Disaster risk assessment early warning method and system based on multi-source heterogeneous data

Abstract

The invention discloses a disaster risk assessment early warning method and system based on multi-source heterogeneous data, and relates to the technical field of data fusion and processing, and the method comprises the steps: obtaining a multi-source heterogeneous disaster risk data stream; constructing a normalized disaster feature matrix; generating a multi-source feature incidence matrix; generating a disaster risk index set according to the disaster risk prediction model; generating a regional disaster risk distribution thermodynamic diagram; constructing a disaster risk time sequence model; and generating a disaster research and judgment report, and performing visual early warning on the disaster research and judgment report through a three-dimensional simulation technology. The technical problem that traditional disaster risk assessment early warning depends on a single or few data sources, data are one-sided and lack of real-time performance, and consequently assessment early warning is inaccurate is solved, comprehensive real-time analysis based on multi-source heterogeneous data is achieved, the accuracy of disaster risk assessment and the timeliness of early warning are improved, and the risk assessment early warning efficiency is improved. And a reliable basis is provided for disaster emergency decision making.

Description

Technical Field

[0001] The present invention relates to the technical field of data fusion and processing, and in particular to a disaster risk assessment and early warning method and system based on multi-source heterogeneous data. Background Art

[0002] Traditional approaches to disaster risk assessment and early warning rely on a single or limited number of data sources, such as meteorological data or historical disaster records. These methods are simple and straightforward in data processing, but lack the ability to deeply mine and integrate complex data. While they can be effective in relatively stable scenarios with a single type of disaster, they are still relatively effective. However, with the acceleration of global climate change and urbanization, disasters are becoming increasingly complex and diverse. Traditional methods struggle to comprehensively capture disaster information and are unable to effectively integrate multiple data points when faced with complex disaster environments. This results in incomplete assessments and delayed early warnings, making it difficult to accurately assess disaster risks and provide timely early warnings. Summary of the Invention

[0003] This application solves the technical problem that traditional disaster risk assessment and early warning rely on a single or a small number of data sources, the data is one-sided and lacks real-time performance, which leads to inaccurate assessment and early warning. This application obtains multi-source heterogeneous disaster risk data streams such as towers, communications, population heat, waterlogging risk assessment and video, and constructs a normalized disaster feature matrix through noise filtering and feature extraction fusion. It uses the least squares method to fit the disaster development curve to mine feature relationships, establishes a disaster risk prediction model to generate a risk index set, and then generates a risk distribution heat map through a threshold segmentation algorithm. It combines the time series model to analyze dynamic changes, and finally generates a disaster assessment report and a visual early warning, thereby achieving a comprehensive and accurate assessment of disaster risks and efficient early warning, making disaster emergency decision-making more scientific and reasonable.

[0004] In response to the above technical problems, this application proposes a technical solution for a disaster risk assessment and early warning method and system based on multi-source heterogeneous data.

[0005] In the first aspect, the present application provides a disaster risk assessment and early warning method based on multi-source heterogeneous data, wherein the method includes: obtaining a multi-source heterogeneous disaster risk data stream, the data sources of the multi-source heterogeneous disaster risk data stream include tower, communication, population heat, urban flood risk assessment and video data; performing noise filtering and feature extraction fusion on the multi-source heterogeneous disaster risk data stream to construct a normalized disaster feature matrix; using the least squares method to fit the normalized disaster feature matrix to the disaster development curve to generate a multi-source feature association matrix; combining the multi-source feature association matrix to perform disaster feature association mining, establish a disaster risk prediction model, and generate a disaster risk index set based on the disaster risk prediction model; based on the disaster risk index set, using a threshold segmentation algorithm to perform risk division and rendering on the target area to generate a regional disaster risk distribution heat map; based on the regional disaster risk distribution heat map, constructing a disaster risk time series model; based on the disaster risk time series model, generating a disaster assessment report, and visualizing the disaster assessment report through three-dimensional simulation technology.

[0006] In the second aspect, the present application provides a disaster risk assessment and early warning system based on multi-source heterogeneous data, wherein the system includes: a data stream acquisition module for acquiring multi-source heterogeneous disaster risk data streams, the data sources of the multi-source heterogeneous disaster risk data streams including towers, communications, population heat, waterlogging risk assessment and video data; a matrix construction module for performing noise filtering and feature extraction fusion on the multi-source heterogeneous disaster risk data streams to construct a normalized disaster feature matrix; a matrix generation module for fitting the disaster development curve to the normalized disaster feature matrix using the least squares method to generate a multi-source feature correlation matrix; a set generation module for Combined with the multi-source feature association matrix, disaster feature association mining is carried out to establish a disaster risk prediction model, and a disaster risk index set is generated based on the disaster risk prediction model; a heat map generation module is used to use a threshold segmentation algorithm based on the disaster risk index set to perform risk division and rendering on the target area to generate a regional disaster risk distribution heat map; a model construction module is used to construct a disaster risk time series model based on the regional disaster risk distribution heat map; a report generation module is used to generate a disaster assessment report based on the disaster risk time series model, and to provide a visual warning for the disaster assessment report through three-dimensional simulation technology.

[0007] This application proposes one or more technical solutions, which have at least the following technical effects:

[0008] This application determines the object of data processing by obtaining multi-source heterogeneous disaster risk data streams. Then, the data stream is filtered for noise, and the filter is designed according to the data characteristics. The normalized disaster feature matrix is ​​constructed by using the method of correlation feature extraction and data fusion mapping. Then, the least squares method is used to fit the disaster development curve, and through matrix decomposition and association rule mining, a multi-source feature correlation matrix is ​​generated and a disaster risk prediction model is established to obtain a set of disaster risk indexes. After that, a threshold segmentation algorithm is used to generate a regional disaster risk distribution heat map, and a time series model is constructed to analyze its dynamic changes. A disaster assessment report is generated and a visual warning is provided to achieve accurate disaster risk assessment and warning, achieving a comprehensive real-time analysis based on multi-source heterogeneous data, improving the accuracy of disaster risk assessment and the timeliness of warning, and providing a reliable basis for disaster emergency decision-making.

[0009] The above content summarizes the present application's method and system for disaster risk assessment and early warning based on multi-source heterogeneous data. The present application will describe the steps of the technical solution in detail in the following specific implementation methods to facilitate technical personnel to have a clear and complete understanding of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 This is a flow chart of a disaster risk assessment and early warning method based on multi-source heterogeneous data provided in an embodiment of the present application.

[0012] Figure 2 This is a structural diagram of a disaster risk assessment and early warning system based on multi-source heterogeneous data provided in an embodiment of the present application.

[0013] Explanation of the accompanying drawings: data flow acquisition module 1, matrix construction module 2, matrix generation module 3, set generation module 4, heat map generation module 5, model construction module 6, report generation module 7. DETAILED DESCRIPTION

[0014] This application obtains multi-source heterogeneous disaster risk data streams, performs noise filtering and feature extraction and fusion, and constructs a normalized disaster feature matrix. The least squares method is used to fit the disaster development curve, generate a multi-source feature association matrix and mine disaster feature association rules, and establish a disaster risk prediction model to generate a risk index set. A threshold segmentation algorithm is used to generate a regional disaster risk distribution heat map, a time series model is constructed to analyze dynamic changes, a disaster assessment report is generated, and a visual warning is provided through three-dimensional simulation technology to achieve a comprehensive assessment and accurate warning of disaster risks, providing strong support for emergency decision-making. It achieves the technical effect of conducting comprehensive real-time analysis based on multi-source heterogeneous data, improving the accuracy of disaster risk assessment and the timeliness of warning, and providing a reliable basis for disaster emergency decision-making.

[0015] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0016] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices.

[0017] Example 1, as Figure 1 As shown, a disaster risk assessment and early warning method based on multi-source heterogeneous data, wherein the method includes:

[0018] Step A100: Acquire a multi-source heterogeneous disaster risk data stream, where the data sources of the multi-source heterogeneous disaster risk data stream include tower, communication, population thermal, waterlogging risk assessment and video data.

[0019] In the embodiment of the present application, waterlogging risk assessment is to evaluate the possibility and degree of waterlogging in different areas by collecting and analyzing various information such as topography, drainage system conditions, historical rainfall data, etc., to obtain waterlogging risk assessment data.

[0020] Specifically, first, data is obtained from multiple channels. Tower data is widely distributed across different regions, and information such as their height, material, and surrounding environment can reflect the region's topography and infrastructure. For example, in mountainous areas, tower stability is correlated with the risk of landslides. Tower data can be obtained through various sensors installed on the towers. Displacement sensors and stress sensors are installed at key structural locations of the towers, such as the base and tower body connections. Displacement sensors can monitor real-time displacement changes, such as tilt and sway, under different environmental conditions, with millimeter-level accuracy. Stress sensors can measure the stress experienced by various parts of the tower, collecting data in megapascals (MPa). Furthermore, temperature sensors are used to collect real-time temperature data of the tower's environment, as temperature changes can affect the performance of the tower's materials and structural stability.

[0021] Communication data includes base station signal strength and user traffic. Abnormal signal weakening or traffic surges in a certain area may indicate a disaster has caused a large gathering of people or damaged communication facilities. For example, during a flood disaster, affected people use communication devices to seek help, causing fluctuations in communication traffic in the area. Communication data can be obtained through the base station systems of local telecommunications operators.

[0022] Population thermal data provides a visual representation of population density. Disasters in densely populated areas, such as city centers and residential areas, can result in greater casualties. By analyzing trends in this data, we can estimate the impact of disasters on populations in different regions. Population thermal data is primarily obtained using mobile phone signaling data and Wi-Fi probe technology. Mobile phone signaling data is the information generated by interactions between mobile phones and base stations. By analyzing this signaling, the approximate location and movement trajectory of mobile phone users can be determined. Telecommunications operators provide desensitized mobile phone signaling data to technical personnel, enabling them to calculate population density distribution within different areas. Furthermore, Wi-Fi probe devices are installed in public and commercial areas. These devices detect the number of Wi-Fi-enabled mobile devices in the vicinity. This, combined with signal strength and time information from these devices, further supplements and optimizes population thermal data.

[0023] Waterlogging risk assessment data focuses on flood disaster risks. Using topographic mapping and drainage system data, it calculates the waterlogging risk level of different regions, providing a key basis for flood disaster warnings. The process of obtaining waterlogging risk assessment data is as follows:

[0024] Step a: First, collect topographic data. High-precision satellite remote sensing imagery can accurately identify subtle ground undulations with sub-meter resolution, capturing key terrain information such as subtle slope changes and low-lying areas. Furthermore, aerial LiDAR measurement technology can rapidly acquire three-dimensional terrain data over large areas, generating high-precision digital elevation models (DEMs) that provide the foundational topographic framework for subsequent water flow simulations.

[0025] Step b: Simultaneously, collect data on the urban drainage system, including information on the layout of the drainage network, pipe diameters, and drainage capacity. This data is typically provided by urban planning or drainage management departments. Additionally, historical rainfall data is required, including data on rainfall intensity, duration, and total rainfall over several years from meteorological authorities.

[0026] Step c: Next, construct the SWMM (Stormwater Flood Management Model) and convert and preprocess the collected topographic data so that it can be recognized and processed by the model. Divide the digital elevation model into appropriate computational grids and determine the topographic parameters of each grid. Then, based on the drainage system data, accurately construct the drainage network system in the model and set the hydraulic parameters of the pipeline, such as roughness. For the pump station, set the control conditions according to its actual operating rules. Next, input the historical rainfall data into the model in a time series to simulate the water flow process under different rainfall scenarios.

[0027] Step d: In order to improve the accuracy of the model, parameter calibration and verification are required. By comparing and analyzing actual waterlogging events, key parameters in the model, such as surface infiltration rate and runoff coefficient, are adjusted to make the model simulation results as consistent as possible with the actual situation. After multiple calibrations and verifications, the waterlogging risk assessment model can more reliably calculate the degree of waterlogging risk in different regions, and output detailed risk levels (such as very low, low, medium, high, and very high) or risk probability (accurate to specific percentages, such as 10%, 30%, etc.) data, that is, waterlogging risk assessment data for different regions, providing accurate waterlogging risk information for disaster risk assessment.

[0028] Video data is captured by surveillance cameras distributed across various areas. These cameras, including traffic monitoring cameras and security cameras, are located on roads, in communities, and around important public facilities. The cameras continuously capture video footage at a certain frame rate (e.g., 25 or 30 frames per second) and transmit the video data to a data storage server via a wired network (e.g., fiber optic) or a wireless network (e.g., 4G or 5G). To ensure efficient data utilization, the video data requires real-time encoding and compression, using encoding standards such as H.264 and H.265 to reduce data transmission volume and storage space usage.

[0029] By uploading the above-mentioned multi-source heterogeneous data to the data storage server and integrating them, a comprehensive disaster risk data system is built, laying a solid foundation for subsequent accurate risk assessment and early warning work.

[0030] Step A200: performing noise filtering and feature extraction fusion on the multi-source heterogeneous disaster risk data stream to construct a normalized disaster feature matrix.

[0031] Optionally, a multi-source heterogeneous data filter is first designed based on the data characteristic information of the multi-source heterogeneous disaster risk data stream, and then the filter is used to filter the noise of the multi-source heterogeneous disaster risk data stream to obtain usable data. Finally, the associated features are extracted and data fusion mapped on these usable data to obtain a normalized disaster feature matrix. The specific steps are described in detail in A210-A230.

[0032] Step A300: Using the least squares method to fit the normalized disaster feature matrix to a disaster development curve, and generating a multi-source feature correlation matrix.

[0033] In the embodiments of this application, the disaster development curve is a curve generated by fitting the model fitting parameters obtained by solving the normalized disaster characteristic matrix using the least squares method. The multi-source characteristic correlation matrix is ​​a matrix formed by arranging the numerical values ​​reflecting the degree of correlation between different disaster characteristics according to certain rules.

[0034] In one embodiment of the present application, the normalized disaster feature matrix is ​​first fitted and solved using the least squares method to obtain the model fitting parameters and then generate a disaster development curve. The disaster feature correlation is then calculated based on the curve to form a disaster feature correlation set. Finally, a multi-source feature correlation matrix is ​​generated based on this set. The specific steps are described in detail in A310-A330.

[0035] Step A400: performing disaster feature association mining in combination with the multi-source feature association matrix, establishing a disaster risk prediction model, and generating a disaster risk index set based on the disaster risk prediction model.

[0036] In the embodiment of the present application, disaster feature association mining is based on a multi-source feature association matrix. By performing matrix decomposition and association rule mining on the matrix, valuable information is extracted from the complex multi-source data feature relationship to obtain a disaster feature association rule data set.

[0037] Specifically, when establishing a disaster risk prediction model, the multi-source feature association matrix is ​​first subjected to matrix decomposition and association rule mining to obtain a disaster feature association rule data set. Then, based on this set, a tower stability prediction model, a casualty estimation model, and a disaster level assessment model are constructed respectively. Finally, these three models are combined to establish a disaster risk prediction model. The specific steps are described in detail in A410-A430.

[0038] Based on the aforementioned disaster risk prediction model, real-time multi-source data is input and, through model calculations and analysis, a set of disaster risk indices is generated. These indices quantitatively reflect the degree of disaster risk in different regions and types, for example, using a numerical range of 0-10 to represent the risk index, with higher values ​​indicating greater risk. This set of disaster risk indices provides a scientific and accurate basis for subsequent disaster warnings and emergency decision-making, helping to improve the efficiency and effectiveness of disaster response and reduce casualties and property losses.

[0039] Generating a set of disaster risk indices through the above steps provides a scientific and accurate basis for subsequent disaster warnings and emergency decision-making, helping to improve the efficiency and effectiveness of disaster response and reduce casualties and property losses.

[0040] Step A500: Based on the disaster risk index set, a threshold segmentation algorithm is used to perform risk division and rendering on the target area to generate a regional disaster risk distribution heat map.

[0041] In this embodiment, the target area refers to a specific area that requires disaster risk assessment before a disaster occurs. Risk segmentation rendering is the process of performing threshold segmentation on a set of disaster risk indices to obtain a set of disaster risk levels, mapping this set onto a spatial map of the target area, and visually rendering it according to the different levels.

[0042] Specifically, a threshold segmentation algorithm is first used to divide the disaster risk index set according to distribution requirements, and the risk division threshold rule is determined. Then, threshold segmentation is performed on the disaster risk index set based on the rule to obtain a disaster risk level set. Finally, this set is mapped to the spatial map of the target area and graded, thereby generating a regional disaster risk distribution heat map. The specific steps are described in detail in A510-A530.

[0043] Step A600: Constructing a disaster risk time series model based on the regional disaster risk distribution heat map.

[0044] Specifically, when constructing a disaster risk time series model, we first analyze the dynamic changes in the regional disaster risk distribution heat map to identify the trend data of disaster risk characteristics, and then use the time series model to fit, train, verify and optimize these data to construct a disaster risk time series model. The specific steps are detailed in A610-A620.

[0045] Step A700: Generate a disaster assessment report based on the disaster risk time series model, and perform a visual early warning on the disaster assessment report through three-dimensional simulation technology.

[0046] Specifically, the disaster risk time series model continuously monitors and analyzes regional disaster risk distribution heat maps to capture changing trends in disaster risk characteristics. For example, during a flood, the model can track in real time the expansion of the affected area, changes in flood depth, and fluctuations in the number of people affected. This data not only reflects the current state of the disaster but also includes information on its trends over time.

[0047] Next, using this dynamic data, technical personnel, combined with historical disaster data and relevant domain knowledge, generate a disaster assessment report. This report covers several key aspects, such as accurately assessing the type, scale, and stage of the disaster. For example, in the case of earthquakes, based on earthquake monitoring data and model analysis, the report can clearly define the magnitude, epicenter, and potential impact area, as well as the likelihood and intensity of aftershocks over time. The report also assesses the impact of the disaster on infrastructure, human safety, and the socioeconomic situation. For example, it analyzes the number of buildings that may collapse, traffic disruptions, estimated casualties, and preliminary estimates of economic losses. The acquisition and analysis of this data relies on the integration of multiple data sources. For example, communication data can help understand communication disruptions in the affected area, population and thermal data can help estimate the size of the affected population, and tower data can help determine the risk of damage to infrastructure such as communication facilities and power towers.

[0048] Then, using 3D simulation modeling software, a visual warning is provided for the disaster assessment report, presenting the various data in the report in an intuitive 3D scene. Taking an urban fire as an example, the 3D simulation scene can accurately display the specific location of the fire, using different colors and brightness to indicate the intensity of the fire, and simulate the spread and direction of the smoke. At the same time, the scene can also mark the affected buildings, roads, and areas where people gather. Through the dynamic simulation of time series data, the audience can clearly see the development of the fire, such as how the fire spreads and how the affected area expands. To enhance the warning effect, various warning signs and prompts can be added, such as flashing red lights to indicate danger zones and voice prompts to guide people in evacuation directions.

[0049] Through the above steps, visual early warning can enable emergency commanders and the general public to quickly understand the severity and development trend of the disaster, so as to take timely and effective response measures and reduce the losses caused by the disaster.

[0050] Furthermore, step A200 in the method provided in the embodiment of the present application includes:

[0051] A210: Design a multi-source heterogeneous data filter based on the data characteristic information of the multi-source heterogeneous disaster risk data stream.

[0052] A220: Use the multi-source heterogeneous data filter to perform noise filtering on the multi-source heterogeneous disaster risk data stream to obtain a usable multi-source heterogeneous disaster risk data stream.

[0053] A230: Perform correlation feature extraction and data fusion mapping on the available multi-source heterogeneous disaster risk data stream to obtain a normalized disaster feature matrix.

[0054] In the embodiment of the present application, the multi-source heterogeneous data filter is designed based on the data characteristic information of the multi-source heterogeneous disaster risk data stream, and its function is to filter the noise of the multi-source heterogeneous disaster risk data stream.

[0055] Specifically, first, a filter is designed for multi-source heterogeneous data. Multi-source heterogeneous disaster risk data streams include tower, communications, population and thermal, flood risk assessment, and video data, each with different formats, frequencies, and noise characteristics. For example, tower data may contain outliers due to sensor failure, and communications data may fluctuate due to signal interference. To remove this noise, those skilled in the art need to deeply analyze the data characteristics. For example, by studying the historical variation patterns of tower data, the fluctuation range of normal data can be determined, and this can be used as a basis for designing a filter that can identify and filter outliers. For communications data, targeted filtering algorithms are designed based on signal transmission principles and common interference patterns. These include Fourier transform-based filtering algorithms for filtering high-frequency noise caused by electromagnetic interference; adaptive filtering algorithms for real-time tracking of signal and interference changes in complex and variable interference environments; and wavelet transform filtering algorithms for filtering bursts in communications data by decomposing the signal into different scales. These denoising steps lay the foundation for subsequent acquisition of high-quality data.

[0056] Next, a designed multi-source heterogeneous data filter is used to filter noise from the multi-source heterogeneous disaster risk data stream. For example, in communications data, if there are signal spikes due to electromagnetic interference, the filter, based on pre-defined rules, can detect these abnormal spikes and correct or remove them, thus obtaining a usable multi-source heterogeneous disaster risk data stream.

[0057] Finally, correlation feature extraction and data fusion mapping are performed on the available multi-source heterogeneous disaster risk data streams to obtain the normalized disaster feature matrix. The specific steps are described in detail in A231-A233.

[0058] By designing filters to filter noise based on data characteristics, and then extracting related features and performing data fusion mapping on the filtered data, the technical effect of constructing a unified, standardized normalized disaster feature matrix that can reflect the characteristic relationships of multi-source data is achieved, providing an effective data basis for subsequent disaster risk assessment and early warning.

[0059] Furthermore, step A230 in the method provided in the embodiment of the present application includes:

[0060] A231: Extract correlation features from the available multi-source heterogeneous disaster risk data stream to obtain a multi-source heterogeneous disaster correlation feature set.

[0061] A232: Mapping the multi-source heterogeneous disaster correlation feature set to the same spatiotemporal scale for time synchronization and spatial alignment to obtain a standard multi-source heterogeneous disaster correlation feature set.

[0062] A233: Perform normalization fusion processing on the standard multi-source heterogeneous disaster correlation feature set to obtain the normalized disaster feature matrix.

[0063] In the embodiment of the present application, time synchronization and spatial alignment are to unify the time information of different data sources to the same time scale and convert the spatial coordinates of various types of data into the same coordinate system to achieve spatial unification.

[0064] Optionally, after obtaining available multi-source heterogeneous disaster risk data streams through the above steps, correlation features are first extracted from the available multi-source heterogeneous disaster risk data streams. Different types of data contain various disaster-related information. For example, tower tilt data may be related to earthquakes or strong wind disasters, while population thermal data can reflect the distribution of people in the affected area.

[0065] Data mining is used to extract relevant features from this data. For example, a clustering algorithm is used to analyze tower data and surrounding geological and meteorological data. Data is collected and cleaned to remove outliers and missing values. Different types of data are standardized to make them comparable. The K-Means clustering algorithm is selected to determine the appropriate number of clusters, K. This parameter can be selected using the elbow method (calculating the error of the clustering model for different K values, plotting the error versus K value, and finding the appropriate number of clusters at the inflection point of the curve). The preprocessed data is input into the selected clustering algorithm, which then divides the data points into clusters based on their characteristics. With each iteration, the algorithm continuously adjusts the cluster centers until convergence is achieved (e.g., the cluster center remains constant or changes very little). After clustering is complete, each cluster is analyzed to examine the distribution of tower stability within the cluster and the corresponding geological and meteorological conditions. For example, it was found that the stability of the towers in a certain cluster was poor under specific geological structures (such as near faults) and meteorological conditions (such as strong winds and heavy rains), and then the correlation characteristics between the stability of the towers and the possibility of disasters under specific geological and meteorological conditions were established.

[0066] An association rule mining algorithm is used to analyze the relationship between signal strength changes in communication data and the extent of disaster-affected areas. Communication data, including information such as signal strength and corresponding geographic location, is first collected, along with data related to the extent of the disaster-affected areas. The data is then organized and converted into a format suitable for association rule mining, such as a transaction dataset. Next, the Apriori association rule mining algorithm is applied, with a support threshold set. The dataset is scanned to identify all frequent itemsets that meet the support threshold. For example, it is found that signal strength changes within a specific range in certain areas frequently coincide with the extent of specific disaster-affected areas. A confidence threshold is then set based on these frequent itemsets to generate association rules that meet the required confidence level. For example, when signal strength drops by more than a certain percentage, a high confidence level indicates that the extent of the disaster-affected area will expand. This establishes a relationship between signal strength changes in the communication data and the extent of the disaster-affected areas, ultimately yielding a multi-source, heterogeneous disaster association feature set.

[0067] Afterwards, the multi-source heterogeneous disaster correlation feature set is mapped to the same spatiotemporal scale for temporal synchronization and spatial alignment. This is because different data sources may have different data collection times and spatial coordinates. For example, the temporal accuracy of video data acquisition may differ from that of communication data, and the spatial resolution of tower location information and population thermal data may also differ. To enable collaborative analysis of these data, temporal synchronization and spatial alignment are required. For example, through timestamp matching and interpolation algorithms, the temporal information of different data sources can be unified to the same time scale. Utilizing Geographic Information System (GIS) technology, the spatial coordinates of various data types can be converted to the same coordinate system to achieve spatial alignment, thereby obtaining a standard multi-source heterogeneous disaster correlation feature set.

[0068] Finally, the standard multi-source heterogeneous disaster-related feature set was normalized and fused to obtain a normalized disaster feature matrix. Because different features have different numerical ranges and dimensions—for example, population thermal data is expressed in terms of number of people, while waterlogging risk assessment data may be expressed in terms of risk level—direct fusion can result in certain features being overweighted or underweighted in subsequent analysis.

[0069] Step e: Use min-max normalization to map all feature values ​​to the same interval, eliminating dimensionality effects. When processing the data, for each feature, first find the minimum and maximum values ​​in the dataset. To calculate this, subtract the minimum value of the feature from the value of each data point, then divide by the difference between the maximum and minimum values. This calculation maps all data for that feature to a specific interval, typically [0, 1]. By performing this operation for every feature in the dataset, all feature values ​​are mapped to the same interval, ensuring that different features have equal weight in subsequent analysis.

[0070] Step f: Using a data fusion algorithm, these normalized features are combined into a matrix according to specific rules, ultimately yielding a normalized disaster feature matrix. The data fusion algorithm combines the normalized features according to pre-defined rules. These rules may be based on various factors, such as the correlation between features, where highly correlated features may be more closely linked when combined; and their importance to the disaster risk assessment task, where more important features may be assigned higher weights. Following these rules, the normalized features are combined, such as by performing a weighted summation based on their weights, to form a new combined feature vector. Finally, these combined feature vectors are arranged in a specific order to form a matrix, the final normalized disaster feature matrix.

[0071] Through the above steps, the normalized disaster characteristic matrix integrates the key information of multi-source data, providing a unified and standardized data basis for subsequent disaster risk assessment and early warning.

[0072] Furthermore, step A300 in the method provided in the embodiment of the present application includes:

[0073] A310: Using the least squares method to fit and solve the normalized disaster characteristic matrix, obtain model fitting parameters, and generate a disaster development curve based on the model fitting parameters.

[0074] A320: Calculate the disaster feature correlation based on the disaster development curve to obtain a disaster feature correlation set.

[0075] A330: Generate the multi-source feature correlation matrix according to the disaster feature correlation degree set.

[0076] In the embodiment of the present application, the model fitting parameters reflect the relative importance and mutual relationship of different disaster characteristics in the model. The disaster development curve presents the development trend of the disaster in time or space in a visual way.

[0077] Specifically, the least squares method is first used to fit and solve the normalized disaster feature matrix. Taking an earthquake disaster scenario as an example, the features in the matrix may include vibration data from seismic monitoring stations, population thermal change data in the affected area, and signal fluctuation data from communication base stations. The core principle of the least squares method is to find the optimal function matching the data by minimizing the sum of squared errors. In this process, the data in the normalized disaster feature matrix are treated as observations. A linear model is assumed to describe the relationship between these data. The model parameters are continuously adjusted to minimize the sum of squared errors between the model predictions and the actual observations. After fitting and solving, the model fitting parameters are obtained. These parameters reflect the relative importance and interrelationships of different disaster features in the model. For example, in the earthquake scenario mentioned above, the model fitting parameters may indicate the degree of correlation between vibration data and population thermal change data, as well as the weight of their impact on the overall disaster situation. Based on the obtained model fitting parameters, a disaster development curve can be generated. This curve visualizes the development trend of the disaster in time or space. For example, the curve can intuitively show the expansion of the earthquake-affected area and the changes in the affected population over time.

[0078] Next, the disaster characteristic correlation is calculated based on the generated disaster development curve. The disaster characteristic correlation is used to measure the degree of correlation between different disaster characteristics. Taking fire disasters as an example, multiple characteristics are involved, such as the location of the fire source, the speed of fire spread, the fire resistance level of surrounding buildings, and the evacuation of the population. When calculating the correlation, methods such as the Pearson correlation coefficient are used. For each pair of disaster characteristics, such as the location of the fire source and the speed of fire spread, the data at the corresponding time points on the disaster development curve are extracted and the Pearson correlation coefficient is calculated. If the coefficient is close to 1, it indicates a strong positive correlation between the two characteristics, that is, the change in the location of the fire source is highly consistent with the change trend of the fire spread rate; if the coefficient is close to -1, it indicates a strong negative correlation; and if the coefficient is close to 0, it indicates a weak correlation. By performing this calculation for all possible pairs of disaster characteristics, a set of disaster characteristic correlations is obtained. This set comprehensively records the degree of correlation between different disaster characteristics.

[0079] Finally, a multi-source feature correlation matrix is ​​generated based on the set of disaster feature correlations. When generating the matrix, each disaster feature is treated as a row and a column of the matrix. The elements in the matrix are the correlation values ​​between the disaster features represented by the corresponding row and column. For example, the element in the i-th row and j-th column of the matrix represents the correlation between the i-th disaster feature and the j-th disaster feature. In this way, the set of disaster feature correlations is presented in the form of a matrix, forming a multi-source feature correlation matrix. For example, in flood disaster assessment, the multi-source feature correlation matrix can quickly understand the correlation between features such as water level changes, rainfall, drainage system conditions, and population distribution in the affected area, providing strong support for accurately assessing flood disaster risks.

[0080] Through the above steps, the complex correlation between different disaster characteristics in multi-source data can be clearly demonstrated, providing an important data foundation for subsequent disaster characteristic correlation mining and the establishment of disaster risk prediction models.

[0081] Furthermore, step A400 in the method provided in the embodiment of the present application includes:

[0082] A410: Perform matrix decomposition and association rule mining on the multi-source feature association matrix to obtain a disaster feature association rule data set.

[0083] A420: Based on the disaster characteristic association rule data set, a tower stability prediction model, a casualty estimation model, and a disaster level assessment model are constructed respectively.

[0084] A430: Establish the disaster risk prediction model by combining the tower stability prediction model, the casualty estimation model and the disaster level assessment model.

[0085] Specifically, the multi-source feature correlation matrix is ​​first decomposed. This matrix contains the characteristic correlation information of multiple sources such as towers, communications, population, and heat. Matrix decomposition techniques (such as principal component analysis and singular value decomposition) can reduce the dimensionality of complex high-dimensional matrices, extract core feature components, and reveal key correlation patterns hidden in the data. For example, by decomposing the correlation matrix between tower stability data and geological disaster data, potential correlations between tower tilt angles, soil moisture content, and earthquake magnitude can be discovered.

[0086] Based on matrix decomposition, an association rule mining algorithm (such as the Apriori algorithm, similar to step A231) is used to analyze the matrix. Rules such as "When the tower vibration frequency exceeds a threshold and the communication signal strength drops sharply, the probability of heavy rainfall causing flooding in the area increases by 80%" are discovered, forming a data set of disaster feature association rules. These rules clearly define the mapping between different feature combinations and disaster types and levels.

[0087] Next, we construct a tower stability prediction model, a casualty estimation model, and a disaster level assessment model:

[0088] Step g: A tower stability prediction model is constructed based on a dataset of disaster feature association rules, integrating tower structural parameters (such as height and material strength), environmental data (such as wind speed and rainfall), and historical fault records. The model is then constructed using a random forest algorithm. The integrated multi-source data undergoes data cleaning, preprocessing, and feature engineering to form a training dataset containing stability labels (such as stable and unstable). Random sampling techniques are then used to select multiple bootstrap samples from the training data. At each node split, some features are randomly selected for optimal splitting, constructing multiple uncorrelated decision trees. The final prediction is then generated by integrating the prediction results of multiple decision trees (e.g., using a majority voting method). The model is trained using historical fault record data, optimizing hyperparameters such as the number and depth of trees. The resulting model can input real-time environmental parameters and tower structural data, output tower stability probabilities, and predict tower stability in disaster scenarios.

[0089] Step h: Developing a casualty estimation model. This model integrates population thermal data, building distribution (e.g., residential density, building seismic rating), and disaster intensity data to form a training dataset containing casualty data from historical disaster scenarios. A statistical regression model is then used to construct the casualty estimation model. The data is then preprocessed, including cleaning and normalization, to identify features highly correlated with casualties (e.g., areas with high residential density and low building seismic ratings are associated with a higher risk of casualties in strong earthquakes). An appropriate regression model (e.g., linear regression, logistic regression, or generalized additive model) is selected based on the data characteristics. The model is then trained using historical data, and model parameters are fitted by minimizing prediction error (e.g., by maximizing the mean squared error or likelihood function). This establishes a quantitative relationship between disaster intensity, building characteristics, population distribution, and the number / probability of casualties. Finally, cross-validation and other methods are used to evaluate model accuracy and optimize model parameters. The resulting model can be fed with real-time disaster intensity and regional building and population data, and outputs casualty estimates or risk levels, providing data support for the deployment of emergency rescue resources.

[0090] Step i: Disaster level assessment model. This model is constructed using the Analytic Hierarchy Process (AHP) by integrating waterlogging risk assessment data (e.g., waterlogging depth and drainage capacity), meteorological data (e.g., rainfall and wind speed), and historical disaster level classification standards. First, the target layer is defined as disaster level assessment. A hierarchical structure is constructed, consisting of a criterion layer (e.g., waterlogging risk assessment data and meteorological data) and an indicator layer (e.g., waterlogging depth, drainage capacity, rainfall, wind speed, and other specific parameters). Next, the relative importance of each indicator layer is quantified through expert scoring or data statistics. A judgment matrix is ​​constructed and the weights of each indicator are calculated. A consistency check is also performed to ensure logical rationality. Based on historical disaster level classification standards, waterlogging risk, meteorological data, and other indicators are substituted into the hierarchical model, and a comprehensive score is calculated through weighted summation. Finally, the disaster level (e.g., mild, moderate, severe) is determined based on the score range. This creates a disaster level assessment model that comprehensively reflects multi-source data and provides a quantitative basis for disaster risk classification.

[0091] Finally, the output results of the above three sub-models are combined, and a weighted fusion method is used to establish a disaster risk prediction model. The weight of each sub-model is determined by training using historical disaster data. For example, the tower stability prediction result accounts for 30% of the weight, the casualty estimate accounts for 40% of the weight, and the disaster level assessment accounts for 30% of the weight. The disaster risk index is obtained by comprehensive calculation.

[0092] Through the above steps, the disaster risk prediction model can output multi-dimensional risk assessment results, including infrastructure stability, personnel safety threats and disaster impact levels, providing core algorithm support for generating a set of disaster risk indexes and realizing comprehensive prediction and early warning of disaster risks.

[0093] Furthermore, step A500 in the method provided in the embodiment of the present application includes:

[0094] A510: Using a threshold segmentation algorithm to divide the disaster risk index set into distribution requirements, and determining a risk division threshold rule.

[0095] A520: Perform threshold segmentation on the disaster risk index set based on the risk classification threshold rule to obtain a disaster risk level set.

[0096] A530: Mapping the disaster risk level set onto the spatial map of the target area and performing level rendering to generate a heat map of the regional disaster risk distribution.

[0097] In the embodiments of this application, threshold segmentation is the process of processing a set of disaster risk indices and classifying them into different levels to obtain a set of disaster risk levels. Level rendering is the process of mapping the set of disaster risk levels onto a spatial map of the target area, visually displaying the different risk levels through different visual effects such as color and depth.

[0098] Specifically, first, an in-depth analysis is conducted on the disaster risk index set. The disaster risk index set is obtained through data processing and model prediction in step A400. Using a threshold segmentation algorithm based on statistics, a large amount of historical disaster risk index data is statistically analyzed to calculate the mean, standard deviation and other statistical quantities of the data. Assuming that the value range of the disaster risk index is 0-100, it is found through statistics that when the risk index exceeds 70, the probability of a serious disaster occurring in the area increases significantly, while when the risk index is between 40-70, the disaster risk is at a medium level, and below 40 is a low risk. Based on such analysis results, the risk classification threshold rule can be determined as follows: a risk index less than 40 is low risk, 40-70 is medium risk, and greater than 70 is high risk. Of course, in actual situations, more factors need to be considered to determine the threshold rules, such as geographical environment differences and infrastructure conditions in different regions, and it may also need to be adjusted in combination with the experience of technical personnel in this field.

[0099] Next, after determining the risk classification threshold rule, threshold segmentation is performed on the disaster risk index set based on this rule. Each index value in the disaster risk index set is compared with the threshold value. Using the aforementioned flooding disaster example, suppose a region has a calculated disaster risk index of 85. Based on the previously determined threshold rule, it would be classified as high risk. If another region has a risk index of 35, it would be classified as low risk. By performing this judgment and classification on each value in the entire disaster risk index set, a set of disaster risk levels can be obtained.

[0100] After obtaining the set of disaster risk levels, they are mapped onto a spatial map of the target area and rendered to generate a regional disaster risk distribution heat map. The spatial map can be an electronic map generated using Geographic Information System (GIS) technology, which contains detailed geographic information of the target area, such as streets, buildings, and topography.

[0101] Finally, using the spatial analysis and visualization capabilities of GIS technology, each risk level in the disaster risk level set is associated with the corresponding area on the map. For example, on the map, find an area with a high risk level and mark it in red; areas with a medium risk level are marked in yellow; and low risk areas are marked in green. Different colors can intuitively reflect the difference in risk level. At the same time, the color depth can be set to further distinguish subtle differences in risk levels, such as dark red indicating extremely high risk and light red indicating relatively low high risk. Through such mapping and rendering operations, a regional disaster risk distribution heat map can be generated.

[0102] Through the above steps, relevant personnel can Figure 1 The disaster risk distribution in different places within the target area can be clearly seen, providing intuitive and powerful data support for emergency decision-making, resource allocation, etc.

[0103] Furthermore, step A600 in the method provided in the embodiment of the present application includes:

[0104] A610: Analyze the dynamic changes in the heat map of the disaster risk distribution in the region to identify and obtain the trend data on the changes in the disaster risk characteristics.

[0105] A620: Use a time series model to perform fitting training and verification optimization on the disaster risk characteristic change trend data to construct the disaster risk time series model.

[0106] In one embodiment, first, the heat map is analyzed to obtain the trend data of disaster risk characteristics. Taking urban flood disasters as an example, different colors on the heat map represent different flood risk levels. The change in color and the expansion or contraction of the coverage area reflect the dynamic change of flood risk. Through image recognition and data analysis tools, the risk level information of each area in the heat map at different time points can be extracted. For example, the heat map is sampled at regular time intervals (such as every hour or every day) to record the change in risk level of each area. The risk level data of these time series are sorted and analyzed to calculate indicators such as the rate of change and fluctuation of risk level.

[0107] This process can also be combined with other relevant data for comprehensive analysis. For example, heat map data can be correlated with rainfall data and river water level data for the same period to identify potential relationships between these factors and changes in disaster risk levels. Through such analysis, it is possible to identify and obtain data on the changing trends of disaster risk characteristics. This data contains information on the temporal changes in disaster risk and its correlation with other factors.

[0108] Next, we constructed and trained a time series model. After obtaining the disaster risk characteristic change trend data, we selected the autoregressive moving average model (ARMA) to fit the data:

[0109] Step j: Examine the data on changing trends in disaster risk characteristics and address any missing or outliers. Missing values ​​can be filled using interpolation methods (such as linear or spline interpolation). Outliers can be identified and corrected based on the statistical characteristics of the data (such as mean and standard deviation). The data is then tested for stationarity, as the ARIMA model requires stationary data. If the data is not stationary, it can be stabilized using differencing (such as first-order or second-order differencing).

[0110] Step k: Determine the ARIMA model parameters p (autoregressive order), d (difference order), and q (moving average order). The parameter range can be preliminarily determined by observing the data's autocorrelation function (ACF) and partial autocorrelation function (PACF). For example, the tailing of the ACF and the truncation of the PACF can help determine the values ​​of p and q. Information criteria (such as AIC and BIC) can also be used to select the optimal parameter combination, ensuring that the model fits the data while avoiding overfitting.

[0111] Step 1: Construct an ARIMA model based on the determined parameters and train the model using the preprocessed data. During training, the model continuously adjusts its parameters based on the input data to minimize the error between the predicted and actual values. For example, methods such as maximum likelihood estimation are used to estimate the model parameters, ensuring that the model best fits the changing trend data of disaster risk characteristics.

[0112] Then, after training is completed, the time series model needs to be verified and optimized:

[0113] Step m: Divide the disaster risk characteristic change trend data into a training set and a test set. (The input data are all disaster risk characteristic change trend data extracted from the regional disaster risk distribution heat map; the output data are all target values ​​used to train the time series model, i.e., the actual disaster risk level or risk degree quantified value corresponding to the input data time point.) Use the training set to train the model, and then use the test set to verify the model's predictive performance. Model accuracy can be measured using a variety of evaluation metrics, such as mean squared error (MSE), root mean square error (RMSE), and mean absolute error (MAE). These metrics quantify the degree of difference between the model's predicted values ​​and the actual values.

[0114] Step n: Optimize the model based on the validation results. If the model's prediction error is large, adjust the model parameters and retrain and validate it. You can also consider adding other relevant variables (such as meteorological data and geographic information) to enrich the model input and improve its predictive capabilities. You can also try different time series models, compare their performance, and select the optimal model as the disaster risk time series model.

[0115] Through the above steps, a time series model that can accurately reflect the dynamic changes in disaster risks is finally constructed. This model can predict future trends in disaster risk changes based on historical data, providing important support for disaster warning and emergency decision-making.

[0116] In summary, the disaster risk assessment and early warning method based on multi-source heterogeneous data provided by the embodiments of the present application has the following technical effects:

[0117] This application integrates multi-source data streams such as towers, communications, population heat, waterlogging risk assessment, and video, processes the data using a noise filtering algorithm, and constructs a normalized disaster feature matrix and a multi-source feature correlation matrix through operations such as feature extraction and fusion, and least squares fitting. A disaster risk index set is generated by mining the association of disaster features and establishing a risk prediction model. Risks are divided based on a threshold segmentation algorithm, heat maps are rendered, and dynamic changes are analyzed using a time series model. Disaster assessment reports are generated and visual warnings are provided. This achieves the technical effect of conducting comprehensive real-time analysis based on multi-source heterogeneous data, improving the accuracy of disaster risk assessment and the timeliness of warnings, and providing a reliable basis for disaster emergency decision-making.

[0118] Example 2, as Figure 2 As shown, based on the same inventive concept as the aforementioned embodiment 1, this embodiment of the present application provides a disaster risk assessment and early warning system based on multi-source heterogeneous data, the system comprising:

[0119] The data stream acquisition module 1 is used to obtain multi-source heterogeneous disaster risk data streams, the data sources of which include tower, communication, population thermal, waterlogging risk assessment and video data.

[0120] The matrix construction module 2 is used to perform noise filtering and feature extraction fusion on the multi-source heterogeneous disaster risk data stream to construct a normalized disaster feature matrix.

[0121] The matrix generation module 3 is used to fit the disaster development curve to the normalized disaster feature matrix using the least square method to generate a multi-source feature correlation matrix.

[0122] The set generation module 4 is used to perform disaster feature association mining in combination with the multi-source feature association matrix, establish a disaster risk prediction model, and generate a disaster risk index set based on the disaster risk prediction model.

[0123] The heat map generation module 5 is used to use a threshold segmentation algorithm based on the disaster risk index set to perform risk division and rendering on the target area to generate a regional disaster risk distribution heat map.

[0124] The model building module 6 is used to build a disaster risk time series model based on the regional disaster risk distribution heat map.

[0125] The report generation module 7 is used to generate a disaster assessment report based on the disaster risk time series model, and to provide a visual warning for the disaster assessment report through three-dimensional simulation technology.

[0126] Furthermore, the matrix construction module 2 is used to perform the following steps:

[0127] According to the data characteristic information of the multi-source heterogeneous disaster risk data stream, a multi-source heterogeneous data filter is designed; the multi-source heterogeneous data filter is used to filter the noise of the multi-source heterogeneous disaster risk data stream to obtain a usable multi-source heterogeneous disaster risk data stream; correlation feature extraction and data fusion mapping are performed on the usable multi-source heterogeneous disaster risk data stream to obtain a normalized disaster feature matrix.

[0128] Furthermore, the matrix construction module 2 is used to perform the following steps:

[0129] Correlation features are extracted from the available multi-source heterogeneous disaster risk data stream to obtain a multi-source heterogeneous disaster correlation feature set; the multi-source heterogeneous disaster correlation feature set is mapped to the same spatiotemporal scale for time synchronization and spatial alignment to obtain a standard multi-source heterogeneous disaster correlation feature set; and the standard multi-source heterogeneous disaster correlation feature set is normalized and fused to obtain the normalized disaster feature matrix.

[0130] Furthermore, the matrix generation module 3 is used to perform the following steps:

[0131] The normalized disaster characteristic matrix is ​​fitted and solved using the least squares method to obtain model fitting parameters, and a disaster development curve is generated based on the model fitting parameters; disaster characteristic correlation is calculated based on the disaster development curve to obtain a disaster characteristic correlation set; and the multi-source characteristic correlation matrix is ​​generated based on the disaster characteristic correlation set.

[0132] Furthermore, the set generation module 4 is configured to perform the following steps:

[0133] The multi-source feature association matrix is ​​subjected to matrix decomposition and association rule mining to obtain a disaster feature association rule data set; based on the disaster feature association rule data set, a tower stability prediction model, a casualty estimation model and a disaster level assessment model are respectively constructed; and the disaster risk prediction model is established by combining the tower stability prediction model, the casualty estimation model and the disaster level assessment model.

[0134] Furthermore, the heat map generation module 5 is configured to perform the following steps:

[0135] A threshold segmentation algorithm is used to divide the disaster risk index set into distribution requirements and determine the risk division threshold rules; threshold segmentation is performed on the disaster risk index set based on the risk division threshold rules to obtain a disaster risk level set; the disaster risk level set is mapped onto the spatial map of the target area and level rendering is performed to generate a regional disaster risk distribution heat map.

[0136] Furthermore, the model building module 6 is used to perform the following steps:

[0137] The dynamic change patterns of the regional disaster risk distribution heat map are analyzed to identify and obtain disaster risk characteristic change trend data; a time series model is used to perform fitting training and verification optimization on the disaster risk characteristic change trend data to construct the disaster risk time series model.

[0138] The disaster risk assessment and early warning system based on multi-source heterogeneous data provided by the embodiment of the present invention can execute the disaster risk assessment and early warning method based on multi-source heterogeneous data provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0139] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other and are not used to limit the scope of protection of the present invention.

[0140] The above specific embodiments do not constitute a limitation to the scope of protection of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of this application should be included in the scope of protection of this application. In some cases, the actions or steps recorded in this application can be performed in an order different from that in the embodiments and can still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A disaster risk assessment and early warning method based on multi-source heterogeneous data, characterized by: The method comprises: Acquire multi-source heterogeneous disaster risk data streams, where the data sources of the multi-source heterogeneous disaster risk data streams include tower, communication, population thermal, waterlogging risk assessment, and video data; Performing noise filtering and feature extraction fusion on the multi-source heterogeneous disaster risk data stream to construct a normalized disaster feature matrix; Using the least squares method to fit the normalized disaster characteristic matrix to a disaster development curve, and generate a multi-source characteristic correlation matrix; Performing disaster feature association mining in combination with the multi-source feature association matrix to establish a disaster risk prediction model, and generating a disaster risk index set based on the disaster risk prediction model; Based on the disaster risk index set, a threshold segmentation algorithm is used to perform risk division and rendering on the target area to generate a regional disaster risk distribution heat map; Constructing a disaster risk time series model based on the regional disaster risk distribution heat map; Based on the disaster risk time series model, a disaster assessment report is generated, and a visual early warning is performed on the disaster assessment report through three-dimensional simulation technology.

2. The disaster risk assessment and early warning method based on multi-source heterogeneous data according to claim 1, characterized in that: The constructing of the normalized disaster characteristic matrix includes: designing a multi-source heterogeneous data filter based on data characteristic information of the multi-source heterogeneous disaster risk data stream; Using the multi-source heterogeneous data filter to filter noise from the multi-source heterogeneous disaster risk data stream to obtain a usable multi-source heterogeneous disaster risk data stream; Correlation feature extraction and data fusion mapping are performed on the available multi-source heterogeneous disaster risk data stream to obtain a normalized disaster feature matrix.

3. The disaster risk assessment and early warning method based on multi-source heterogeneous data according to claim 2, characterized in that: The obtaining of the normalized disaster characteristic matrix includes: Extracting correlation features from the available multi-source heterogeneous disaster risk data stream to obtain a multi-source heterogeneous disaster correlation feature set; Mapping the multi-source heterogeneous disaster correlation feature set to the same spatiotemporal scale for time synchronization and spatial alignment to obtain a standard multi-source heterogeneous disaster correlation feature set; The standard multi-source heterogeneous disaster correlation feature set is normalized and fused to obtain the normalized disaster feature matrix.

4. The disaster risk assessment and early warning method based on multi-source heterogeneous data according to claim 1, characterized in that: Generating a multi-source feature association matrix includes: Using the least squares method to fit and solve the normalized disaster characteristic matrix to obtain model fitting parameters, and generating a disaster development curve based on the model fitting parameters; Calculating the disaster characteristic correlation based on the disaster development curve to obtain a disaster characteristic correlation set; The multi-source feature correlation matrix is ​​generated according to the disaster feature correlation degree set.

5. The disaster risk assessment and early warning method based on multi-source heterogeneous data according to claim 1, characterized in that: The establishment of a disaster risk prediction model includes: Performing matrix decomposition and association rule mining on the multi-source feature association matrix to obtain a disaster feature association rule data set; Based on the disaster characteristic association rule data set, a tower stability prediction model, a casualty estimation model and a disaster level assessment model are respectively constructed; The disaster risk prediction model is established by combining the tower stability prediction model, the casualty estimation model and the disaster level assessment model.

6. The disaster risk assessment and early warning method based on multi-source heterogeneous data according to claim 1, characterized in that: Generating a regional disaster risk distribution heat map includes: A threshold segmentation algorithm is used to divide the disaster risk index set into distribution requirements and determine a risk division threshold rule; Performing threshold segmentation on the disaster risk index set based on the risk segmentation threshold rule to obtain a disaster risk level set; The disaster risk level set is mapped onto the spatial map of the target area and level rendering is performed to generate a heat map of the regional disaster risk distribution.

7. The disaster risk assessment and early warning method based on multi-source heterogeneous data according to claim 1, characterized in that: The construction of the disaster risk time series model includes: Analyze the dynamic change patterns of the regional disaster risk distribution heat map to identify and obtain disaster risk characteristic change trend data; A time series model is used to perform fitting training and verification optimization on the disaster risk characteristic change trend data to construct the disaster risk time series model.

8. Disaster risk assessment and early warning system based on multi-source heterogeneous data, characterized by: A system for implementing the disaster risk assessment and early warning method based on multi-source heterogeneous data according to any one of claims 1 to 7, comprising: A data stream acquisition module is used to acquire multi-source heterogeneous disaster risk data streams, where the data sources of the multi-source heterogeneous disaster risk data streams include tower, communication, population thermal, waterlogging risk assessment and video data; A matrix construction module is used to perform noise filtering and feature extraction fusion on the multi-source heterogeneous disaster risk data stream to construct a normalized disaster feature matrix; A matrix generation module is used to fit the disaster development curve to the normalized disaster feature matrix using the least squares method to generate a multi-source feature correlation matrix; A set generation module is used to perform disaster feature association mining in combination with the multi-source feature association matrix, establish a disaster risk prediction model, and generate a disaster risk index set based on the disaster risk prediction model; A heat map generation module is used to perform risk division and rendering on the target area based on the disaster risk index set using a threshold segmentation algorithm to generate a regional disaster risk distribution heat map; A model building module, configured to build a disaster risk time series model based on the regional disaster risk distribution heat map; The report generation module is used to generate a disaster assessment report based on the disaster risk time series model, and to provide a visual warning for the disaster assessment report through three-dimensional simulation technology.

Images