Natural disaster damage intelligent assessment method and system based on artificial intelligence

Abstract

The invention discloses an artificial intelligence-based natural disaster damage intelligent assessment method and system. The method comprises the following steps of: acquiring a restoration period image and a disaster sensitive factor through a satellite remote sensing technology; performing land utilization type intelligent interpretation and generating a grid map layer, and automatically extracting geological disaster stress factors at the same time; innovatively coupling an InVEST ecological environment quality evaluation model and an AI disaster loss evaluation engine, synchronously calculating an ecological environment quality index and a disaster loss risk index, and generating an ecological quality thermodynamic diagram and a disaster sensitivity dynamic map; analyzing an evolvement rule of an ecological quality spatial pattern by adopting a space-time diagram convolutional network, and quantifying a synergistic effect in combination with a natural disaster damage degree-disaster resistance incidence matrix; and finally, multi-dimensional visual deduction of the natural disaster damage degree and the disaster resistance capability is realized on a three-dimensional digital twinborn platform, and an ecological restoration-disaster toughness double-target intelligent evaluation closed loop is formed.

Description

Technical Field

[0001] This invention relates to the field of ecological protection natural disaster damage assessment technology, and in particular to an intelligent assessment method and system for natural disaster damage based on artificial intelligence. Background Technology

[0002] The assessment of the degree of damage caused by natural disasters in the study area is an important means of verifying the effectiveness of ecological restoration projects. It is of great significance for further adjusting and optimizing restoration plans and provides guidance for subsequent ecological supervision work.

[0003] Currently, the assessment of natural disaster damage in the study area mainly relies on traditional monitoring technologies, and has not yet achieved deep integration with remote sensing technology, geographic information systems (GIS), and ecological assessment models, resulting in limitations in the spatiotemporal variation analysis of the assessment results. Furthermore, there is currently a lack of a widely applicable method and system for evaluating the effectiveness of ecological restoration projects already implemented in the study area in my country.

[0004] Therefore, from the perspective of ecological supervision, there is an urgent need to develop a set of intelligent assessment methods and supporting systems for natural disaster damage based on artificial intelligence. Summary of the Invention

[0005] To address the technical problems existing in related technologies, this disclosure provides an intelligent assessment method and system for natural disaster damage based on artificial intelligence.

[0006] First, data required for assessing the extent of natural disaster damage in the study area was collected. Then, preprocessing steps including data cleaning, data integration, and data standardization were performed on the collected data. Based on the preprocessed data, spatial analysis was conducted to extract spatial information related to ecological environment quality. The InVEST ecological environment quality model was used to obtain the ecological environment quality index and the ecological environment degradation index, and an artificial intelligence scoring model was used to comprehensively assess and monitor environmental quality. Finally, 3D renderings of the restoration projects and heat maps showing spatial differences in ecological indicators were provided. An interactive map was used to dynamically display the restoration process, analyze trends in ecological environment quality changes, and provide decision support for policymakers.

[0007] In this application, the step of performing land use classification on the preprocessed result to obtain a land use type raster layer includes:

[0008] The preprocessed results were classified for land use using eCognition software to obtain the classification results.

[0009] The classification results are corrected using an ArcGIS model to obtain a land use type raster layer.

[0010] Understandably, when preprocessing results are used for land use classification, the ability to correct the classification results ensures the accuracy of the classification data, thereby enabling a more accurate land use type raster layer to be obtained.

[0011] In this application, the step of extracting the stress factor raster layer based on the land use type raster layer includes:

[0012] The land use type raster layer is converted from vector to raster using the ArcGIS model to obtain a land use type vector map layer.

[0013] The attribute values ​​of the land use type vector layer are reassigned, with the stress factor assigned a value of 1 and the others assigned a value of 0, to obtain the reclassified land use type vector layer.

[0014] The stress factor raster layer is obtained by performing vector-raster conversion on the reclassified land use type vector layer using the ArcGIS model.

[0015] Understandably, extracting stress factor raster layers based on land use type raster layers can achieve data compatibility, thereby promoting flexible data integration within the system. This method can also adapt to long-term monitoring needs. For example, after land use changes, only the corresponding raster area needs to be replaced to quickly update the stress factor layer.

[0016] In this application, the stress factor attribute table includes: the stress factor label, the stress factor maximum influence distance, the stress factor weight, the stress factor attenuation type, and the stress factor raster layer path;

[0017] The stress factors are clustered according to the clustering rules of the stress factors to obtain the labels of the stress factors;

[0018] Based on the buffer analysis model, multi-ring buffers are established on the stress factor raster layer at different distance intervals, and the degradation index in each region is statistically analyzed to obtain the influence range of the stress factor.

[0019] The influence range of the stress factor was verified in the field and the model was optimized. The maximum value was selected to obtain the maximum influence distance of the stress factor.

[0020] The relative importance of the stress factors to the current regional ecological environment quality is scored based on the artificial intelligence scoring model, and the scoring results are obtained.

[0021] The scoring results are averaged to obtain the average value corresponding to the stress factor weights;

[0022] The influence of the stress factor on the current regional ecological environment quality is analyzed using the long-term trend analysis method, and the relationship between the current region and the distance from the stress factor is obtained to determine the attenuation type of the stress factor.

[0023] The stress factor raster layer is traced using an artificial intelligence data tracing model to obtain the path of the stress factor raster layer.

[0024] It is understood that the stress factor attribute table includes: the label of the stress factor, the maximum influence distance of the stress factor, the weight of the stress factor, the attenuation type of the stress factor, and the path of the stress factor raster layer. This attribute table can standardize the data interface, accurately quantify the spatial influence, ensure the transparency and traceability of the model input data, and facilitate subsequent verification work.

[0025] In this application, the sensitivity table includes: land category code, land category label, ecological environment suitability, and the relative sensitivity of each land category to each stress factor;

[0026] The land use type is numbered according to the land use type raster layer to obtain the land use type code;

[0027] The land use types are clustered according to the land use type raster layer to obtain the land use type labels;

[0028] Based on the current regional information, the ecological environment suitability scores are assigned to different land types to obtain the ecological environment suitability score.

[0029] Based on the current regional information, a score is assigned to the relative sensitivity of each land type to each stress factor, thus obtaining the relative sensitivity of each land type to each stress factor.

[0030] It is understood that the sensitivity table includes: land category code, land category label, ecological environment suitability, and the relative sensitivity of each land category to each stress factor. This table can standardize land category classification management and quantify the ecological baseline value and its sensitivity.

[0031] In this application, the calculation steps for the half-saturation parameter include:

[0032] Based on the current region information, determine the range of values ​​for the half-saturation parameter;

[0033] By repeatedly inputting values ​​within the range of the half-saturation parameter into the InVEST ecological environment quality assessment model, the changes in the model calculation results are analyzed, the impact of the half-saturation parameter on the model calculation results is evaluated, and the value input when the best model calculation result is obtained is taken as the half-saturation parameter.

[0034] Understandably, the calculation steps for the half-saturation parameter can balance theoretical assumptions and empirical data, thereby improving model accuracy.

[0035] In this application, the InVEST ecological environment quality assessment model is based on ecological formulas and uses spatial relationships to calculate pixel by pixel to quantify the ecological environment quality of the current area.

[0036] It is understood that the InVEST ecological environment quality assessment model is based on ecological formulas and uses spatial relationships to calculate pixel by pixel to quantify the ecological environment quality of the current area. This method can scientifically quantify complex ecological processes and accurately reflect the spatial distribution differences of ecosystem services.

[0037] In this application, the calculation results of the InVEST ecological environment quality assessment model are presented in the form of a raster map, including an ecological environment quality map and an ecological environment degradation map, which can intuitively reflect the spatial distribution and changing trend of the degree of natural disaster damage in the study area.

[0038] It is understandable that the calculation results of the InVEST ecological environment quality assessment model are presented in the form of a raster map, which can achieve a balance between calculation efficiency and accuracy requirements by adjusting the raster size, and flexibly respond to different management needs at multiple scales.

[0039] In this application, the spatial exploratory evaluation model obtains the spatial pattern evolution characteristics of ecological environment quality by analyzing the spatial clustering of ecological environment quality, and then obtains the evaluation results of the degree of natural disaster damage in the study area.

[0040] It is understood that the spatial exploratory evaluation model can identify spatial clusters, gradient changes, and outliers in the data, revealing complex spatial relationships.

[0041] In this application, the system includes:

[0042] The data preprocessing module is used to collect data on the extent of natural disaster damage in the study area and to perform preprocessing procedures such as data cleaning, data integration, and data standardization on the collected data.

[0043] The GIS spatial analysis module is used to perform spatial analysis on data after preprocessing procedures such as data cleaning, data integration, and data standardization, and to extract spatial information related to ecological environment quality.

[0044] The ecological environment assessment module is used to build a comprehensive key indicator library. It uses an artificial intelligence scoring model to quantify the importance of specific stress factors relative to the current regional ecological environment quality. It uses the InVEST ecological environment quality model to calculate the ecological environment quality index and the ecological environment degradation index, so as to achieve comprehensive assessment and monitoring of ecological environment quality.

[0045] The decision support and visualization module is used for interactive map display, providing 3D renderings of restoration projects and heat maps of spatial differences in ecological indicators, dynamically displaying the restoration process, analyzing the trend of changes in ecological environment quality, and providing decision support for decision-makers.

[0046] It is understood that the system includes a data preprocessing module, a GIS spatial analysis module, an ecological environment assessment module, and a decision support and visualization module. The system has the function of eliminating redundancy and conflict in the original data, supports multi-scale spatial statistical analysis, and can display the spatial distribution and temporal evolution characteristics of the assessment results through interactive visualization technology, thereby reducing the difficulty of information understanding.

[0047] The technical solutions provided by the embodiments of this disclosure may include the following beneficial effects.

[0048] An AI-based intelligent assessment method and system for natural disaster damage integrates multi-dimensional data and can customize assessment indicators based on watershed specificity, adapting to different climatic and geographical conditions to achieve a comprehensive evaluation of the ecological environment quality of the study area. The modular design makes the assessment process more flexible and efficient, enabling precise analysis based on specific watershed characteristics. Furthermore, the method and system possess powerful data processing and analysis capabilities, enabling real-time monitoring of ecological environment change trends, timely identification of potential problems, and providing a scientific basis for restoration work.

[0049] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this disclosure. Attached Figure Description

[0050] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the specification, serve to explain the principles of this application.

[0051] Figure 1 This is a flowchart illustrating an artificial intelligence-based intelligent assessment method for natural disaster damage provided in an embodiment of this application. Detailed Implementation

[0052] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of methods and systems consistent with some aspects of this application as detailed in the appended claims.

[0053] This application provides an artificial intelligence-based intelligent assessment method and system for natural disaster damage, addressing the shortcomings of existing methods that lack deep integration of multiple technological approaches. This artificial intelligence-based intelligent assessment method and system for natural disaster damage can achieve deep integration with remote sensing technology, geographic information systems (GIS), and ecological assessment models.

[0054] Remote sensing technology is a scientific and technological system that uses electromagnetic waves to conduct non-contact detection and imaging of the Earth's surface through satellites, aircraft, or other long-distance sensor platforms, to acquire spatial, spectral, and temporal information about ground features, and then analyzes and processes this information.

[0055] Geographic Information Systems (GIS) are comprehensive information management systems that utilize computer hardware and software technologies to collect, store, manage, analyze, and visualize geospatial data in order to support spatial decision-making.

[0056] Ecological assessment models are a systematic methodology that uses mathematical modeling, data analysis, and computer simulation to quantitatively assess the structure, function, health status, or service value of ecosystems, providing a scientific basis for environmental protection and sustainable management.

[0057] In the embodiments of this application, the target area is a research area where ecological restoration has been implemented, an area where the degree of damage from natural disasters needs to be assessed, such as a typical research area in the Loess Hilly Gully region where gully management and artificial grass planting are carried out, or a research area in a key water source protection area where systematic ecological restoration is carried out.

[0058] Example 1

[0059] like Figure 1 As shown, this application provides an intelligent assessment method for natural disaster damage based on artificial intelligence, the method comprising:

[0060] Step S101: Collect remote sensing image data of the early, middle and late stages of the ecological restoration project in the study area, preprocess the remote sensing image data, and obtain the preprocessing results.

[0061] Remote sensing image data can be selected from Landsat-8 or Gaofen-1 satellite remote sensing image data. In this embodiment, remote sensing image data of the ecological restoration project in the study area acquired by Gaofen-1 satellite in the early, middle and late stages were collected.

[0062] The evaluation cycle consists of three phases: early stage, mid-stage, and late stage.

[0063] The initial period refers to the year preceding the implementation of ecological environment protection and restoration policies, plans, and projects within the target area, or the base year, serving as the initial time for comparing various evaluation indicators.

[0064] The medium-term phase involves the implementation of relevant policies, plans, and projects for ecological environmental protection and restoration, and will be carried out as needed in a timely manner.

[0065] Two years after the completion and acceptance of the overall policies, plans, and projects related to ecological environment protection and restoration, the vegetation will undergo a selection and reproduction process during the two-year growing season.

[0066] The preliminary remote sensing image data includes: remote sensing image data of the target area before the implementation of policies, plans and projects related to ecological environment protection and restoration;

[0067] The mid-term remote sensing imagery data includes: remote sensing imagery data of the target area after the completion of each sub-task of the ecological restoration project. These sub-tasks include site remediation, soil improvement, vegetation planting, water body restoration, and landscape planning and design.

[0068] The subsequent remote sensing image data includes: remote sensing image data of the target area two years after the overall completion and acceptance of the ecological restoration project;

[0069] Landsat-8 satellite carries the OLI and TIRS pushbroom imagers. The OLI includes nine bands, with a spatial resolution of 30m for the panchromatic band and 15m for the other bands. The TIRS includes two thermal infrared bands with a spatial resolution of 100m. It has advantages in distinguishing the spectral characteristics of forest land and farmland and is suitable for long-term time series variation analysis of large areas.

[0070] The Gaofen-1 satellite carries two 2m resolution panchromatic and 8m resolution multispectral high-resolution cameras, and four 16m resolution multispectral wide-field cameras. It has advantages in fine-grained classification of urban land use (roads, buildings), scattered wetlands or road boundaries, and identification of fragmented features. Compared with remote sensing image data acquired by Landsat-8 satellite, Gaofen-1 satellite has higher image resolution, making it more suitable for evaluating the effectiveness of restoration in the study area.

[0071] Raw remote sensing image data, acquired by satellite sensors, typically contains defects such as outliers, missing data, geometric distortions, and radiometric distortions, thus requiring preprocessing. The specific contents of preprocessing include:

[0072] The periodic offset of the sensors on the satellite or the electromagnetic interference between the payload components can cause outliers in the remote sensing image data. Fourier transform is calculated to filter the remote sensing image data and remove outliers.

[0073] Remote sensing image data may be missing due to multiple factors such as cloud cover and sensor failure. For areas with missing data, the median of remote sensing image data that are adjacent in time series within the same geographical area is used for interpolation to fill in the missing data.

[0074] When remote sensing images are formed, factors such as photographic material deformation, lens distortion, atmospheric refraction, Earth curvature, Earth rotation, and topographic relief can cause the geometric position, shape, size, orientation, and other features of various objects in the remote sensing image data to be inconsistent with the expression requirements in the reference system, resulting in geometric distortion. By selecting the control point dataset between the distorted remote sensing image and the standard map as samples, a geometric distortion model is trained, and then the model is used to perform geometric distortion correction.

[0075] Satellite sensors are affected by factors such as atmospheric aerosols, topographic features, and nearby land features, resulting in remote sensing image data that integrates comprehensive information from object surfaces, the atmosphere, and solar radiation. To extract the spectral characteristics of surfaces of specific land use types, an atmospheric radiative transfer correction model is used to convert the digital quantization values ​​of remote sensing image data into radiance values, and then the radiance values ​​are converted into the actual reflectance of the land use type surface.

[0076] Step S102: Classify the preprocessed results for land use to obtain a land use type raster layer;

[0077] In this embodiment, land use classification is performed on the preprocessed results using eCognition software to obtain classification results;

[0078] The land use classification is based on the first-level categories in the "Classification of Current Land Use" (GB / T 21010—2007);

[0079] In the field of remote sensing image processing, land use classification focuses on dividing pixels or cells in an image into different categories based on specific or multiple features. This classification process simulates the recognition mechanism of human vision to achieve automatic identification and understanding of land use types. In this embodiment, the land use types in the ecological restoration area of ​​the study area are classified into forest land, grassland, cultivated land, urban residential land, transportation land, water bodies, wasteland, and geological disaster land.

[0080] The key technology for land use classification using eCognition software is object-oriented classification. This method integrates geometric, topological, spectral, and textural information from imagery, combining image pixels into meaningful regions or objects with unique characteristics related to their surrounding environment. Therefore, the classification process includes two main steps: object generation and object classification.

[0081] In the object generation stage, this study adopted a multi-scale segmentation algorithm to segment the image and form homogeneous objects at different scales. The procedure involves the following steps: First, the homogeneity of pixels within adjacent areas and the heterogeneity between areas are calculated; second, based on the calculation results of homogeneity and heterogeneity, the most suitable merging strategy is selected to construct objects; finally, the most suitable segmentation scale is determined through set parameters or adaptive algorithms to ensure that the surface features of land use types can be fully expressed at multiple scales.

[0082] In the object classification stage, objects are assigned category labels based on training samples or a specific rule set. This process involves the following steps: First, spectral, texture, and shape features are extracted from the objects. Second, a classifier is trained using a precisely labeled training sample dataset, ensuring that this dataset adequately represents the features of each category. Subsequently, the trained classifier is applied to unlabeled image objects, assigning them to the appropriate category based on feature similarity. Finally, cross-validation is used to evaluate the accuracy of the classification results, and the classification rules are adjusted based on the evaluation results to improve the quality of the classification results.

[0083] The classification results are corrected using an ArcGIS model to obtain a land use type raster layer.

[0084] Specifically, the land use type results generated by eCognition software are imported into the ArcGIS model for correction and processing, which involves a series of processes such as removing small patches, optimizing classification results, smoothing boundaries, calculating area, and improving visualization effects.

[0085] Step S103: Obtain stress factors affecting the ecological environment of the study area through historical data, and extract stress factor raster layers based on land use type raster layers and vector layers.

[0086] In this embodiment, the stressors on the ecological environment of the study area are arable land, urban residential land, transportation land, geological disasters, etc.

[0087] Among them, the stress factors of farmland reclamation and use on the ecological environment are reflected in large-scale monoculture, the widespread use of chemical fertilizers and pesticides, and deforestation, wetland filling or grassland destruction caused during the reclamation process.

[0088] The development and use of urban residential land poses a threat to the ecological environment in the following ways: untreated domestic sewage is discharged into water bodies, carbon dioxide emissions, excessive groundwater extraction, hardening of the ground, accumulation of construction waste, and improper disposal of urban domestic waste.

[0089] The development and use of transportation land has a stress on the ecological environment in the following ways: transportation networks cut off continuous natural landscapes, occupy habitats of animals and plants, bridge construction changes the shape of river channels, fuel vehicles emit harmful gases, and the habitats around transportation lines are disturbed by sunlight, noise and pollution, leading to an aggravated edge effect.

[0090] The stress factors of geological disasters on the ecological environment are manifested in the direct damage of surface vegetation and soil layers caused by geological processes such as earthquakes, landslides, collapses, and debris flows. Debris flows block river channels and change downstream hydrological conditions. The ecological restoration of areas that have experienced geological disasters has a significant lag.

[0091] Based on land use type raster layers, raster layers of stress factors such as cultivated land, urban residential land, transportation land, and geological hazards are extracted.

[0092] By performing clustering and using ArcGIS model spatialization processing to convert the land use type raster layer to a vector layer, a stress factor vector layer is obtained.

[0093] In this embodiment, the K-means clustering method is used to perform cluster analysis on cultivated land, urban residential land, transportation land and geological hazards. The cluster analysis results are input into the ArcGIS model for spatialization processing. The stress factor is assigned a value of 1 and the others are assigned a value of 0 using binary encoding. The stress factor vector layer is obtained after clustering and spatialization processing.

[0094] K-means clustering is an unsupervised machine learning method. Its principle is to divide the dataset into several clusters with high internal similarity, while the differences between data points in different clusters are relatively large. The core of the algorithm is to continuously optimize through an iterative process to achieve the goal of minimizing the sum of distances between each data point and the centroid of its corresponding cluster.

[0095] The stress factor vector layer is converted from vector to raster using the ArcGIS model to obtain a stress factor raster layer.

[0096] Step S104: Obtain the stress factor attribute table, sensitivity table, and half-saturation parameter using the current region information.

[0097] In this embodiment, the information of the stress factor attribute table, sensitivity table, and half-saturation parameter is limited to the range of the influence factor determination box of the maximum permissible area;

[0098] Among them, the impact factor determination box of the maximum permissible area is constructed based on the natural geographical boundary of the study area, combined with the sensitivity of the ecosystem and the impact range of human activities, which limits the evaluation scope to a complete ecological functional unit and avoids the impact of cross-basin disturbance factors on the evaluation results.

[0099] The natural geographical boundary of the study area is essentially a watershed line, which usually corresponds to a topographic ridge. Precipitation flows to different rivers or catchment areas on either side of this boundary.

[0100] The stress factor attribute table includes: the stress factor's label, the stress factor's maximum influence distance, the stress factor's weight, the stress factor's attenuation type, and the stress factor's raster layer path.

[0101] The stress factors are clustered according to the clustering rules of the stress factors to obtain the labels of the stress factors.

[0102] In this embodiment, the stress factor is labeled with the corresponding land use type name, namely: cultivated land, urban land, highway, and debris flow.

[0103] Based on the buffer analysis model, multi-ring buffers are established on the stress factor raster layer at different distance intervals, and the degradation index in each region is statistically analyzed to obtain the influence range of the stress factor.

[0104] The influence range of the stress factor was verified in the field and the model was optimized. The maximum value was selected to obtain the maximum influence distance of the stress factor.

[0105] Buffer analysis is a spatial analysis method for solving proximity problems. For example, it automatically generates a polygon layer of a specific width around features such as points, lines, and polygons—this is the buffer. By overlaying this buffer layer onto the target layer, the desired spatial analysis results can be obtained, thus addressing various proximity-related issues.

[0106] The scoring results are obtained by scoring the relative importance of the stress factors to the current regional ecological environment quality based on the artificial intelligence scoring model.

[0107] The rules of AI-based scoring models are constructed based on pre-defined mathematical formulas or logical frameworks, such as the analytic hierarchy process (AHP) and entropy weighting. The model assigns different weights and scoring standards to stress factors based on their attributes, scope of influence, and degree of impact. For example, for arable land as a stress factor, the model comprehensively considers its area, degradation index, and interaction with the surrounding ecological environment, and uses algorithms to calculate its relative importance score to the current regional ecological environment quality. For other stress factors such as urban residential land, transportation land, and geological hazards, the model similarly performs corresponding scoring calculations based on their respective characteristics and impact mechanisms.

[0108] The Analytic Hierarchy Process (AHP) is a subjective weighting method that integrates qualitative and quantitative analysis. It hierarchically and quantifies influencing factors, using experience or expert consultation to determine and measure the relative importance of lower-level factors to higher-level factors, and rationally allocates weights to each decision-making option under different evaluation criteria. The ranking of the options is then determined through these weights, providing a scientific basis for decision-making.

[0109] Entropy weighting is an objective weighting method based on the theory of information entropy. It holds that the greater the data dispersion of an indicator, the more information it provides, and therefore it is given a high weight; conversely, the smaller the data dispersion of an indicator, the less information it provides, and therefore it is given a low weight.

[0110] In this embodiment, when stress factors are assigned weights as evaluation indicators, their attributes, scope of influence, and degree of influence need to be comprehensively considered. The entropy weight method assigns weights entirely based on data, requiring high data quality, but it ignores the subjective factors of the indicators. However, the analytic hierarchy process (AHP) can comprehensively consider the practical significance of the indicators. Therefore, an artificial intelligence scoring model based on AHP is used to assign weights to stress factors. This model is based on a multi-level analytical framework, including a target layer, a criterion layer, and an indicator layer, to decompose complex ecological stress problems and transform them into quantifiable and comparable subsystems, clarifying the logical relationships between various stress factors. Since stress factors contain both quantifiable data and subjective judgments, the model uses a comparative method, combining expert experience with objective data. Through a unified weight assignment process, subjective and objective information is integrated into relative weight values. The resulting scoring directly reflects the relative importance of each stress factor.

[0111] The scoring results are averaged to obtain the average value corresponding to the stress factor weights.

[0112] The influence of the stress factor on the current regional ecological environment quality is analyzed using the long-term trend analysis method, and the relationship between the current region and the distance from the stress factor is obtained to determine the attenuation type of the stress factor.

[0113] Long-term trend analysis is a method for identifying and predicting the overall direction or pattern of data over time. Through in-depth analysis of historical data, it eliminates short-term fluctuations such as seasonal and cyclical changes, identifying stable trends in the data over the long term. Commonly used statistical algorithms in this method include moving averages and trend line fitting.

[0114] The moving average method smooths out short-term fluctuations by calculating the average value over a continuous interval in a time series, eliminating the influence of trend fluctuations, seasonal fluctuations, periodic fluctuations, and random fluctuations on the time series values, thereby identifying the trend changes within them.

[0115] Trendline fitting is a method that uses mathematical functions to model the long-term trend of a time series, determining the best-fit curve to describe the overall direction of data evolution over time. The mathematical functions involved in this method include linear regression analysis, multinomial regression analysis, and exponential growth models.

[0116] The moving average method is suitable for data sequences that fluctuate significantly in the short term and lack a clear fixed trend, while the trend line fitting method is suitable for data sequences with a clear mathematical trend. It is known that grid cells closer to the stress factor will suffer a more significant impact, and the impact of the stress factor on the ecological environment decreases with increasing distance from the degradation source. Therefore, in this embodiment, the trend line fitting method is used to explore the attenuation law of different stress factors over spatial distance. Linear regression analysis or an exponential growth model can be used to characterize the spatial attenuation effect of the stress factor's threat to the ecological environment. The specific formula is as follows:

[0117]

[0118] In the formula: i rxy This represents the effect of the stress factor r on the habitat of raster x on raster y, d xy It is the linear distance between grid x and grid y; d rmax It is the maximum effective distance of the stress factor r.

[0119] The stress factor raster layer is traced using an artificial intelligence data tracing model to obtain the path of the stress factor raster layer.

[0120] The artificial intelligence data tracing model performs tracing processing on the stress factor raster layer. The core steps include: metadata extraction, hierarchical path design, unique identifier injection, version control system integration, and storage adaptation processing, ultimately obtaining the path of the stress factor raster layer.

[0121] Metadata extraction involves extracting structured information from a dataset that describes and interprets the data, while providing background information such as the data’s origin, data type, and relationships with other datasets.

[0122] Hierarchical path design is a directory structure built based on dimensions such as project, model, version, and time.

[0123] Unique identifier injection uses unique identifiers (UUIDs), globally unique identifiers (GUIDs), or serialization numbers to ensure the uniqueness of a path.

[0124] Version control system integration is a systematic approach to managing file changes, used to track and record the modification history of code, documents, and other files throughout the project lifecycle.

[0125] Storage adaptation automatically ensures compatibility between local and cloud storage path formats.

[0126] The sensitivity table includes: land type code, land type label, ecological environment suitability, and the relative sensitivity of each land type to each stress factor.

[0127] The land use type is numbered according to the land use type raster layer to obtain the land use type code.

[0128] The land use coding system includes three levels: basic land use, derived land use, and refined land use. Basic land use includes agricultural land, construction land, and unused land. Derived land use is based on the basic land use and is divided according to factors such as the current use status, potential, and suitability of the land. Refined land use is a further subdivision of the derived land use.

[0129] The land use coding rules use three digits, where the first digit represents the basic land use type, the second digit represents the derived land use type, and the third digit represents the refined land use type.

[0130] In this embodiment, the national standard "Classification of Current Land Use" (GB / T 21010—2007) is referenced. The land use types are coded as follows: forest land as 03; grassland as 04; cultivated land as 01; urban land as 20; highways as 10; water bodies as 11; and wasteland as 12.

[0131] The land use types are clustered based on the land use type raster layer to obtain the land use type labels.

[0132] In this embodiment, the land use type is labeled with the corresponding land use type name, which are: forest land, grassland, cultivated land, urban land, highway, water body, and wasteland.

[0133] Based on the current regional information, the ecological environment suitability of different land types is scored to obtain the ecological environment suitability score.

[0134] Using binary coding, land types that are unsuitable for the ecological environment are scored as 0, and land types that are perfectly suitable for the ecological environment are scored as 1.

[0135] Based on the current regional information, a score is assigned to the relative sensitivity of each land type to each stress factor, thus obtaining the relative sensitivity of each land type to each stress factor.

[0136] In this embodiment, the relative sensitivity scores of various land use types to different stress factors are set between 0 and 1, where 1 represents high sensitivity and 0 indicates that the land use type is not affected by the stress factor.

[0137] The steps for calculating the half-saturation parameter include:

[0138] Based on the current area information, the range of values ​​for the half-saturation parameter is determined. The value of the half-saturation parameter is usually set to 50% of the maximum value of habitat degradation.

[0139] By repeatedly inputting values ​​within the range of the half-saturation parameter into the InVEST ecological environment quality assessment model, the changes in the model calculation results are analyzed, the impact of the half-saturation parameter on the model calculation results is evaluated, and the value input when the best model calculation result is obtained is taken as the half-saturation parameter.

[0140] Step S105: Construct the InVEST ecological environment quality assessment model by inputting the land use type raster layer, the stress factor raster layer, the stress factor attribute table, the sensitivity table, and the half-saturation parameter into the InVEST ecological environment quality assessment model for calculation to obtain the calculation results.

[0141] The InVEST ecological environment quality assessment model evaluates ecological environment quality based on land use type and the impact of each stress factor on the ecological environment, and further assesses the sustainability and resilience of biodiversity. This model assumes that areas with better ecological environment quality have richer biodiversity and stronger resilience. To obtain the ecological environment quality index, the degree of ecological environment degradation must first be calculated, using the following formula:

[0142]

[0143] Among them, D xj R represents the degree of habitat degradation in grid x within land use j, where R is the number of stress factors, and C is the number of stress factors. r Y represents the weight of the stress factor r. r r represents the total number of raster cells in the stress factor layer on the land cover layer. y Let i be the number of a certain stress factor. rxy This represents the effect of the stress factor r on the habitat of raster x on raster y, β x It is the level at which the grid cells are protected, S jr This indicates the relative sensitivity of land use type j to stress factor r.

[0144] After calculating the degree of ecological degradation, the raster degradation score is interpreted into a habitat quality score using a half-saturation function, as shown in the following formula:

[0145] In the formula, Q xj H represents the habitat quality of grid x in land use j. jFor the habitat attributes of land use j, D xj Z represents the degree of habitat degradation of grid x in land use j, Z is the normalization index with the default values ​​of the model, and K is the half-saturation parameter.

[0146] The calculation results of the InVEST ecological environment quality assessment model are presented in the form of a raster map, including an ecological environment quality map and an ecological environment degradation map, which can intuitively reflect the spatial distribution and changing trend of the degree of natural disaster damage in the study area.

[0147] Step S106: The calculation results are processed through a spatial exploratory evaluation model to obtain the evaluation results of the degree of natural disaster damage in the study area.

[0148] Spatial exploratory evaluation models analyze the spatial clustering of ecological and environmental quality to obtain the spatial pattern evolution characteristics of ecological and environmental quality, and then obtain the evaluation results of the degree of natural disaster damage in the study area.

[0149] In this embodiment, the spatial exploratory evaluation model uses spatial autocorrelation analysis and principal component analysis to detect and analyze spatial change trends and driving factors.

[0150] The input data of the spatial exploratory evaluation model is divided into target variables and driving factors. The data is in the form of raster data. The target variable is the ecological environment quality map. The driving factors include: stress factor raster layer, land use raster layer, and the influencing factor determination box of the maximum permissible area, etc.

[0151] Spatial autocorrelation analysis tools in the spatial exploratory evaluation model are used to determine the overall dispersion of ecological environment quality in the study area, thereby describing the spatial heterogeneity distribution characteristics of ecological environment quality. Further analysis is conducted on the degree of correlation of ecological environment quality among different regions within the study area, revealing the spatial clustering pattern of ecological environment quality, and identifying hotspot areas where high or low ecological environment quality values ​​are clustered.

[0152] Using principal component analysis in a spatial exploratory evaluation model, the main components of the driving factors were extracted, the correlation coefficients between each main component and the ecological environment quality were calculated, and the degree of influence and contribution rate of each main component on the spatial pattern evolution of the ecological environment quality in the study area were determined. Among them, the main component with the highest degree of influence and contribution rate on the spatial pattern evolution of the ecological environment quality in the study area was identified as the key driving factor. The analysis results revealed the main reasons affecting the spatial pattern evolution of the ecological environment quality in the study area.

[0153] Example 2

[0154] Based on the same inventive concept as the artificial intelligence-based intelligent assessment method for natural disaster damage in the foregoing embodiments, this application provides a system for assessing the degree of natural disaster damage in a study area. The specific description of the artificial intelligence-based intelligent assessment method for natural disaster damage in Embodiment 1 is also applicable to this system, wherein the system includes:

[0155] The disaster data intelligent preprocessing module is used to repair missing data and filter noise by integrating high-resolution remote sensing images, real-time monitoring data and historical disaster damage databases, and to construct a spatiotemporally aligned disaster risk assessment quantity.

[0156] The disaster chain spatial diagnosis module is used to extract disaster-causing factors such as slope stability in real time based on edge computing, construct a vulnerability matrix by combining the identified distribution of disaster-bearing bodies, and use a three-dimensional particle tracking algorithm to simulate the transmission path of disasters such as debris flows to generate a disaster chain impact radiation map.

[0157] The intelligent disaster risk assessment module is used to innovate and expand the InVEST model, integrate dynamic disaster loss formulas, and quantify the assessment results of direct economic losses, ecosystem service losses, and lifeline interruption risks.

[0158] The three-dimensional simulation module for prevention and control decision-making is used to dynamically render disaster heat maps and prevention and control engineering sand tables on the digital twin platform. It generates the optimal engineering plan through a reinforcement learning decision engine and optimizes emergency evacuation routes based on an intelligent agent model, forming a "assessment-simulation-optimization" decision-making closed loop.

[0159] In summary, when applying the above-mentioned scheme, an AI-based intelligent assessment method and system for natural disaster damage, by integrating multi-dimensional data, can customize assessment indicators according to watershed specificity, adapting to different climatic and geographical conditions, and achieving a comprehensive evaluation of the ecological environment quality of the study area. The introduction of modular design makes the assessment process more flexible and efficient, enabling precise analysis based on specific watershed characteristics. Furthermore, the method and system possess powerful data processing and analysis capabilities, enabling real-time monitoring of ecological environment change trends, timely identification of potential problems, and providing a scientific basis for restoration work.

[0160] It should be understood that this application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method and system for intelligent assessment of natural disaster damage based on artificial intelligence, characterized in that, Includes the following steps: Remote sensing image data of the early, middle and late stages of ecological restoration projects in the study area were collected, and geological disaster monitoring data were simultaneously accessed. The multi-source data were spatiotemporally aligned and adaptively filtered for preprocessing to obtain preprocessed data. The preprocessed results are classified for land use to obtain a land use type raster layer; Construct an intelligent extraction engine for disaster-sensitive factors, input the preprocessed data and the land use type raster layer data, and output a stress factor raster layer that integrates ecological and disaster attributes; Based on the current region information, obtain the stress factor attribute table, sensitivity table, and half-saturation parameter; An InVEST-AI joint assessment model is constructed. The land use type raster layer, the stress factor raster layer, the stress factor attribute table, the sensitivity table, and the half-saturation parameter are input into the InVEST-AI joint assessment model for calculation to obtain the ecological restoration-disaster resilience synergistic index map. The calculation results are input into a spatial exploratory evaluation model for processing to obtain the assessment results of the direct economic losses of geological disasters to the ecological environment in the study area.

2. The intelligent assessment method for natural disaster damage based on artificial intelligence according to claim 1, characterized in that, The step of classifying the preprocessed results for land use to obtain a land use type raster layer includes: The preprocessed results were classified for land use using eCognition software to obtain the classification results. The classification results are corrected using an ArcGIS model to obtain a land use type raster layer.

3. The intelligent assessment method for natural disaster damage based on artificial intelligence according to claim 1, characterized in that, The steps for constructing a disaster-sensitive factor intelligent extraction engine to extract stress factor raster layers include: Input the real-time monitoring data stream of geological disasters, the land use raster layer, and the geological vector map layer; The real-time data stream is spatiotemporally aligned with the raster data using a sliding window dynamic registration algorithm. Adaptive resolution conversion technology is used to convert geological vector layers into weighted raster layers; Dynamically quantify disaster-sensitive factors and calculate parameters such as slope stability index and soil erosion risk value; The slope stability index, soil erosion risk value and other parameters are converted into raster layers to obtain slope stability raster, soil erosion risk raster and other layers. The slope stability raster, soil erosion risk raster, and other layers are fused with the stress factor raster layer by factor weighting and normalization to obtain the stress factor raster layer of ecological-hazard attributes.

4. The artificial intelligence-based intelligent assessment method for natural disaster damage according to claim 1, characterized in that, The stress factor attribute table includes: the stress factor label, the maximum influence distance of the stress factor, the stress factor weight, the stress factor attenuation type, and the stress factor raster layer path. The stress factors are clustered according to the clustering rules of the stress factors to obtain the labels of the stress factors; Based on the buffer analysis model, multi-ring buffers are established on the stress factor raster layer at different distance intervals, and the degradation index in each region is statistically analyzed to obtain the influence range of the stress factor. The influence range of the stress factor was verified in the field and the model was optimized. The maximum value was selected to obtain the maximum influence distance of the stress factor. The relative importance of the stress factors to the current regional ecological environment quality is scored based on the artificial intelligence scoring model, and the scoring results are obtained. The scoring results are averaged to obtain the average value corresponding to the stress factor weights; The influence of the stress factor on the current regional ecological environment quality is analyzed using the long-term trend analysis method, and the relationship between the current region and the distance from the stress factor is obtained to determine the attenuation type of the stress factor. The stress factor raster layer is traced using an artificial intelligence data tracing model to obtain the path of the stress factor raster layer.

5. The intelligent assessment method for natural disaster damage based on artificial intelligence according to claim 1, characterized in that, The sensitivity table includes: land type code, land type label, ecological environment suitability, and the relative sensitivity of each land type to each stress factor; The land use type is numbered according to the land use type raster layer to obtain the land use type code; The land use types are clustered according to the land use type raster layer to obtain the land use type labels; Based on the current regional information, the ecological environment suitability scores are assigned to different land types to obtain the ecological environment suitability score. Based on the current regional information, a score is assigned to the relative sensitivity of each land type to each stress factor, thus obtaining the relative sensitivity of each land type to each stress factor.

6. The intelligent assessment method for natural disaster damage based on artificial intelligence according to claim 1, characterized in that, The calculation steps for the half-saturation parameter include: Based on the current region information, determine the range of values ​​for the half-saturation parameter; By repeatedly inputting values ​​within the range of the half-saturation parameter into the InVEST ecological environment quality assessment model, the changes in the model calculation results are analyzed, the impact of the half-saturation parameter on the model calculation results is evaluated, and the value input when the best model calculation result is obtained is taken as the half-saturation parameter.

7. The intelligent assessment method for natural disaster damage based on artificial intelligence according to claim 1, characterized in that, The InVEST ecological environment quality assessment model is based on ecological formulas and uses spatial relationships to calculate pixel by pixel to quantify the ecological environment quality of the current area.

8. The intelligent assessment method for natural disaster damage based on artificial intelligence according to claim 1, characterized in that, The calculation results of the InVEST ecological environment quality assessment model are presented in the form of a raster map, including an ecological environment quality map and an ecological environment degradation map, which can intuitively reflect the spatial distribution and changing trend of the degree of natural disaster damage in the study area.

9. The intelligent assessment method for natural disaster damage based on artificial intelligence according to claim 1, characterized in that, The AI ​​disaster damage assessment engine uses deep learning algorithms, combining historical disaster data and real-time monitoring data, to automatically learn and identify key factors in the occurrence of disasters, such as rainfall and topography, as well as the degree of impact of these factors on disaster losses, and to intelligently predict and assess the degree of damage from natural disasters. The spatial exploratory evaluation model analyzes the spatial clustering of ecological and environmental quality to obtain the spatial pattern evolution characteristics of ecological and environmental quality, and then obtains the evaluation results of the degree of natural disaster damage in the study area.

10. A system for assessing natural disaster damage in a study area based on artificial intelligence, characterized in that, The system includes: The disaster data intelligent preprocessing module integrates high-resolution remote sensing images, real-time monitoring data, and historical disaster damage databases to repair missing data and filter noise, and construct a spatiotemporally aligned disaster risk assessment quantity. The disaster chain spatial diagnosis module extracts disaster-causing factors such as slope stability in real time based on edge computing, constructs a vulnerability matrix by combining the identified distribution of disaster-bearing bodies, and uses a three-dimensional particle tracking algorithm to simulate the transmission path of disasters such as debris flows, generating a disaster chain impact radiation map. The intelligent disaster risk assessment module innovatively expands the InVEST model, integrates dynamic disaster loss formulas, and quantifies the assessment results of direct economic losses, ecosystem service losses, and lifeline interruption risks. The three-dimensional simulation module for prevention and control decision-making dynamically renders disaster heat maps and prevention and control engineering sand tables on the digital twin platform. It generates the optimal engineering plan through a reinforcement learning decision engine and optimizes emergency evacuation routes based on an intelligent agent model, forming a "assessment-simulation-optimization" decision-making closed loop.

Images