High-temperature dry heat event monitoring method and system based on multi-source data fusion
By integrating information from meteorological records and satellite imagery through multi-source data fusion technology, the heterogeneity problem in monitoring high-temperature and dry heat events has been solved, enabling efficient event feature extraction and early warning, and improving the accuracy and responsiveness of monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- STATE QIHOU CENT
- Filing Date
- 2026-02-04
- Publication Date
- 2026-05-15
AI Technical Summary
Existing methods for monitoring high-temperature and dry heat events are insufficient to fully reflect the true characteristics of events when faced with complex environments. In particular, under dynamically changing weather conditions, monitoring results are prone to deviation, and the heterogeneity of data from different sources makes information integration complex, making it difficult to form accurate analysis results in a short period of time.
By using multi-source data fusion methods, including format unification, scale alignment, multimodal fusion algorithms, and convolutional neural networks, meteorological records, satellite imagery, and ground observation information are integrated to extract the spatiotemporal evolution patterns of high-temperature and dry heat events and calculate the similarity with historical benchmarks to generate early warning signals.
It enables timely early warning and dynamic management of high-temperature and dry heat events, significantly improving disaster prevention and control effectiveness, and enhancing the accuracy and responsiveness of monitoring results.
Smart Images
Figure CN122045976A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of meteorological monitoring technology, specifically to a method and system for monitoring high-temperature and dry heat events based on multi-source data fusion. Background Technology
[0002] Against the backdrop of increasingly severe climate change, the monitoring and early warning of high-temperature and dry heat events has become an indispensable research area for ensuring social security and economic stability. These extreme weather phenomena have a profound impact on agricultural production, public health, and energy supply; therefore, accurate detection and timely early warning of high-temperature and dry heat events are particularly crucial. Related research is not only an important part of responding to natural disasters but also a core support for promoting the construction of disaster prevention and mitigation systems.
[0003] However, current methods for monitoring high-temperature and dry heat events often fail to fully reflect the true characteristics of the events when faced with complex environments. Many existing methods lack the ability to deeply explore the correlations between data from different sources, resulting in an insufficiently comprehensive portrayal of event characteristics. This is especially true under dynamically changing weather conditions, where monitoring results are prone to bias. This deficiency limits the timeliness and reliability of early warnings, making it difficult to meet practical needs.
[0004] A deeper technical challenge lies in effectively integrating information from different sources and extracting key features. The most immediate challenge is the heterogeneity of data from different sources. For example, meteorological records, satellite imagery, and ground-based observations differ significantly in format, time scale, and spatial resolution, making information integration exceptionally complex. Furthermore, this heterogeneity directly impacts the comprehensive assessment of the characteristics of high-temperature and dry heat events, especially in scenarios requiring rapid response, making it difficult to generate consistent and accurate analytical results in a short time. For instance, when a high-temperature and dry heat event occurs in a region, satellite imagery may show a large area of high temperature, but ground-based observation data, due to uneven distribution, cannot accurately pinpoint the specific affected locations. This inconsistency often leads to a disconnect between monitoring results and actual conditions.
[0005] Therefore, how to overcome the integration challenges caused by data heterogeneity in a complex and ever-changing environment, and on this basis, accurately characterize the dynamic features of high-temperature dry heat events, has become a key issue that urgently needs to be addressed. Summary of the Invention
[0006] To address the above technical problems, this invention provides a method for monitoring high-temperature dry heat events based on multi-source data fusion, comprising the following steps: By acquiring multi-source information from meteorological satellite imagery and ground observation channels, a heterogeneous dataset containing differences in format, time scale, and spatial resolution was obtained. The heterogeneous dataset is formatted and scaled to handle time series inconsistencies and resolution variations, resulting in a standardized dataset. Based on the spatiotemporal attributes in the standardized dataset, a multimodal fusion algorithm is applied to merge features from different sources, and the fused feature set is filtered to obtain an optimized feature set. For the optimized feature set, a convolutional neural network is used to extract high-level abstract representations, capture the spatiotemporal evolution pattern of high-temperature dry heat events, and obtain the event dynamic feature vector. The similarity between the event's dynamic feature vector and a historical benchmark is calculated. If the similarity exceeds a first preset threshold, an early warning signal is generated, and the risk level of a potential high-temperature and dry heat event is determined.
[0007] Preferably, the method for obtaining the standardized dataset includes: For the heterogeneous dataset, data parsing tools are used for preliminary classification and format recognition to obtain a preliminarily organized dataset; The data format of the dataset is uniformly processed using standardized conversion technology. If a mismatch in data fields is detected, the fields are aligned according to a preset mapping rule to determine a data group with a consistent format. For the data set, time series related information is obtained. If the timestamp intervals are not uniform, the time series is smoothed by interpolation to obtain a time-aligned data series. Based on the data sequence, sampling techniques are used to scale the data resolution to address resolution differences. If the resolution is lower than a second preset threshold, an upsampling operation is performed to obtain a data matrix with uniform scale. The data integrity of the data matrix is checked by a data verification tool. If missing values are found, they are filled in by the mean imputation method to determine the data framework after integrity verification. Cluster analysis is used to group the data in the data framework, obtain the grouped data clusters, determine the correlation between the data clusters, and obtain the standardized dataset.
[0008] Preferably, the method for obtaining the optimized feature set includes: By using the spatiotemporal attributes and multi-source data obtained from the standardized dataset, a pre-established multimodal fusion framework is used to initially integrate the data from different sources to obtain initial fusion features; Cross-modal correlation mining is performed on the initial fused features, and the correlation between features is filtered to remove irrelevant features, thereby determining the core correlation feature set; By using the core associated feature set, a feature set reflecting the dynamics of the event is constructed, and the feature set is reorganized to obtain a dynamic feature set; The dynamic feature set is filtered to obtain the optimized feature set.
[0009] Preferably, the method for obtaining the event dynamic feature vector includes: For the optimized feature set, a convolutional network is applied to extract multi-layer features from the data, capturing deep patterns in spatiotemporal evolution to obtain a high-level abstract feature representation; Based on the high-level abstract feature representation, feature optimization operation is performed to filter out the feature subset that is highly correlated with high temperature events and dry heat phenomena, and a refined feature combination is obtained. By constructing an event representation through the refined feature combination, an intermediate feature vector that can reflect the spatiotemporal evolution law is generated, and the core representation of the event is obtained. For the core representation of an event, a mapping method is used to transform it into a dynamic vector to obtain a dynamic feature description. If there are inconsistencies in the data dimensions in the dynamic feature description, the dynamic vector is compressed through dimensionality reduction to unify the data structure and obtain the event dynamic feature vector.
[0010] The present invention also provides a high-temperature dry heat event monitoring system based on multi-source data fusion. The system applies the above-mentioned method and includes: a data acquisition module, a data processing module, a feature optimization module, a dynamic feature acquisition module, and an early warning module. The data acquisition module obtains multi-source information from meteorological satellite imagery and ground observation channels to obtain a heterogeneous dataset containing differences in format, time scale, and spatial resolution. The data processing module is used to unify the format and scale of the heterogeneous dataset, handle time series inconsistencies and resolution variations, and obtain a standardized dataset. The feature optimization module applies a multimodal fusion algorithm to merge features from different sources based on the spatiotemporal attributes in the standardized dataset, and filters the fused feature set to obtain an optimized feature set. The dynamic feature acquisition module is used to extract high-level abstract representations from the optimized feature set using a convolutional neural network, capture the spatiotemporal evolution pattern of high-temperature dry heat events, and obtain dynamic feature vectors of the events. The early warning module is used to calculate the similarity between the dynamic feature vector of the event and the historical benchmark. If the similarity exceeds a first preset threshold, an early warning signal is generated, and the risk level of potential high temperature and dry heat events is determined.
[0011] Preferably, the workflow of the data processing module includes: For the heterogeneous dataset, data parsing tools are used for preliminary classification and format recognition to obtain a preliminarily organized dataset; The data format of the dataset is uniformly processed using standardized conversion technology. If a mismatch in data fields is detected, the fields are aligned according to a preset mapping rule to determine a data group with a consistent format. For the data set, time series related information is obtained. If the timestamp intervals are not uniform, the time series is smoothed by interpolation to obtain a time-aligned data series. Based on the data sequence, sampling techniques are used to scale the data resolution to address resolution differences. If the resolution is lower than a second preset threshold, an upsampling operation is performed to obtain a data matrix with uniform scale. The data integrity of the data matrix is checked by a data verification tool. If missing values are found, they are filled in by the mean imputation method to determine the data framework after integrity verification. Cluster analysis is used to group the data in the data framework, obtain the grouped data clusters, determine the correlation between the data clusters, and obtain the standardized dataset.
[0012] Preferably, the workflow of the feature optimization module includes: By using the spatiotemporal attributes and multi-source data obtained from the standardized dataset, a pre-established multimodal fusion framework is used to initially integrate the data from different sources to obtain initial fusion features; Cross-modal correlation mining is performed on the initial fused features, and the correlation between features is filtered to remove irrelevant features, thereby determining the core correlation feature set; By using the core associated feature set, a feature set reflecting the dynamics of the event is constructed, and the feature set is reorganized to obtain a dynamic feature set; The dynamic feature set is filtered to obtain the optimized feature set.
[0013] Preferably, the workflow of the dynamic feature acquisition module includes: For the optimized feature set, a convolutional network is applied to extract multi-layer features from the data, capturing deep patterns in spatiotemporal evolution to obtain a high-level abstract feature representation; Based on the high-level abstract feature representation, feature optimization operation is performed to filter out the feature subset that is highly correlated with high temperature events and dry heat phenomena, and a refined feature combination is obtained. By constructing an event representation through the refined feature combination, an intermediate feature vector that can reflect the spatiotemporal evolution law is generated, and the core representation of the event is obtained. For the core representation of an event, a mapping method is used to transform it into a dynamic vector to obtain a dynamic feature description. If there are inconsistencies in the data dimensions in the dynamic feature description, the dynamic vector is compressed through dimensionality reduction to unify the data structure and obtain the event dynamic feature vector.
[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention standardizes and unifies data formats, employs multimodal fusion algorithms to extract cross-modal correlation features, utilizes filtering techniques to optimize data quality, combines convolutional neural networks to capture the spatiotemporal dynamics of events, calculates similarity with historical benchmarks to trigger early warnings, and enhances responsiveness by continuously updating feature data in real time. Its core lies in efficiently integrating multi-source information and accurately predicting risk levels, ultimately achieving timely early warning and dynamic management of high-temperature and dry heat events, significantly improving disaster prevention and control effectiveness. Attached Figure Description
[0015] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0016] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention.
[0017] Explanation of reference numerals in the attached figures: 1010, Processor; 1020, Memory; 1030, Input / Output Interface; 1040, Communication Interface; 1050, Bus. Detailed Implementation
[0018] 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.
[0019] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this disclosure should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar terms used in the embodiments of this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0020] Example 1 In this embodiment, as Figure 1 As shown, a method for monitoring high-temperature dry heat events based on multi-source data fusion includes the following steps: S1. By acquiring multi-source information from meteorological satellite imagery and ground observation channels, a heterogeneous dataset containing differences in format, time scale, and spatial resolution is obtained.
[0021] In this embodiment, meteorological records mainly come from meteorological stations and reanalysis data, including daily or hourly temperature, precipitation, humidity, and other elements. The storage format is mostly CSV or NetCDF, and the time scale ranges from hours to years. Satellite imagery data uses the thermal infrared and visible light bands of remote sensing platforms such as MODIS and Landsat to retrieve information such as surface temperature and vegetation index. Its spatial resolution is between meters and kilometers, and it is usually stored in HDF or GeoTIFF format. Ground observation data covers field measurement information such as soil moisture and flux observations. The time series may have uneven acquisition intervals.
[0022] S2. Standardize the format and scale of heterogeneous datasets, handle time series inconsistencies and resolution variations, and obtain standardized datasets.
[0023] The method for obtaining a standardized dataset includes: for heterogeneous datasets, using data parsing tools for preliminary classification and format recognition to obtain a pre-organized dataset; employing standardization transformation techniques to unify the data format of the dataset, and if data field mismatches are detected, aligning the fields according to preset mapping rules to determine data groups with consistent formats; for the data groups, obtaining time series information, and if the timestamp intervals are uneven, smoothing the time series using interpolation methods to obtain time-aligned data sequences; based on the data sequences, scaling the data resolution using sampling techniques to address resolution differences, and if the resolution is below a second preset threshold, performing upsampling to obtain a data matrix with uniform scale; using data validation tools to check the data integrity of the data matrix, and if missing values are found, imputing them using the mean imputation method to determine the data framework after integrity verification; and using cluster analysis to group the data in the data framework to obtain grouped data clusters, determining the correlation between data clusters to obtain a standardized dataset.
[0024] In this embodiment, firstly, the GDAL data parsing tool is used to automatically identify and initially classify data from diverse sources. For different formats such as CSV, NetCDF, HDF, and GeoTIFF, standardization conversion techniques are used to unify them into an efficient array structure that can be manipulated in memory. During this process, if data field mismatches (such as differences in naming "temperature" across different datasets) or unit mismatches are detected, the system will invoke preset mapping rules and conversion factors to perform semantic alignment and unit standardization, generating data sets with consistent formats.
[0025] Then, temporal scale alignment and spatial resolution normalization are performed. For time series, due to the uneven timestamp intervals caused by hourly observations from meteorological stations, daily satellite transits, and irregular ground-based collections, this method employs linear interpolation or cubic spline interpolation algorithms based on the time axis to resample all data sequences to a unified time frequency, generating strictly time-aligned data sequences. At the spatial scale, for resolution variations ranging from meters to kilometers, resampling techniques are used for grid adjustment. Specifically, for low-resolution data (such as reanalysis data), if its resolution is below the second preset threshold set by the system, an upsampling operation is performed, using bilinear or bicubic convolution interpolation to refine it to the target grid; for high-resolution data (such as Landsat imagery), an aggregation method is used to downsample to the same target resolution, ultimately forming a multi-layer data matrix with perfectly matched spatial coordinates.
[0026] Finally, data quality verification and enhancement are performed. A data integrity verification tool is used to scan the multi-layered data matrix to identify missing or outlier values. For missing values, instead of simply using the global mean for imputation, selection is made based on their spatiotemporal characteristics: spatially continuous missing values are imputed using Inverse Distance Weighted (IDW) or Kriging interpolation; temporal series missing values are imputed using temporal interpolation or machine learning models based on neighboring pixel information (such as random forests). Building upon this, to further improve the internal consistency and structural clarity of the dataset, unsupervised clustering analysis is used to initially group multidimensional features, identifying and removing outlier samples that significantly deviate from the main data clusters. The final output is a standardized dataset that achieves a high degree of uniformity in time, space, format, and quality.
[0027] S3. Based on the spatiotemporal attributes in the standardized dataset, apply a multimodal fusion algorithm to merge features from different sources, and filter the fused feature set to obtain an optimized feature set.
[0028] The method for obtaining the optimized feature set includes: using spatiotemporal attributes and multi-source data obtained from standardized datasets, a pre-established multimodal fusion framework is used to initially integrate data from different sources to obtain initial fused features; cross-modal correlation mining is performed on the initial fused features, and the correlation between features is filtered to remove irrelevant features, thus determining the core associated feature set; through the core associated feature set, a feature set reflecting the dynamics of events is constructed, and the feature set is reorganized to obtain a dynamic feature set; the dynamic feature set is filtered to obtain the optimized feature set.
[0029] In this embodiment, feature-level integration is first performed using a standardized dataset with a unified spatiotemporal grid as a benchmark, employing a multimodal fusion framework. Specifically, multidimensional data from meteorological records (such as air temperature and precipitation deficit), satellite inversion products (such as LST surface temperature and NDVI vegetation index), and ground observations (such as soil volumetric water content) are concatenated at each spatiotemporal grid point to form an initial fused feature vector. Subsequently, to uncover deep physical correlations between different modal data and eliminate redundancy, methods such as mutual information or canonical correlation analysis (CCA) are used to quantitatively assess the statistical dependencies between cross-modal features such as meteorological factors, surface conditions, and soil conditions, screening out the core correlation feature set most sensitive to the coordinated changes in high temperature and dry heat, such as combinations of sustained high temperature days, diurnal surface temperature range, soil moisture anomalies, and vegetation stress indices. Based on this, the core features are reorganized and filtered to accurately capture event dynamics. Feature reconstruction aims to construct temporal features that reflect the evolution of events, such as calculating the trend, volatility, or cumulative anomalies of key indicators (e.g., LST, soil moisture) within a sliding time window. Subsequently, spatiotemporal filtering techniques are used to optimize the reconstructed dynamic feature set. In the spatial domain, adaptive Gaussian filtering or bilateral filtering is used to smooth out minor noise while preserving significant thermal anomaly boundaries between pixels. In the temporal domain, Savitzky-Golay filters or wavelet transforms are applied to remove high-frequency noise and highlight the low-frequency trend signal dominated by the dry-heat event; thus, the optimized feature set is obtained.
[0030] S4. For the optimized feature set, a convolutional neural network is used to extract high-level abstract representations, capture the spatiotemporal evolution pattern of high-temperature dry heat events, and obtain the event dynamic feature vector.
[0031] The method for obtaining the dynamic feature vector of an event includes: applying a convolutional network to extract multi-layer features from the optimized feature set, capturing deep patterns in spatiotemporal evolution, and obtaining a high-level abstract feature representation; performing feature optimization based on the high-level abstract feature representation, selecting a subset of features with high correlation to high-temperature events and dry heat phenomena, and obtaining a refined feature combination; constructing an event representation through the refined feature combination, generating an intermediate feature vector that reflects the spatiotemporal evolution law, and obtaining the core representation of the event; and using a mapping method to transform the core representation of the event into a dynamic vector, obtaining a dynamic feature description. If there are inconsistencies in the data dimensions in the dynamic feature description, the dynamic vector is compressed through dimensionality reduction to unify the data structure and obtain the event dynamic feature vector.
[0032] In this embodiment, step S4 optimizes the feature set by using a spatiotemporal convolutional neural network to extract high-level abstract representations. This network employs a dual-branch architecture to process spatiotemporal information in parallel: the spatial feature extraction branch consists of multiple consecutive two-dimensional convolutional and pooling layers, using 3×3 convolutional kernels to slide along the spatial dimension, capturing the spatial distribution patterns and spatial correlation patterns of high temperatures from local pixels to regional scales layer by layer; the temporal evolution capture branch uses one-dimensional convolutional layers to operate along the time axis, learning the dynamic trends and periodic changes of key indicators (such as surface temperature and soil moisture anomalies) within continuous time windows. The outputs of the two branches are concatenated in a fusion layer, and then further integrated across spatiotemporal features through additional convolutional layers, thereby capturing the co-evolution process of high temperature and drought conditions under spatiotemporal coupling, generating high-level abstract feature representations containing complex nonlinear relationships. Based on the obtained high-level abstract features, further feature optimization and vectorization are performed. First, an attention mechanism module is introduced, which automatically selects the feature maps with the highest correlation to the core driving factors of high-temperature and dry-heat events (such as sustained high temperature and water stress) by calculating the weights of feature channels, suppressing redundant information, and obtaining a refined feature combination. Subsequently, these features are combined, flattened, and aggregated and mapped through a fully connected layer to construct a core representation that comprehensively reflects the spatiotemporal evolution of events. Finally, to avoid the curse of dimensionality and unify the data structure, principal component analysis (PCA) or an autoencoder is used to reduce the dimensionality of this high-dimensional core representation, transforming it into a low-dimensional, dense, and highly condensed dynamic feature vector of events.
[0033] S5. Calculate the similarity between the event's dynamic feature vector and the historical benchmark. If the similarity exceeds the first preset threshold, generate an early warning signal and determine the risk level of potential high-temperature and dry heat events.
[0034] In this embodiment, firstly, a historical benchmark database is constructed: using a long-term historical standardized dataset (such as data from the same period over the past 30 years), steps S2 to S4 are used to process the historical standardized data to generate a set of "historical normal state" feature vectors corresponding to each spatiotemporal grid unit, and their statistical distribution is calculated. Subsequently, the similarity between the event dynamic feature vector of each grid point in the current monitoring period and the corresponding historical benchmark is calculated: Mahalanobis distance is used for quantification, where Mahalanobis distance effectively considers the correlation between features; the smaller the value, the more significant the deviation from historical normality, i.e., the stronger the dry heat anomaly. The early warning judgment module compares the real-time calculated Mahalanobis distance with a first preset threshold (95% of the historical distribution of the similarity index). If the similarity index of a grid point exceeds (or falls below, depending on the specific measurement direction) this threshold, it is judged as an anomaly, and the system automatically generates an early warning signal containing geographical location, intensity, and duration. Ultimately, the risk level is dynamically divided based on the degree to which the similarity index exceeds the threshold: multiple risk level thresholds are set, and the degree of exceeding the first preset threshold is mapped to "mild", "moderate", "severe" and "extremely severe" risk levels. The risk level is then corrected by incorporating spatiotemporal persistence analysis (such as the number of consecutive days exceeding the threshold) to generate a global risk level distribution map as the monitoring output.
[0035] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.
[0036] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, it should be understood that the sequence number of each step in the above embodiments does not imply the order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention. The actions or steps recorded in the claims can be performed in a different order than that in the above embodiments and can still achieve the desired result. In addition, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0037] Example 2 In this embodiment, a high-temperature dry heat event monitoring system based on multi-source data fusion includes: a data acquisition module, a data processing module, a feature optimization module, a dynamic feature acquisition module, and an early warning module.
[0038] The data acquisition module obtains multi-source information from meteorological satellite imagery and ground observation channels, resulting in a heterogeneous dataset containing differences in format, time scale, and spatial resolution.
[0039] The data processing module is used to unify the format and scale of heterogeneous datasets, handle time series inconsistencies and resolution variations, and obtain standardized datasets.
[0040] The data processing module's workflow includes: for heterogeneous datasets, using data parsing tools for preliminary classification and format recognition to obtain a pre-organized dataset; employing standardization transformation techniques to unify the data format of the dataset; if data field mismatches are detected, field alignment is performed using preset mapping rules to determine data groups with consistent formats; for data groups, time-series related information is obtained; if timestamp intervals are uneven, interpolation methods are used to smooth the time series to obtain time-aligned data sequences; based on the data sequences, sampling techniques are used to scale the data resolution to address resolution differences; if the resolution is below a second preset threshold, upsampling is performed to obtain a scaled data matrix; data integrity is checked using data validation tools; if missing values are found, they are filled using the mean imputation method to determine the data framework after integrity verification; cluster analysis is used to group the data in the data framework, obtaining grouped data clusters, and the correlation between data clusters is determined to obtain a standardized dataset.
[0041] The feature optimization module applies a multimodal fusion algorithm to merge features from different sources based on the spatiotemporal attributes of the standardized dataset, and then filters the fused feature set to obtain an optimized feature set.
[0042] The workflow of the feature optimization module includes: using spatiotemporal attributes and multi-source data obtained from standardized datasets, a pre-established multimodal fusion framework is used to initially integrate data from different sources to obtain initial fused features; cross-modal correlation mining is performed on the initial fused features, and the correlation between features is filtered to remove irrelevant features, thus determining the core associated feature set; through the core associated feature set, a feature set reflecting the dynamics of events is constructed, and the feature set is reorganized to obtain a dynamic feature set; the dynamic feature set is filtered to obtain an optimized feature set.
[0043] The dynamic feature acquisition module is used to extract high-level abstract representations from the optimized feature set using a convolutional neural network, capture the spatiotemporal evolution pattern of high-temperature dry heat events, and obtain the event dynamic feature vector.
[0044] The workflow of the dynamic feature acquisition module includes: for the optimized feature set, applying a convolutional network to perform multi-layer feature extraction on the data, capturing deep patterns in spatiotemporal evolution, and obtaining a high-level abstract feature representation; based on the high-level abstract feature representation, performing feature optimization operations, filtering out a subset of features with high correlation to high-temperature events and dry heat phenomena, and obtaining a refined feature combination; constructing an event representation through the refined feature combination, generating an intermediate feature vector that can reflect the spatiotemporal evolution law, and obtaining the core representation of the event; for the core representation of the event, using a mapping method to transform it into a dynamic vector, obtaining a dynamic feature description; if there are inconsistent data dimensions in the dynamic feature description, then compressing the dynamic vector through dimensionality reduction processing, unifying the data structure, and obtaining the event dynamic feature vector.
[0045] The early warning module is used to calculate the similarity between the dynamic feature vector of the event and the historical benchmark. If the similarity exceeds the first preset threshold, an early warning signal is generated and the risk level of potential high temperature and dry heat events is determined.
[0046] The system described in the above embodiments is used to implement the corresponding high-temperature dry heat event monitoring method based on multi-source data fusion in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0047] It should be noted that the aforementioned high-temperature dry heat event monitoring system based on multi-source data fusion is embodied in the form of functional units. The term "module" here can be implemented in software and / or hardware, without specific limitations.
[0048] For example, a "module" can be a software program, hardware circuit, or a combination of both that implements the above functions. Hardware circuits may include application-specific integrated circuits (ASICs), electronic circuits, processors (e.g., shared processors, proprietary processors, or group processors) and memory for executing one or more software or firmware programs, combined logic circuits, and / or other suitable components that support the described functions.
[0049] Example 3 Based on the same inventive concept, corresponding to the methods of any of the above embodiments, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the high-temperature dry heat event monitoring method based on multi-source data fusion as described in any of the above embodiments.
[0050] Figure 2This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.
[0051] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.
[0052] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.
[0053] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.
[0054] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB (Universal Serial Bus), network cable, etc.) or wireless means (such as mobile network, WIFI (Wireless Fidelity), Bluetooth, etc.).
[0055] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.
[0056] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.
[0057] The system described in the above embodiments is used to implement the corresponding high-temperature dry heat event monitoring method based on multi-source data fusion in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0058] Example 4 Based on the same inventive concept, corresponding to the methods of any of the above embodiments, the present invention also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the high-temperature dry heat event monitoring method based on multi-source data fusion as described in any of the above embodiments.
[0059] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.
[0060] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the high-temperature dry heat event monitoring method based on multi-source data fusion as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0061] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this disclosure (including the claims) is limited to these examples; within the framework of this disclosure, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this disclosure as described above, which are not provided in detail for the sake of brevity.
[0062] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this disclosure, the provided drawings may or may not show well-known power / ground connections to integrated circuit (IC) chips and other components. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this disclosure, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this disclosure will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuitry) have been set forth to describe exemplary embodiments of this disclosure, it will be apparent to those skilled in the art that the embodiments of this disclosure may be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.
[0063] Although this disclosure has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.
[0064] Therefore, the units of the various examples described in the embodiments of this application 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.
[0065] This disclosure is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.
Claims
1. A method for monitoring high-temperature dry heat events based on multi-source data fusion, characterized in that, Includes the following steps: By acquiring multi-source information from meteorological satellite imagery and ground observation channels, a heterogeneous dataset containing differences in format, time scale, and spatial resolution was obtained. The heterogeneous dataset is formatted and scaled to handle time series inconsistencies and resolution variations, resulting in a standardized dataset. Based on the spatiotemporal attributes in the standardized dataset, a multimodal fusion algorithm is applied to merge features from different sources, and the fused feature set is filtered to obtain an optimized feature set. For the optimized feature set, a convolutional neural network is used to extract high-level abstract representations, capture the spatiotemporal evolution pattern of high-temperature dry heat events, and obtain the event dynamic feature vector. The similarity between the event's dynamic feature vector and a historical benchmark is calculated. If the similarity exceeds a first preset threshold, an early warning signal is generated, and the risk level of a potential high-temperature and dry heat event is determined.
2. The method for monitoring high-temperature dry heat events based on multi-source data fusion according to claim 1, characterized in that, Methods for obtaining the standardized dataset include: For the heterogeneous dataset, data parsing tools are used for preliminary classification and format recognition to obtain a preliminarily organized dataset; The data format of the dataset is uniformly processed using standardized conversion technology. If a mismatch in data fields is detected, the fields are aligned according to a preset mapping rule to determine a data group with a consistent format. For the data set, time series related information is obtained. If the timestamp intervals are not uniform, the time series is smoothed by interpolation to obtain a time-aligned data series. Based on the data sequence, sampling techniques are used to scale the data resolution to address resolution differences. If the resolution is lower than a second preset threshold, an upsampling operation is performed to obtain a data matrix with uniform scale. The data integrity of the data matrix is checked by a data verification tool. If missing values are found, they are filled in by the mean imputation method to determine the data framework after integrity verification. Cluster analysis is used to group the data in the data framework, obtain the grouped data clusters, determine the correlation between the data clusters, and obtain the standardized dataset.
3. The method for monitoring high-temperature dry heat events based on multi-source data fusion according to claim 1, characterized in that, The methods for obtaining the optimized feature set include: By using the spatiotemporal attributes and multi-source data obtained from the standardized dataset, a pre-established multimodal fusion framework is used to initially integrate the data from different sources to obtain initial fusion features; Cross-modal correlation mining is performed on the initial fused features, and the correlation between features is filtered to remove irrelevant features, thereby determining the core correlation feature set; By using the core associated feature set, a feature set reflecting the dynamics of the event is constructed, and the feature set is reorganized to obtain a dynamic feature set; The dynamic feature set is filtered to obtain the optimized feature set.
4. The method for monitoring high-temperature dry heat events based on multi-source data fusion according to claim 1, characterized in that, The methods for obtaining the dynamic feature vector of the event include: For the optimized feature set, a convolutional network is applied to extract multi-layer features from the data, capturing deep patterns in spatiotemporal evolution to obtain a high-level abstract feature representation; Based on the high-level abstract feature representation, feature optimization operation is performed to filter out the feature subset that is highly correlated with high temperature events and dry heat phenomena, and a refined feature combination is obtained. By constructing an event representation through the refined feature combination, an intermediate feature vector that can reflect the spatiotemporal evolution law is generated, and the core representation of the event is obtained. For the core representation of an event, a mapping method is used to transform it into a dynamic vector to obtain a dynamic feature description. If there are inconsistencies in the data dimensions in the dynamic feature description, the dynamic vector is compressed through dimensionality reduction to unify the data structure and obtain the event dynamic feature vector.
5. A high-temperature dry heat event monitoring system based on multi-source data fusion, wherein the system applies the method described in any one of claims 1-4, characterized in that, include: The system includes a data acquisition module, a data processing module, a feature optimization module, a dynamic feature acquisition module, and an early warning module. The data acquisition module obtains multi-source information from meteorological satellite imagery and ground observation channels to obtain a heterogeneous dataset containing differences in format, time scale, and spatial resolution. The data processing module is used to unify the format and scale of the heterogeneous dataset, handle time series inconsistencies and resolution variations, and obtain a standardized dataset. The feature optimization module applies a multimodal fusion algorithm to merge features from different sources based on the spatiotemporal attributes in the standardized dataset, and filters the fused feature set to obtain an optimized feature set. The dynamic feature acquisition module is used to extract high-level abstract representations from the optimized feature set using a convolutional neural network, capture the spatiotemporal evolution pattern of high-temperature dry heat events, and obtain dynamic feature vectors of the events. The early warning module is used to calculate the similarity between the dynamic feature vector of the event and the historical benchmark. If the similarity exceeds a first preset threshold, an early warning signal is generated, and the risk level of potential high temperature and dry heat events is determined.
6. The high-temperature dry heat event monitoring system based on multi-source data fusion according to claim 5, characterized in that, The workflow of the data processing module includes: For the heterogeneous dataset, data parsing tools are used for preliminary classification and format recognition to obtain a preliminarily organized dataset; The data format of the dataset is uniformly processed using standardized conversion technology. If a mismatch in data fields is detected, the fields are aligned according to a preset mapping rule to determine a data group with a consistent format. For the data set, time series related information is obtained. If the timestamp intervals are not uniform, the time series is smoothed by interpolation to obtain a time-aligned data series. Based on the data sequence, sampling techniques are used to scale the data resolution to address resolution differences. If the resolution is lower than a second preset threshold, an upsampling operation is performed to obtain a data matrix with uniform scale. The data integrity of the data matrix is checked by a data verification tool. If missing values are found, they are filled in by the mean imputation method to determine the data framework after integrity verification. Cluster analysis is used to group the data in the data framework, obtain the grouped data clusters, determine the correlation between the data clusters, and obtain the standardized dataset.
7. The high-temperature dry heat event monitoring system based on multi-source data fusion according to claim 5, characterized in that, The workflow of the feature optimization module includes: By using the spatiotemporal attributes and multi-source data obtained from the standardized dataset, a pre-established multimodal fusion framework is used to initially integrate the data from different sources to obtain initial fusion features; Cross-modal correlation mining is performed on the initial fused features, and the correlation between features is filtered to remove irrelevant features, thereby determining the core correlation feature set; By using the core associated feature set, a feature set reflecting the dynamics of the event is constructed, and the feature set is reorganized to obtain a dynamic feature set; The dynamic feature set is filtered to obtain the optimized feature set.
8. The high-temperature dry heat event monitoring system based on multi-source data fusion according to claim 5, characterized in that, The workflow of the dynamic feature acquisition module includes: For the optimized feature set, a convolutional network is applied to extract multi-layer features from the data, capturing deep patterns in spatiotemporal evolution to obtain a high-level abstract feature representation; Based on the high-level abstract feature representation, feature optimization operation is performed to filter out the feature subset that is highly correlated with high temperature events and dry heat phenomena, and a refined feature combination is obtained. By constructing an event representation through the refined feature combination, an intermediate feature vector that can reflect the spatiotemporal evolution law is generated, and the core representation of the event is obtained. For the core representation of an event, a mapping method is used to transform it into a dynamic vector to obtain a dynamic feature description. If there are inconsistencies in the data dimensions in the dynamic feature description, the dynamic vector is compressed through dimensionality reduction to unify the data structure and obtain the event dynamic feature vector.