Rainfall monitoring and early warning system and method based on AI calibration

Abstract

The present application relates to the technical field of rainfall monitoring and early warning, in particular to a rainfall monitoring and early warning system and method based on AI calibration. The method comprises the following steps: collecting rainfall data and rainfall radar images of ground observation stations in each management period, analyzing rainfall trend change conditions, recording grid cells with the same rainfall trend change conditions as a coarse combination group, calculating rainfall change indexes of each grid cell in the coarse combination group based on the rainfall data, obtaining a rainfall statistical path of the ground observation stations based on the rainfall change indexes and a coarse combination algorithm, comparing and analyzing two rainfall radar images of adjacent nodes on the rainfall statistical path to obtain an analysis result, calibrating rainfall data corresponding to the nodes on the rainfall statistical path based on the analysis result, and generating an early warning signal based on the calibrated rainfall data, thereby effectively improving the fine degree of rainfall monitoring and the precise timeliness of early warning.

Description

Technical Field

[0001] This invention relates to the field of rainfall monitoring and early warning technology, specifically to a rainfall monitoring and early warning system and method based on AI calibration. Background Technology

[0002] Rainfall, as a core monitoring indicator in meteorology and hydrology, directly impacts the effectiveness of decision-making in critical scenarios such as flood control and disaster reduction, water resource allocation, and urban operational safety. Traditional rainfall monitoring and early warning technologies are no longer sufficient to meet the current practical needs for refined perception, dynamic tracking, and precise early warning of rainfall processes. Their technological limitations are becoming increasingly apparent, mainly in the following aspects:

[0003] First, traditional methods often employ uniform grid division or simple administrative region division, failing to fully consider the spatial distribution differences of ground observation stations and the spatial heterogeneity of rainfall. This results in the divided areas not accurately matching actual rainfall conditions, affecting the reliability of rainfall trend analysis. Second, in the rainfall data calibration and early warning signal generation stages, existing systems mostly use single calibration methods (such as calibration based on the mean of historical data or linear regression calibration), without incorporating the dynamic characteristics of rainfall changes and spatial correlation information to perform real-time calibration of rainfall data. This leads to significant errors between the calibrated rainfall data and actual rainfall conditions. Furthermore, early warning signal generation often relies on rainfall thresholds from single stations, failing to fully utilize the correlation information of nodes along the rainfall statistical path. This makes it difficult to achieve dynamic prediction and accurate early warning of rainfall development trends, often resulting in delayed or false alarms, and failing to provide timely and effective support for disaster prevention and mitigation decision-making.

[0004] Therefore, there is an urgent need for AI-calibrated rainfall monitoring and early warning systems and methods to solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide a rainfall monitoring and early warning system and method based on AI calibration, aiming to solve the problems of low accuracy, poor timeliness and weak adaptability of traditional rainfall monitoring and early warning technologies.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] On the one hand, an AI-calibrated rainfall monitoring and early warning system, the system includes:

[0008] The data acquisition module is used to generate management cycles, divide the management cycle into several management periods, and collect rainfall data and rainfall radar images from ground observation stations within each management period.

[0009] The grid cell division module is used to divide the spatial distribution of ground observation stations into grid cells based on the Thiessen polygon method, resulting in n grid cells, where each ground observation station corresponds to one grid cell.

[0010] The rainfall statistical path generation module is used to analyze the rainfall trend changes of raster units under the same rainfall threshold within the management period based on the ITA innovative trend change test method. Raster units with the same rainfall trend changes are recorded as raster unit coarse combination groups. The rainfall change index of each raster unit in the raster unit coarse combination group is calculated based on the rainfall data. The rainfall statistical path of the ground observation station is obtained based on the rainfall change index and the coarse combination algorithm. Each raster unit is a node on the rainfall statistical path.

[0011] The calibration and early warning module is used to compare and analyze the pairwise rainfall radar images of adjacent nodes on the rainfall statistics path, obtain the analysis results, calibrate the rainfall data corresponding to the nodes on the rainfall statistics path based on the analysis results, and generate early warning signals based on the calibrated rainfall data.

[0012] Furthermore, based on the Thiessen polygon method, the spatial distribution of ground observation stations is divided into grid cells to obtain n grid cells. The specific process includes the following steps:

[0013] Step 1: Obtain the spatial distribution data of ground observation stations and the raster template of the rainfall monitoring area. The ground observation station data includes the geographic coordinates and unique identifier of the ground observation station. The raster template of the rainfall monitoring area specifies the spatial resolution, geographic coordinate system and boundary of the target raster, and the geographic coordinate system is consistent with the geographic coordinates of the ground observation station.

[0014] Step 2: Generate the Thiessen polygon in vector format: Based on the geographic coordinates of the ground observation stations, construct the Thiessen polygon vector layer using geographic information system tools. First, perform Delaunay triangulation on all ground observation stations, and then draw perpendicular bisectors on each side of each triangle. All perpendicular bisectors intersect to form a closed polygon, which is the Thiessen polygon corresponding to each ground observation station. Each Thiessen polygon contains only one ground observation station, and the polygon boundary is the bisector of the distance between adjacent ground observation stations.

[0015] Step 3: Using the grid template in Step 1 as a reference, determine the number of rows and columns and the spatial range of the grid. Then, rasterize and assign values ​​to the Thiessen polygon vector layer obtained in Step 2. Use the unique identifier corresponding to each Thiessen polygon as the attribute value of the grid unit to ensure that each grid unit corresponds to only one ground observation station.

[0016] Step 4: Use the study area boundary of the raster template in Step 1 to clip the raster layer obtained in Step 3, and remove invalid raster cells that are outside the study area, finally obtaining n raster cells, where n is the number of ground observation stations in the rainfall monitoring area.

[0017] Furthermore, the analysis of rainfall trend changes in grid cells under the same rainfall threshold within the management period based on the ITA innovation trend change test method specifically includes the following process:

[0018] Step 1: Obtain rainfall data for the grid cells within the management period. The rainfall data includes the rainfall in each management period and determines the rainfall threshold.

[0019] Step 2: Divide the rainfall time series of each grid cell into two equal sub-periods. Arrange the rainfall in the two sub-periods in ascending order. Plot the rainfall values ​​of the first half of the sub-period on the X-axis and the rainfall values ​​of the second half of the sub-period on the Y-axis to form a scatter plot.

[0020] Step 3: If all points in the scatter plot fall on the contour lines, it is determined that the rainfall of the grid cell shows a gradual trend as the rainfall threshold decreases. If more than half of the points in the scatter plot fall above the contour lines, it is determined that the rainfall of the grid cell shows an increasing trend as the rainfall threshold decreases. If more than half of the points in the scatter plot fall below the contour lines, it is determined that the rainfall of the grid cell shows a decreasing trend as the rainfall threshold decreases.

[0021] Furthermore, the calculation of the rainfall variation index of each grid cell in the coarse-grained group of grid cells based on rainfall data specifically includes the following process:

[0022] Based on rainfall data, the rainfall corresponding to each management period in each grid cell within the coarse combination group of grid cells is obtained within the management cycle. A rectangular coordinate system is established with the execution time of the management period as the X-axis and the rainfall corresponding to the management period as the Y-axis. Rainfall curves are plotted in the rectangular coordinate system by plotting points. A rainfall constant line is set, and the part of the rainfall curve above the rainfall constant line is marked. The integral value of the marked curve part is calculated, and the integral value is recorded as the rainfall change index of each grid cell.

[0023] Furthermore, the specific process for obtaining the rainfall statistical path of ground observation stations based on the rainfall variation index and coarse combination algorithm includes the following steps:

[0024] Step 1: Obtain the two-dimensional coordinates of the ground observation station corresponding to each raster cell based on the GIS system. The rainfall variation index for each grid cell and Combinations are represented as nodes, and the node representation is as follows: ;

[0025] Step 2: Combine the first and second nodes of each coarse group to form a reference vector. , and the remaining first The node and the first The nodes form a vector ,in, Calculate separately and Angle between Among them, angle The calculation formula is as follows: ;

[0026] Step 3: Put With threshold angle Compare, if the first points Angle greater than the threshold Then it will be arranged in the th order. The nodes preceding each node are subdivided into groups, and the nodes are arranged in order. Repeat steps two and three for the nodes preceding each node to subdivide the coarse combination, and record each subdivided group as a segmented path for rainfall statistics.

[0027] Step 4: Combine the segmented rainfall statistics paths according to the geographical location relationship of the nodes into a rainfall statistics path.

[0028] Furthermore, the pairwise rainfall radar images of adjacent nodes along the rainfall statistics path are compared and analyzed to obtain the following analysis results:

[0029] Obtain pairwise rainfall radar images of adjacent nodes along the rainfall statistics path. and ,in, Represents the set of real numbers. , Images ,image height, , Images ,image The width is 3, where 3 represents the three color channels of the image (RGB).

[0030] Image and images Each image is split into multiple non-overlapping image patch sequences, denoted as follows: , ,in, , These are the dimensions of the image patches, The size of each image patch is represented by [insert image patch size here]. All image patches are flattened, and the image embedding representation is obtained through a linear mapping function, denoted as [insert image embedding representation here]. , ,in, , The length of the embedded representation, , For the dimension of the embedded representation, and Embedding means that the input is fed into the Transformer encoding layer for feature extraction, and the resulting features are as follows: , ;

[0031] based on and Generate graph structure ,in, The node set is used to represent rainfall information. , For the number of nodes, To represent the edge set of rainfall information, The number of sides is ;

[0032] Compute the edge centrality of two adjacent nodes in a graph structure. : ;

[0033] in, For two adjacent nodes , Corresponding edges Edge centrality is used to represent the influence of similarity between two connected nodes. , Representing nodes respectively , Degree centrality;

[0034] set up To mitigate the impact of dense nodes, after normalization, the values ​​are converted into rainfall similarity using a transformation formula. The conversion formula is as follows: ;

[0035] in, yes The maximum value, yes The average value, To limit the maximum similarity of rainfall conditions and ;

[0036] Similarity of rainfall patterns Record this as the analysis result.

[0037] Furthermore, the calibration of rainfall data corresponding to nodes on the rainfall statistics path based on the analysis results specifically includes the following process:

[0038] The similarity of rainfall conditions among all adjacent nodes on the rainfall statistics path is calculated, and the average similarity is determined. For any node on the rainfall statistics path, its similarity to the rainfall conditions of the preceding node is determined. Similarity to the rainfall situation of the next node ,calculate Difference from the average ,calculate Difference from the average ,like Greater than the preset difference threshold and If the difference exceeds a preset threshold, the rainfall data for that node is discarded, and the discarded rainfall data is replaced by the rainfall data from either of its two adjacent nodes.

[0039] Furthermore, a rainfall radar image is an echo image displayed on a radar screen of rainfall information corresponding to ground observation stations detected by a meteorological radar transmitting microwave signals.

[0040] Furthermore, the process of generating an early warning signal based on the calibrated rainfall data includes the following steps:

[0041] Based on the calibrated rainfall data, the rainfall at ground observation stations is statistically analyzed for each management period. When the rainfall during a management period exceeds a preset rainfall threshold, an early warning message is generated and sent to the management personnel.

[0042] On the other hand, there is an AI-calibrated rainfall monitoring and early warning method, which includes:

[0043] The management cycle is generated and divided into several management periods. Rainfall data and rainfall radar images from ground observation stations are collected during each management period.

[0044] The spatial distribution of ground observation stations is divided into grid cells based on the Thiessen polygon method, resulting in n grid cells, where each ground observation station corresponds to one grid cell.

[0045] Based on the ITA innovation trend change test method, the rainfall trend change of grid cells under the same rainfall threshold within the management period is analyzed. Grid cells with the same rainfall trend change are recorded as a coarse combination group of grid cells. The rainfall change index of each grid cell in the coarse combination group is calculated based on the rainfall data. The rainfall statistical path of the ground observation station is obtained based on the rainfall change index and the coarse combination algorithm. Each grid cell is a node on the rainfall statistical path.

[0046] The radar images of adjacent nodes on the rainfall statistics path are compared and analyzed to obtain the analysis results. Based on the analysis results, the rainfall data corresponding to the nodes on the rainfall statistics path are calibrated, and the early warning signal is generated based on the calibrated rainfall data.

[0047] Compared to existing solutions, the beneficial effects achieved by this invention are:

[0048] On the one hand, this invention performs pairwise comparative analysis of rainfall radar images of adjacent nodes on the rainfall statistical path. By utilizing the spatial correlation characteristics of adjacent nodes and the visualization advantages of rainfall radar images, it accurately identifies the sources of deviation in rainfall data, thereby achieving real-time dynamic calibration of rainfall data. Compared with the traditional method of "static calibration based on historical data", the calibration results are more in line with the current rainfall situation, and the error is significantly reduced.

[0049] On the other hand, generating early warning signals based on calibrated full-path node rainfall data, rather than relying on single-site thresholds, can comprehensively consider the spatial propagation trend and intensity changes of rainfall, achieving "dynamic predictive early warning" of rainfall development. This effectively avoids the problems of "local data misjudging the overall situation" and "lagging behind rainfall development" in traditional early warning systems, making early warning signals more accurate and timely. It provides a scientific and effective basis for disaster prevention and mitigation decisions, significantly improves the emergency response capability to extreme rainfall events, and reduces disaster losses. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0051] Figure 1 This is a system block diagram of a rainfall monitoring and early warning system based on AI calibration according to an embodiment of the present invention;

[0052] Figure 2 This is a flowchart illustrating the workflow of an AI-calibrated rainfall monitoring and early warning system according to an embodiment of the present invention.

[0053] Figure 3 This is a flowchart of another AI-calibrated rainfall monitoring and early warning system according to an embodiment of the present invention.

[0054] Figure 4 This is a flowchart illustrating the workflow of an AI-calibrated rainfall monitoring and early warning method according to an embodiment of the present invention. Detailed Implementation

[0055] 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more exemplary embodiments. Numerous specific details are provided in the following description to give a full understanding of exemplary embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced with one or more of the specific details omitted, or other methods, components, steps, etc., can be employed. In other instances, well-known structures, methods, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this disclosure.

[0057] This embodiment provides a rainfall monitoring and early warning system based on AI calibration. Figure 1 This is a system block diagram of a rainfall monitoring and early warning system based on AI calibration, according to an embodiment of the present invention. Figure 1 As shown, the system includes:

[0058] The data acquisition module is used to generate management cycles, divide the management cycle into several management periods, and collect rainfall data and rainfall radar images from ground observation stations within each management period.

[0059] The start time of the management cycle can be the moment when rainfall begins, and the end time of the management cycle can be any moment during the rainfall process. The rainfall radar image is the echo image displayed on the radar screen of the rainfall information corresponding to the ground observation station detected by the microwave signal emitted by the meteorological radar.

[0060] The grid cell division module is used to divide the spatial distribution of ground observation stations into grid cells based on the Thiessen polygon method, resulting in n grid cells, where each ground observation station corresponds to one grid cell.

[0061] The rainfall statistical path generation module is used to analyze the rainfall trend changes of raster units under the same rainfall threshold within the management period based on the ITA innovative trend change test method. Raster units with the same rainfall trend changes are recorded as raster unit coarse combination groups. The rainfall change index of each raster unit in the raster unit coarse combination group is calculated based on the rainfall data. The rainfall statistical path of the ground observation station is obtained based on the rainfall change index and the coarse combination algorithm. Each raster unit is a node on the rainfall statistical path.

[0062] The calibration and early warning module is used to compare and analyze the pairwise rainfall radar images of adjacent nodes on the rainfall statistics path, obtain the analysis results, calibrate the rainfall data corresponding to the nodes on the rainfall statistics path based on the analysis results, and generate early warning signals based on the calibrated rainfall data.

[0063] In summary, this invention collects rainfall data and rainfall radar images from ground observation stations during various management periods. Based on the Thiessen polygon method, the spatial distribution of ground observation stations is divided into grid cells, resulting in n grid cells. Rainfall trend changes are analyzed, and grid cells with similar rainfall trend changes are grouped into coarse-coordinated grid cell sets. The rainfall change index of each grid cell in the coarse-coordinated grid cell set is calculated based on the rainfall change index and the coarse-coordinated algorithm. The rainfall statistical path of the ground observation stations is obtained based on the rainfall change index and the coarse-coordinated algorithm. Pairwise rainfall radar images of adjacent nodes on the rainfall statistical path are compared and analyzed to obtain the analysis results. Based on the analysis results, the rainfall data corresponding to the nodes on the rainfall statistical path is calibrated, and an early warning signal is generated based on the calibrated rainfall data. Through multi-module collaborative innovation, a comprehensive technological upgrade is achieved in rainfall data collection, regional division, path generation, data calibration, and early warning generation, effectively improving the precision of rainfall monitoring and the accuracy and timeliness of early warnings.

[0064] In some embodiments, Figure 2 This is a flowchart illustrating the workflow of an AI-calibrated rainfall monitoring and early warning system according to an embodiment of the present invention. Figure 2 As shown, the spatial distribution of ground observation stations is divided into grid cells based on the Thiessen polygon method to obtain n grid cells. The specific process includes the following steps:

[0065] Step 1: Obtain spatial distribution data of ground observation stations and grid templates for rainfall monitoring areas;

[0066] Among them, the ground observation station data includes the geographic coordinates and unique identifier of the ground observation station, and the raster template of the rainfall monitoring area clearly defines the spatial resolution, geographic coordinate system and boundary of the rainfall monitoring area of ​​the target raster, and the geographic coordinate system is consistent with the geographic coordinates of the ground observation station.

[0067] Step 2: Generate the Thiessen polygon in vector format;

[0068] Specifically, based on the geographic coordinates of ground observation stations, a Thiessen polygon vector layer is constructed using Geographic Information System (GIS) tools. First, Delaunay triangulation is performed on all ground observation stations. Then, perpendicular bisectors are drawn for each side of each triangle. All perpendicular bisectors intersect to form a closed polygon, which is the Thiessen polygon corresponding to each ground observation station. Each Thiessen polygon contains only one ground observation station, and the polygon boundary is the bisector of the distance between adjacent ground observation stations. The GIS tool can be a Geographic Information System (GIS).

[0069] Step 3: Using the grid template in Step 1 as a reference, determine the number of rows and columns and the spatial range of the grid. Then, rasterize and assign values ​​to the Thiessen polygon vector layer obtained in Step 2. Use the unique identifier corresponding to each Thiessen polygon as the attribute value of the grid unit to ensure that each grid unit corresponds to only one ground observation station.

[0070] Step 4: Use the study area boundary of the raster template in Step 1 to clip the raster layer obtained in Step 3, and remove invalid raster cells that are outside the study area, finally obtaining n raster cells, where n is the number of ground observation stations in the rainfall monitoring area.

[0071] In some embodiments, Figure 3 This is a flowchart of another AI-calibrated rainfall monitoring and early warning system according to an embodiment of the present invention, as shown below. Figure 3 As shown, the analysis of rainfall trend changes in raster cells under the same rainfall threshold within the management period based on the ITA (Innovative Trend Analysis Method) method specifically includes the following process:

[0072] Step 1: Obtain rainfall data for the grid cells within the management period. The rainfall data includes the rainfall in each management period and determines the rainfall threshold.

[0073] Determining rainfall thresholds can be done using empirical statistical methods. For example, in a study on the "impact of rainfall on the trend of soil erosion in a certain area," field observations revealed that when the rainfall amount is ≥15mm, significant erosion begins to appear in the area (the runoff threshold is 15mm). If the management period occurs during the "flood season," then "single rainfall amount ≥15mm during the flood season" is set as the threshold, and only the trend of such rainfall events is analyzed.

[0074] Step 2: Divide the rainfall time series of each grid cell into two equal sub-periods. Arrange the rainfall in the two sub-periods in ascending order. Plot the rainfall values ​​of the first half of the sub-period on the X-axis and the rainfall values ​​of the second half of the sub-period on the Y-axis to form a scatter plot.

[0075] Step 3: If all points in the scatter plot fall on the contour lines, it is determined that the rainfall of the grid cell shows a gradual trend as the rainfall threshold decreases. If more than half of the points in the scatter plot fall above the contour lines, it is determined that the rainfall of the grid cell shows an increasing trend as the rainfall threshold decreases. If more than half of the points in the scatter plot fall below the contour lines, it is determined that the rainfall of the grid cell shows a decreasing trend as the rainfall threshold decreases.

[0076] Among them, contour lines are essentially "XY equal lines".

[0077] In some embodiments, calculating the rainfall variation index of each grid cell in a coarse grid cell group based on rainfall data specifically includes the following process:

[0078] Based on rainfall data, the rainfall corresponding to each management period in each grid cell within the coarse combination group of grid cells is obtained within the management cycle. A rectangular coordinate system is established with the execution time of the management period as the X-axis and the rainfall corresponding to the management period as the Y-axis. Rainfall curves are plotted in the rectangular coordinate system by plotting points. A rainfall constant line is set, and the part of the rainfall curve above the rainfall constant line is marked. The integral value of the marked curve part is calculated, and the integral value is recorded as the rainfall change index of each grid cell.

[0079] Specifically, the corresponding rainfall constant is set directly according to the "rainfall event type" of the research focus: If the focus is on heavy rainfall / rainstorm events: refer to the rainstorm level standards of the meteorological department (e.g., the Chinese standard: 24-hour rainfall ≥ 50 mm is rainstorm, so the rainfall constant can be set to 50 mm); If the focus is on effective rainfall for crops: refer to the crop water requirement model (e.g., the critical water requirement value for wheat jointing stage is 15 mm / ten-day period, so the rainfall constant is set to 15 mm / ten-day period); If the focus is on rainfall associated with soil erosion: refer to the "erosive rainfall threshold" in soil erosion research (e.g., the erosive rainfall threshold for exposed soil in a certain area is 12 mm / time, so the rainfall constant is set to 12 mm).

[0080] In some embodiments, obtaining the rainfall statistical path of ground observation stations based on the rainfall variation index and coarse combination algorithm specifically includes the following process:

[0081] Step 1: Obtain the two-dimensional coordinates of the ground observation station corresponding to each raster cell based on the GIS system. The rainfall variation index for each grid cell and Combinations are represented as nodes, and the node representation is as follows: ;

[0082] Step 2: Combine the first and second nodes of each coarse group to form a reference vector. , and the remaining first The node and the first The nodes form a vector ,in, Calculate separately and Angle between Among them, angle The calculation formula is as follows: ;

[0083] Step 3: Put With threshold angle Compare, if the first points Angle greater than the threshold Then it will be arranged in the th order. The nodes preceding each node are subdivided into groups, and the nodes are arranged in order. Repeat steps two and three for each node preceding it to further subdivide the coarse combination, and record each subdivided group as a rainfall statistics segment path; this involves acquiring relevant data from multiple coarse combination groups and calculating the differences in data within each group. Angle between F Analyzing the distribution of angles: This involves analyzing a large number of calculated data... Statistical analysis is performed, and frequency distribution histograms or cumulative distribution curves are plotted to observe the distribution characteristics of angle values. A "critical point" is identified where, near this point, the angle values ​​can effectively distinguish different data categories. A threshold is then determined based on the distribution characteristics. For example, if statistics show that when When the temperature exceeds 25°, the statistical characteristics of rainfall (such as the rate of change of rainfall) will change significantly, so 25° will be set as the maximum value. .

[0084] Step 4: Combine the segmented rainfall statistics paths according to the geographical location relationship of the nodes into a rainfall statistics path.

[0085] In some embodiments, comparing and analyzing the pairwise rainfall radar images of adjacent nodes on the rainfall statistics path to obtain the analysis results specifically includes the following process:

[0086] Obtain pairwise rainfall radar images of adjacent nodes along the rainfall statistics path. and ,in, Represents the set of real numbers. , Images ,image height, , Images ,image The width is 3, where 3 represents the three color channels of the image (RGB).

[0087] Image and images Each image is split into multiple non-overlapping image patch sequences, denoted as follows: , ,in, , These are the dimensions of the image patches, The size of each image patch is represented by [insert image patch size here]. All image patches are flattened, and the image embedding representation is obtained through a linear mapping function, denoted as [insert image embedding representation here]. , ,in, , The length of the embedded representation, , For the dimension of the embedded representation, and Embedding means that the input is fed into the Transformer encoding layer for feature extraction, and the resulting features are as follows: , ;

[0088] based on and Generate graph structure ,in, The node set is used to represent rainfall information. , For the number of nodes, To represent the edge set of rainfall information, The number of sides is ;

[0089] Compute the edge centrality of two adjacent nodes in a graph structure. : ;

[0090] in, For two adjacent nodes , Corresponding edges Edge centrality is used to represent the influence of similarity between two connected nodes. , Representing nodes respectively , Degree centrality;

[0091] set up To mitigate the impact of dense nodes, after normalization, the values ​​are converted into rainfall similarity using a transformation formula. The conversion formula is as follows: ;

[0092] in, yes The maximum value, yes The average value, To limit the maximum similarity of rainfall conditions and ;

[0093] Similarity of rainfall patterns Record this as the analysis result.

[0094] Furthermore, the calibration of rainfall data corresponding to nodes on the rainfall statistics path based on the analysis results specifically includes the following process:

[0095] The similarity of rainfall conditions among all adjacent nodes on the rainfall statistics path is calculated, and the average similarity is determined. For any node on the rainfall statistics path, its similarity to the rainfall conditions of the preceding node is determined. Similarity to the rainfall situation of the next node ,calculate Difference from the average ,calculate Difference from the average ,like Greater than the preset difference threshold and If the difference exceeds a preset threshold, the rainfall data for that node is discarded, and replaced with the rainfall data from either of its two adjacent nodes. Typically, the difference threshold can be set as a multiple of the standard deviation of rainfall similarity, such as 1, 2, or 3 times the standard deviation. Preferably, the difference threshold is 1 standard deviation.

[0096] Furthermore, the process of generating an early warning signal based on the calibrated rainfall data includes the following steps:

[0097] Based on the calibrated rainfall data, the rainfall at ground observation stations is statistically analyzed for each management period. When the rainfall during a management period exceeds a preset rainfall threshold, an early warning message is generated and sent to the management personnel.

[0098] In some embodiments, the present invention also provides a vehicle-mounted mobile weather detection method. Figure 4 This is a flowchart illustrating the workflow of an AI-calibrated rainfall monitoring and early warning method according to an embodiment of the present invention. Figure 4 As shown, the method includes the following steps:

[0099] Step S401: Generate a management cycle, divide the management cycle into several management periods, and collect rainfall data and rainfall radar images from ground observation stations during each management period;

[0100] Step S402: Divide the spatial distribution of ground observation stations into grid cells based on the Thiessen polygon method to obtain n grid cells, where each ground observation station corresponds to one grid cell;

[0101] Step S403: Based on the ITA innovation trend change test method, analyze the rainfall trend change of grid cells under the same rainfall threshold within the management period. Grid cells with the same rainfall trend change are recorded as a coarse combination group of grid cells. The rainfall change index of each grid cell in the coarse combination group is calculated based on the rainfall data. The rainfall statistical path of the ground observation station is obtained based on the rainfall change index and the coarse combination algorithm. Each grid cell is a node on the rainfall statistical path.

[0102] Step S404: Compare and analyze the pairwise rainfall radar images of adjacent nodes on the rainfall statistics path to obtain the analysis results. Based on the analysis results, calibrate the rainfall data corresponding to the nodes on the rainfall statistics path, and generate an early warning signal based on the calibrated rainfall data.

[0103] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0104] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software 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.

[0105] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0106] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods 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 apparatuses or units may be electrical, mechanical, or other forms.

[0107] 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.

[0108] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A rainfall monitoring and early warning system based on AI calibration, characterized in that, The system includes: The data acquisition module is used to generate management cycles, divide the management cycle into several management periods, and collect rainfall data and rainfall radar images from ground observation stations within each management period. The grid cell division module is used to divide the spatial distribution of ground observation stations into grid cells based on the Thiessen polygon method, resulting in n grid cells, where each ground observation station corresponds to one grid cell. The rainfall statistical path generation module is used to analyze the rainfall trend changes of raster units under the same rainfall threshold within the management period based on the ITA innovative trend change test method. Raster units with the same rainfall trend changes are recorded as raster unit coarse combination groups. The rainfall change index of each raster unit in the raster unit coarse combination group is calculated based on the rainfall data. The rainfall statistical path of the ground observation station is obtained based on the rainfall change index and the coarse combination algorithm. Each raster unit is a node on the rainfall statistical path. The specific process for obtaining the rainfall statistical path of ground observation stations based on the rainfall variation index and coarse combination algorithm includes the following steps: Step 1: Obtain the two-dimensional coordinates of the ground observation station corresponding to each raster cell based on the GIS system. The rainfall variation index for each grid cell and Combinations are represented as nodes, and the node representation is as follows: ; Step 2: Combine the first and second nodes of each coarse group to form a reference vector. , and the remaining first The node and the first The vector consists of nodes. ,in, Calculate separately and Angle between Among them, angle The calculation formula is as follows: ; Step 3: Put With threshold angle Compare, if the first points greater than the threshold angle Then it will be arranged in the th order. The nodes preceding each node are subdivided into groups, and the nodes are arranged in order. Repeat steps two and three for the nodes preceding each node to subdivide the coarse combination, and record each subdivided group as a segmented path for rainfall statistics. Step 4: Combine the segmented rainfall statistics paths according to the geographical location relationships of the nodes to form a total rainfall statistics path; The calibration and early warning module is used to compare and analyze the pairwise rainfall radar images of adjacent nodes on the rainfall statistics path, obtain the analysis results, calibrate the rainfall data corresponding to the nodes on the rainfall statistics path based on the analysis results, and generate an early warning signal based on the calibrated rainfall data. The analysis of pairwise rainfall radar images of adjacent nodes along the rainfall statistics path, specifically including the following process, yields the following results: Obtain pairwise rainfall radar images of adjacent nodes along the rainfall statistics path. and ,in, Represents the set of real numbers. , Images ,image height, , Images ,image The width is 3, where 3 represents the three color channels of the image (RGB). Image and images Each image is split into multiple non-overlapping image patch sequences, denoted as follows: , ,in, , These are the dimensions of the image patches, respectively. The size of each image patch is represented by [insert image patch size here]. All image patches are flattened, and the image embedding representation is obtained through a linear mapping function, denoted as [insert image embedding representation here]. , ,in, , The length of the embedded representation, , For the dimension of the embedded representation, and Embedding means that the input is fed into the Transformer encoding layer for feature extraction, and the resulting features are as follows: , ; based on and Generate graph structure ,in, The node set is used to represent rainfall information. , For the number of nodes, To represent the edge set of rainfall information, The number of sides is ; Compute the edge centrality of two adjacent nodes in a graph structure. : ; in, For two adjacent nodes , Corresponding edges Edge centrality is used to represent the influence of similarity between two connected nodes. , Representing nodes respectively , Degree centrality; set up To mitigate the impact of dense nodes, after normalization, the values ​​are converted into rainfall similarity using a transformation formula. The conversion formula is as follows: ; in, yes The maximum value, yes The average value, To limit the maximum similarity of rainfall conditions and ; Similarity of rainfall patterns Record this as the analysis result.

2. The rainfall monitoring and early warning system based on AI calibration according to claim 1, characterized in that, The spatial distribution of ground observation stations is divided into grid cells based on the Thiessen polygon method to obtain n grid cells. The specific process includes the following steps: Step 1: Obtain the spatial distribution data of ground observation stations and the raster template of the rainfall monitoring area. The ground observation station data includes the geographic coordinates and unique identifier of the ground observation station. The raster template of the rainfall monitoring area specifies the spatial resolution, geographic coordinate system and boundary of the target raster, and the geographic coordinate system is consistent with the geographic coordinates of the ground observation station. Step 2: Generate the Thiessen polygon in vector format: Based on the geographic coordinates of the ground observation stations, construct the Thiessen polygon vector layer using geographic information system tools. First, perform Delaunay triangulation on all ground observation stations, and then draw perpendicular bisectors on each side of each triangle. All perpendicular bisectors intersect to form a closed polygon, which is the Thiessen polygon corresponding to each ground observation station. Each Thiessen polygon contains only one ground observation station, and the polygon boundary is the bisector of the distance between adjacent ground observation stations. Step 3: Using the grid template in Step 1 as a reference, determine the number of rows and columns and the spatial range of the grid. Then, rasterize and assign values ​​to the Thiessen polygon vector layer obtained in Step 2. Use the unique identifier corresponding to each Thiessen polygon as the attribute value of the grid unit to ensure that each grid unit corresponds to only one ground observation station. Step 4: Use the study area boundary of the raster template in Step 1 to clip the raster layer obtained in Step 3, and remove invalid raster cells that are outside the study area, finally obtaining n raster cells, where n is the number of ground observation stations in the rainfall monitoring area.

3. The rainfall monitoring and early warning system based on AI calibration according to claim 1, characterized in that, The analysis of rainfall trend changes in grid cells under the same rainfall threshold within the management cycle based on the ITA innovation trend change test method includes the following process: Step 1: Obtain rainfall data for the grid cells within the management period. The rainfall data includes the rainfall in each management period and determines the rainfall threshold. Step 2: Divide the rainfall time series of each grid cell into two equal sub-periods. Arrange the rainfall in the two sub-periods in ascending order. Plot the rainfall values ​​of the first half of the sub-period on the X-axis and the rainfall values ​​of the second half of the sub-period on the Y-axis to form a scatter plot. Step 3: If all points in the scatter plot fall on the contour lines, it is determined that the rainfall of the grid cell shows a gradual trend as the rainfall threshold decreases. If more than half of the points in the scatter plot fall above the contour lines, it is determined that the rainfall of the grid cell shows an increasing trend as the rainfall threshold decreases. If more than half of the points in the scatter plot fall below the contour lines, it is determined that the rainfall of the grid cell shows a decreasing trend as the rainfall threshold decreases.

4. The rainfall monitoring and early warning system based on AI calibration according to claim 1, characterized in that, The calculation of the rainfall variation index of each grid cell in the coarse combination of grid cells based on rainfall data includes the following process: Based on rainfall data, the rainfall corresponding to each management period in each grid cell within the coarse combination group of grid cells is obtained within the management cycle. A rectangular coordinate system is established with the execution time of the management period as the X-axis and the rainfall corresponding to the management period as the Y-axis. Rainfall curves are plotted in the rectangular coordinate system by plotting points. A rainfall constant line is set, and the part of the rainfall curve above the rainfall constant line is marked. The integral value of the marked curve part is calculated, and the integral value is recorded as the rainfall change index of each grid cell.

5. The rainfall monitoring and early warning system based on AI calibration according to claim 1, characterized in that, The calibration of rainfall data corresponding to nodes on the rainfall statistics path based on the analysis results includes the following process: The similarity of rainfall conditions among all adjacent nodes on the rainfall statistics path is calculated, and the average similarity is determined. For any node on the rainfall statistics path, its similarity to the rainfall conditions of the preceding node is determined. Similarity to the rainfall situation at the previous and next nodes ,calculate Difference from the average ,calculate Difference from the average ,like Greater than the preset difference threshold and If the difference exceeds a preset threshold, the rainfall data for that node is discarded, and the discarded rainfall data is replaced by the rainfall data from either of its two adjacent nodes.

6. The rainfall monitoring and early warning system based on AI calibration according to claim 1, characterized in that, Rainfall radar images are echo images displayed on radar screens of rainfall information corresponding to ground observation stations detected by meteorological radar transmitting microwave signals.

7. The rainfall monitoring and early warning system based on AI calibration according to claim 1, characterized in that, Specifically, early warning signals are generated based on calibrated rainfall data. Includes the following processes: Based on the calibrated rainfall data, the rainfall at ground observation stations is statistically analyzed for each management period. When the rainfall during a management period exceeds a preset rainfall threshold, an early warning message is generated and sent to the management personnel.

8. A rainfall monitoring and early warning method based on AI calibration, characterized in that, The method applicable to the AI-calibrated rainfall monitoring and early warning system according to any one of claims 1 to 7, comprising: The management cycle is generated and divided into several management periods. Rainfall data and rainfall radar images from ground observation stations are collected during each management period. The spatial distribution of ground observation stations is divided into grid cells based on the Thiessen polygon method, resulting in n grid cells, where each ground observation station corresponds to one grid cell. Based on the ITA innovation trend change test method, the rainfall trend change of grid cells under the same rainfall threshold within the management period is analyzed. Grid cells with the same rainfall trend change are recorded as a coarse combination group of grid cells. The rainfall change index of each grid cell in the coarse combination group is calculated based on the rainfall data. The rainfall statistical path of the ground observation station is obtained based on the rainfall change index and the coarse combination algorithm. Each grid cell is a node on the rainfall statistical path. The radar images of adjacent nodes on the rainfall statistics path are compared and analyzed to obtain the analysis results. Based on the analysis results, the rainfall data corresponding to the nodes on the rainfall statistics path are calibrated, and the early warning signal is generated based on the calibrated rainfall data.

Images