagery to determine the location of the center of the real-time situation Asymmetric parameters High-frequency observation compensation was performed to obtain the real-time center position after compensation. Asymmetric parameters Output high-frequency observation compensation results, including real-time center position. Compensated strong wind radius Maximum wind speed radius Asymmetric parameters Specifically: to obtain Satellite cloud top brightness temperature at any moment At the center of the live broadcast Search for the nearest satellite cloud top brightness temperature (TBB) point to obtain the real-time center location. , is represented as: Where S is the location of the center of the real-world position. The search area is centered on a predefined region; by analyzing the gradient and morphology of the satellite cloud top brightness temperature (TBB) field around the eye region, such as eccentricity and helicity, the symmetry index of the cloud image is calculated as an asymmetry parameter. .

[0041] S5. The model prior identification results, strong constraint correction results, and high-frequency observation compensation results are fused to generate real-time correction results; specifically including:

[0042] S51, Definition Pattern prior vector at time step Strongly constrained vectors and high-frequency observation vector , is represented as: , , ; where the strong constraint vector like If there is no CMA update at any time, the most recent one can be used. ;

[0043] S52, Given the pattern prior vector Strongly constrained vectors and high-frequency observation vector Assign prior foundation weights separately Strongly constrained basic weights =0.2 and high-frequency observation basic weight =0.7, and to address data gaps (e.g., no radar in the far sea). Unavailable, or mode (crash), respectively introduce binary availability flags , and ; S53. Calculate the normalized weights and obtain the assigned prior weights. Strongly constrained weights and high-frequency observation weights , is represented as: , , , ,in, This refers to an intermediate variable in the dynamic weighted fusion step, representing the sum of the basic weights of the currently available data sources. It serves as the denominator for normalizing the weights of each data source, because... , and It's a binary flag, taking either 0 or 1. 0 means the merged data is unavailable, so... The numerical range is from 0 to 1;

[0044] S54, Calculate T i The final fused state vector at time step As a result of real-time correction, it is represented as: .

[0045] The beneficial effect of this step is: complete data (B equals 1): When observations are missing (e.g., in the open ocean) ): , The weights are automatically redistributed to ensure that the weights of CMA and pattern increase, which is logical; in extreme cases (only the pattern is affected). ): , The system automatically downgrades to the original output mode, ensuring robustness.

[0046] For all time-limited ( ), the original pattern recognition results Replace with real-time correction results .

[0047] S6. The final fused state vector As initial conditions, the GFDL-VortexTracker algorithm was used to recalculate. All subsequent future typhoon tracks and parameters will be used as re-extrapolated results; this initial condition It includes the typhoon's center location, intensity (maximum wind speed, central pressure), and structural parameters (strong wind radius, symmetry index) corrected by multiple sources such as CMA, radar, and satellite. The "re-extrapolation" mechanism in this step ensures that the future path is calculated based on the most accurate current situation (rather than the original model forecast), thus avoiding "jumps" in the forecast path or physical inconsistencies caused by direct replacement.

[0048] The real-time correction results (T0 time limit) will be corrected in real time. The timeliness) and the re-extrapolation result of step 6(2) The subsequent timeframes are merged to generate a complete, self-consistent, standardized typhoon track, intensity level, and structural parameter sequence covering the entire timeframe (T0+0H to T0+240H). The data format of this output sequence is designed to be completely consistent with the original GFDL-VortexTracker, allowing for seamless direct use by downstream operational early warning systems, achieving plug-and-play functionality.

[0049] A multi-source real-time correction device based on GFDL-VortexTracker recognition results includes:

[0050] Storage; storage is used to store computer programs;

[0051] An actuator; the actuator is used to execute a computer program in the storage, which, when executed, implements the multi-source real-time correction method based on the GFDL-VortexTracker identification results as described above.

[0052] Example: Comparison of real-time correction effects before near-shore typhoon landfall

[0053] This embodiment is used to illustrate the significant technical advantages of the present invention during the critical stage of typhoon landfall compared to "existing technology" (using GFDL-VortexTracker alone) and "routine operations" (relying on CMA real-time data).

[0054] 1) Scenario setting: Typhoon status: A typhoon is currently at time T0+48H, about 100 kilometers from the coastline, and is about to make landfall.

[0055] 2) Data Status:

[0056] a) CMA authoritative live report: The latest issue was released at T0+45H (i.e., 3 hours ago), and the data is subject to lag.

[0057] b) Radar / Satellite: Available in real time (T0+48H).

[0058] 3) Comparative Examples (Performance of Prior Art)

[0059] a) Method A: Using GFDL-VortexTracker alone (pattern prior)

[0060] Results: Due to accumulated initial field errors, the typhoon center location identified at T0+48H systematically deviated northward by 50 kilometers compared to the actual location. Simultaneously, the estimated radius of strong winds... The actual distance is 180 kilometers, which is about 30% higher than the actual distance.

[0061] Drawbacks: Issuing warnings based on these results will lead to incorrect landfall point forecasts and an excessive expansion of the strong wind warning area, resulting in wasted resources and a "crying wolf" effect.

[0062] b) Method B: Relying solely on CMA Live (regular business)

[0063] Result: The salesperson referenced the actual CMA position at T0+45H.

[0064] Limitations: Although the location at T0+45H is accurate, the typhoon may have already shifted its path or accelerated within 3 hours near the coast. Relying on "old data" from 3 hours prior to guide the real-time warning at T0+48H is severely lacking in timeliness and cannot capture short-term changes before landfall.

[0065] c) Embodiment of the present invention, method C:

[0066] At time T0+48H, the method of the present invention automatically performs the following steps:

[0067] i. Strongly constrained authoritative real-world application:

[0068] The system acquires the real-time CMA data at T0+45H and compares it with the model results at T0+45H, identifying a systematic bias in the model that is "50 kilometers northward".

[0069] Based on this, the system performs a preliminary translation correction on the prior model results of T0+48H (from comparative example A).

[0070] ii. High-frequency compensation correction:

[0071] Satellite correction: FY-4 satellite cloud image at time T0+48H The data was retrieved, pinpointing the exact location of the typhoon's eye. The true position was further corrected by adjusting the position of the authoritative reality after the strong constraint correction.

[0072] Radar Correction: Since the typhoon has entered the coverage area of ​​the ground-based radar (CINRAD), the system calls upon radial wind field data and... The formula accurately calculates the true strong wind radius at T0+48H. The distance is 130 kilometers (instead of the 180 kilometers estimated by the model).

[0073] iii. Dynamic weighted fusion: System execution Fusion formula. At this point, (Radar / Satellite) and The weight is the highest.

[0074] The T0+48H position is output by this method. It combines the authority of CMA with the real-time capability of satellites, keeping the position error within 10 kilometers and solving the problem of inaccurate landing point predictions.

[0075] Parameter accuracy: Strong wind radius compared to radar data The distance was corrected to 130 km (error <10%), which solved the problems of coarse model parameters and excessively large warning range.

[0076] Timeliness: Compared to the 3-hour lag of CMA, this invention achieves real-time updates at the minute level (satellite) and hour level (radar), solving the problem of delayed business early warning.

[0077] The technical solutions of the present invention are not limited to the specific embodiments described above. Any technical modifications made in accordance with the technical solutions of the present invention fall within the protection scope of the present invention.

Claims

1. A multi-source real-time correction method based on GFDL-VortexTracker recognition results, characterized in that, include: S1. Obtain model data and multi-source observation data from the numerical weather prediction model output field. The multi-source observation data includes authoritative real-time track data released by the China Meteorological Administration (CMA). Radar mosaics and satellite cloud images provide authoritative real-time path data. Includes live center location Center pressure of intensity level and maximum wind speed ; S2. Initially identify the pattern data of the output field using the GFDL-VortexTracker algorithm to obtain the pattern prior identification result; S3. Utilize pattern prior recognition results to analyze authoritative real-world path data. Perform strong constraint correction to obtain strong constraint correction results; specifically, this includes position deviation calculation, position threshold correction, and intensity threshold correction. S4. Utilize radar mosaics and satellite cloud images to perform high-frequency observation compensation on the authoritative real-time track data after strong constraint correction, obtaining the high-frequency observation compensation results; specifically, within the update interval of the authoritative real-time track data released by the China Meteorological Administration (CMA), use radar mosaics to perform high-frequency observation compensation on the strong wind radius and maximum wind speed radius, obtaining the compensated strong wind radius. and the radius of maximum wind speed Using satellite cloud imagery to pinpoint the location of the real-time center Asymmetric parameters High-frequency observation compensation was performed to obtain the real-time center position after compensation. Asymmetric parameters ; Output high-frequency observation compensation results, including real-time center position. Compensated strong wind radius Maximum wind speed radius Asymmetric parameters ; S5. The model prior identification results, strong constraint correction results, and high-frequency observation compensation results are fused to generate real-time correction results; specifically including: S51, Definition Pattern prior vector at time step Strongly constrained vectors and high-frequency observation vector , is represented as: , , ; S52, Given the pattern prior vector Strongly constrained vectors and high-frequency observation vector Assign prior foundation weights separately Strongly constrained basic weights =0.2 and high-frequency observation basic weight =0.7, and introduce binary availability flags respectively. , and ; S53. Calculate the normalized weights and obtain the assigned prior weights. Strongly constrained weights and high-frequency observation weights , is represented as: , , , ,in, The sum of the basic weights; S54, Calculation The final fused state vector at time step As a result of real-time correction, it is represented as: .

2. The multi-source real-time correction method based on GFDL-VortexTracker recognition results according to claim 1, characterized in that, The model data includes zonal winds at the start time of T0. , direction of wind Sea level pressure and the potential height of different isobaric surfaces In S2, the GFDL-VortexTracker algorithm uses Barnes analysis and iterative search to locate the maximum vortex closure region and extract the point of lowest central pressure as the initial position of the typhoon center. =( Simultaneously, the initial structural parameters are calculated, including the maximum wind speed. strong wind radius and symmetry index Initial parameters and initial position are used as pattern prior recognition results. .

3. The multi-source real-time correction method based on GFDL-VortexTracker recognition results according to claim 2, characterized in that, S3 includes: utilizing pattern prior recognition results to analyze authoritative real-world path data. Perform strong constraint correction to obtain the strong constraint correction result, specifically: Position deviation calculation: Calculation Time-based prior position Authoritative real-time path data spherical distance between , is represented as: ； Where R is the Earth's radius; , These are the prior positions of the patterns. and authoritative real-time path data Latitude; , These are the differences in latitude and longitude, respectively. Position threshold correction: Set the position correction threshold And adjust the threshold according to the location. Perform position correction to obtain the corrected live center position. ; Intensity threshold correction: Sets the intensity level deviation threshold. And according to the strength level deviation threshold Maximum wind speed for intensity parameters Make corrections to obtain the corrected maximum wind speed. ; Output the strongly constrained correction results, which include the maximum wind speed. and the live center location .

4. The multi-source real-time correction method based on GFDL-VortexTracker recognition results according to claim 3, characterized in that, Threshold adjustment based on location The specific position correction is as follows: if the spherical distance If the determination mode result is distorted, a position translation correction will be performed, which will correct the background wind center. As the corrected live center position , is represented as: ; Conversely, the center position of the live feed. As the corrected live center position ; Based on the intensity level deviation threshold Maximum wind speed for intensity parameters The specific correction is as follows: If Then, a linear ratio is used for the maximum wind speed. Adjustments were made to obtain the corrected maximum wind speed. , is represented as: , among which, among which It is an adjustment factor used for smooth transition; conversely, the maximum wind speed is adjusted accordingly. As the corrected maximum wind speed In this context, Grade represents the intensity level.

5. The multi-source real-time correction method based on GFDL-VortexTracker recognition results according to claim 1, characterized in that, The specific steps for high-frequency observation compensation of the strong wind radius and maximum wind speed radius using radar mosaic are as follows: Obtaining... CINRAD radar radial wind field at any given time ,in For radius, For azimuth; set the wind speed threshold for the preset typhoon level. And based on the wind speed threshold Extracting the radius of strong winds and the radius of maximum wind speed , is represented as: , .

6. The multi-source real-time correction method based on GFDL-VortexTracker recognition results according to claim 1, characterized in that, Using satellite cloud imagery to determine the location of the center of the real-time situation Asymmetric parameters The specific steps for high-frequency observation compensation are: acquiring... Satellite cloud top brightness temperature at any moment At the center of the live broadcast Search for the nearest satellite cloud top brightness temperature (TBB) point to obtain the real-time center location. , is represented as: Where S is the location of the center of the real-world position. The search area is centered on a predefined region; by analyzing the gradient and morphology of the satellite cloud top brightness temperature (TBB) field around the eye region, the symmetry index of the cloud image is calculated as an asymmetry parameter. .

7. The multi-source real-time correction method based on GFDL-VortexTracker recognition results according to claim 1, characterized in that, S6, following S5, will finalize the fused state vector. As initial conditions, the GFDL-VortexTracker algorithm was used to recalculate. All subsequent future typhoon trajectories and parameters are used as re-extrapolation results.

8. A multi-source real-time correction device based on GFDL-VortexTracker recognition results, characterized in that, include: Storage; storage is used to store computer programs; Actuator; The actuator is used to execute a computer program in the storage, which, when executed, implements the multi-source real-time correction method based on the identification results of GFDL-VortexTracker as described in any one of claims 1-7.

Images