A geological disaster hazard identification system based on artificial intelligence

Abstract

The present invention discloses a geological hazard hazard identification system based on artificial intelligence, belonging to the field of geological hazard analysis technology, and includes a multi-source data integration module, a regional feature learning module, a hazard identification module, a disaster coupling simulation module, and an emergency strategy generation module. The present invention adopts a hazard identification and disaster coupling simulation method driven by regional feature learning, which improves the accuracy of hazard identification while enhancing the modeling capability of the multi-hazard linkage mechanism in complex geological environments; adopts a lightweight multi-layer perception method enhanced by geological unit division for regional feature learning, which improves the expressive power of feature learning and the discrimination between geological units; adopts a graph convolutional network improved by density peak clustering for hazard identification, and adopts a graph convolutional neural network combined with an improved geological response coupling index for disaster coupling simulation, which improves the simulation credibility of the geological hazard disaster mechanism and the modeling accuracy of the regional risk transmission path.

Description

Technical Field

[0001] The present invention belongs to the technical field of geological disaster analysis, and specifically refers to a geological disaster hidden danger identification system based on artificial intelligence. Background Art

[0002] The artificial intelligence-based geological disaster risk identification system integrates remote sensing perception, multi-source data fusion and artificial intelligence technology to realize the intelligent identification of potential geological disaster risk areas. It aims to improve the timeliness and accuracy of geological disaster warnings, assist relevant departments in scientifically formulating disaster prevention measures, and improve the response speed of disaster prevention and control, effectively reducing geological disaster losses.

[0003] However, in the process of identifying geological hazard risks, there is a technical problem that the recognition accuracy is insufficient due to the influence of terrain differences and the complexity of geological background, which in turn leads to blurred boundaries of hazard areas; in the process of analyzing the characteristics of geological hazard risks, there is a technical problem that traditional methods roughly divide geological units, resulting in insufficient unit expression dimensions and weak spatial detail perception ability; in the process of identifying geological hazard risks, there is a technical problem that common graph neural methods lack the ability to model complex coupling relationships, making it difficult to effectively distinguish the response differences between similar units. Summary of the Invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a geological hazard hazard identification system based on artificial intelligence. In view of the technical problem that in the process of geological hazard hazard identification, the system is easily affected by terrain differences and geological background complexity, resulting in insufficient recognition accuracy, and then leading to blurred boundaries of hazard hazard areas, this solution creatively adopts a hazard identification and disaster coupling simulation method driven by regional feature learning, which can deeply explore the multi-source feature differences and spatial correlations between the internal and adjacent geological units of the region, and at the same time improve the hazard identification accuracy, enhance the modeling ability of the multi-hazard linkage mechanism under complex geological environments; in view of the technical problem that in the process of geological hazard hazard feature analysis, the traditional method has a rough division of geological units, resulting in insufficient unit expression dimensions and weak spatial detail perception ability, this solution creatively adopts a lightweight multi-layer perception method combined with geological unit division enhancement to conduct regional feature learning, by introducing Fine-scale geomorphic zoning information and hierarchical structure recognition mechanism, multi-scale integrated expression of geological attribute factors such as lithologic differences, fault structures, and slope aspect and gradient changes in the region, improve the expression ability of feature learning and the discrimination between geological units, so that the identification of geological hazard risks is no longer limited to rough terrain indicators, but is more targeted and accurate; in the process of identifying geological hazard risks, there are common graph neural methods that have insufficient ability to model complex coupling relationships and have difficulty in effectively distinguishing response differences between similar geological units. This scheme creatively uses a graph convolutional network improved by density peak clustering for hazard identification, and uses a graph convolutional neural network combined with an improved geological response coupling index to perform disaster coupling simulation, so that the disaster coupling simulation process can more realistically reflect the inherent linkage and differential response mechanism between geological units, thereby improving the simulation credibility of geological hazard mechanisms and the modeling accuracy of regional risk transmission paths.

[0005] The technical solution adopted by the present invention is as follows: the present invention provides an artificial intelligence-based geological disaster hazard identification system, comprising: a multi-source data integration module, a regional feature learning module, a hazard identification module, a disaster coupling simulation module and an emergency strategy generation module;

[0006] The multi-source data integration module is used for multi-source data integration, obtains multi-source geological raster data through multi-source data integration, and sends the multi-source geological raster data to the regional feature learning module and the hidden danger identification module;

[0007] The regional feature learning module is used for regional feature learning. It uses a lightweight multi-layer perception method combined with geological unit division enhancement to perform regional feature learning based on multi-source geological raster data to obtain geological hazard feature information, and sends the geological hazard feature information to the hazard identification module and the disaster coupling simulation module;

[0008] The hidden danger identification module is used to identify hidden dangers. Based on multi-source geological raster data and geological hidden danger feature information, it uses a graph convolutional network improved with density peak clustering to identify hidden dangers, obtain geological hidden danger risk information, and send the geological hidden danger risk information to the disaster coupling simulation module and the emergency strategy generation module;

[0009] The disaster coupling simulation module is used for disaster coupling simulation. Based on geological hazard characteristic information and geological hazard risk information, a graph convolutional neural network combined with an improved geological response coupling index is used to perform disaster coupling simulation, obtain disaster coupling simulation reference information, and send the disaster coupling simulation reference information to the emergency strategy generation module.

[0010] The emergency strategy generation module is used for emergency strategy generation, and obtains emergency strategy support reference information through emergency strategy generation.

[0011] Furthermore, the multi-source data integration includes the following steps: multi-source data acquisition, data conversion, geological text data collection and multi-source raster data construction;

[0012] The multi-source data collection specifically involves collecting remote sensing image data, geological survey data, terrain elevation data, meteorological and hydrological data, and engineering activity record data to obtain original geological data;

[0013] The data conversion is specifically to obtain geological raster data by performing coordinate registration, semantic rasterization, spatial resampling and channel coding operations on the original geological data;

[0014] The geological text data collection specifically involves collecting geological survey reports and geological knowledge bases, and performing text preprocessing to obtain geological text data;

[0015] The multi-source raster data construction is specifically to introduce spatial semantic information in geological text data on the basis of geological raster data, structure the text information in the geological text data through geological entity extraction, spatial position association and semantic variable rasterization, and map it to the corresponding spatial position to obtain multi-source geological raster data.

[0016] Furthermore, the regional feature learning is specifically based on multi-source geological raster data, using a lightweight multi-layer perception method combined with geological unit division enhancement to perform regional feature learning to obtain geological hazard feature information, including the following steps: geological unit division, geological unit boundary enhancement, spatial coupling enhancement, feature mapping, sensitivity scoring, and generating regional feature learning results;

[0017] The geological unit division is specifically to calculate the terrain gradient of each grid point based on multi-source geological grid data, and automatically identify the geological unit boundaries through the watershed algorithm, perform geological unit division, and obtain a geological unit boundary map;

[0018] The geological unit boundary enhancement specifically comprises extracting the geological attribute factors of each geological unit from the geological unit boundary map, calculating the boundary consistency index between each pair of adjacent geological units, fusing the boundaries of each pair of geological units whose boundary consistency index is less than a set threshold, merging and generating a new geological unit, and obtaining a geological unit boundary enhancement map; extracting the geological attribute factors of each geological unit from the geological unit boundary enhancement map, performing numerical coding on them, and obtaining geological unit enhancement features;

[0019] The geological attribute factors include slope, aspect, stratum lithology, fault distance, vegetation coverage and landform type index;

[0020] The spatial coupling enhancement is specifically performed by normalizing the geological unit enhancement features to obtain the geological unit standard features, and by encoding the geographical location of each geological unit center point in the geological unit boundary enhancement map to obtain the geographical location features, and then splicing the geological unit standard features and the geographical location features to obtain the spatial coupling enhancement features;

[0021] The feature mapping is specifically to introduce geologically inspired residual connections and sparse activation control mechanisms into a lightweight multilayer perceptron to construct an improved lightweight multilayer perceptron; through the improved lightweight multilayer perceptron, feature mapping is performed on the spatial coupling enhancement features to obtain geological semantic coding features;

[0022] The sensitivity score is used to estimate the catastrophic sensitivity of geological units, specifically by calculating the sensitivity score of geological semantic coding features through linear mapping and converting the sensitivity score into a sensitivity level;

[0023] The generated regional feature learning results are specifically obtained by the geological unit division, the geological unit boundary enhancement, the spatial coupling enhancement, the feature mapping and the sensitivity scoring to obtain geological hazard feature information, wherein the geological hazard feature information includes geological semantic coding features, sensitivity scores and sensitivity levels.

[0024] Furthermore, the hidden danger identification is used to identify geological hazard hidden dangers. Specifically, based on multi-source geological raster data and geological hazard characteristic information, a graph convolutional network improved with density peak clustering is used to identify hidden dangers to obtain geological hazard risk information. The method includes the following steps: risk sequence construction, mutation factor extraction, hidden danger evolution modeling, and hidden danger level classification.

[0025] The risk sequence construction specifically involves extracting time series dynamic factors from multi-source geological raster data, constructing multi-dimensional risk features based on the time series dynamic factors and geological hazard characteristic information, and constructing a risk sequence matrix;

[0026] The time series dynamic factors include rainfall, surface displacement, groundwater level, soil moisture content and engineering disturbance intensity index;

[0027] The mutation factor extraction is used to extract mutation factors that have a leading effect on the evolution of hidden dangers. Specifically, a mutation sensitivity index is constructed; a factor similarity graph is constructed based on the similarity between temporal dynamic factors; factor embedding features are extracted by performing graph convolution on the factor similarity graph, and then the factor embedding density is calculated based on the factor embedding features; an improved density peak clustering scoring function is constructed by combining the factor embedding density and the mutation sensitivity index, and a joint modeling score is generated. The N temporal dynamic factors with the highest joint modeling scores are selected as mutation factors;

[0028] The nodes of the factor similarity graph are used to represent time series dynamic factors;

[0029] The edges of the factor similarity graph are used to represent the similarity between temporal dynamic factors;

[0030] The mutation sensitivity index is used to indicate the degree of fluctuation of the time series dynamic factor in the time series;

[0031] The factor embedding density is used to represent the degree of aggregation of temporal dynamic factors in the feature space;

[0032] The improved density peak clustering scoring function is used to measure the leading role of temporal dynamic factors in the evolution of hidden dangers;

[0033] The hidden danger evolution modeling is used to construct a joint graph structure of the geological unit spatial adjacency graph and the factor similarity graph, and to mine the hidden danger evolution characteristics. Specifically, the geological adjacency graph is constructed by combining the geological unit spatial proximity and the mutation factor similarity. By constructing a time-series graph convolutional network, the geological adjacency graph is subjected to feature aggregation to obtain the hidden danger evolution characteristics.

[0034] The nodes of the geological adjacency graph are used to represent geological units;

[0035] The edges of the geological adjacency graph are used to represent the strength of association between geological units;

[0036] The strength of association between the geological units is used to represent the spatial proximity of the geological units and the similarity of the mutation factors;

[0037] The hazard level classification is used to predict the future risk level and evolution trend of geological units. Specifically, based on the hazard evolution characteristics, a fully connected classification network is constructed to perform hazard level classification to obtain geological hazard risk information. The geological hazard risk information specifically refers to the geological hazard risk level.

[0038] Furthermore, the disaster coupling simulation is specifically based on geological hazard characteristic information and geological hazard risk information, using a graph convolutional neural network combined with an improved geological response coupling index to perform disaster coupling simulation to obtain disaster coupling simulation reference information, including the following steps: coupling graph construction, geological response coupling index construction, graph convolution disaster propagation simulation and disaster coupling risk assessment;

[0039] The coupling map is constructed by adopting a standard graph structure construction method based on the spatial proximity of geological units and geological semantic similarity to construct a disaster coupling map to obtain a disaster coupling map;

[0040] The nodes of the hazard coupling diagram are used to represent geological units;

[0041] The edges of the hazard coupling graph are used to represent the spatial proximity and geological semantic similarity of geological units;

[0042] A weighted adjacency matrix of the hazard coupling graph is used to represent a geological response coupling index;

[0043] The geological response coupling index is constructed by introducing geological structure coupling terms and hidden danger risk response coupling terms to construct the coupling index of geological units, and using the geological response index as a weighted adjacency matrix of the disaster coupling graph;

[0044] The graph convolution disaster propagation simulation is specifically based on the disaster coupling graph, by constructing a standard graph convolution network, performing graph convolution disaster propagation simulation, and obtaining disaster coupling feature data;

[0045] The disaster coupling risk assessment is used to evaluate the probability of hidden dangers becoming disasters. Specifically, based on disaster coupling characteristic data, a disaster coupling risk assessment is performed by introducing standard linear mapping and softmax classifier function to obtain disaster coupling simulation reference data. The disaster coupling simulation reference data specifically refers to the probability of hidden dangers becoming disasters in geological units.

[0046] Furthermore, the emergency strategy is generated specifically by generating an emergency response strategy based on the geological hazard risk information and the disaster coupling simulation reference data through a rule-matching method, and obtaining emergency strategy support reference information by constructing an early warning trigger threshold mechanism.

[0047] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0048] (1) In order to address the technical problem that the identification of geological hazard hazards is easily affected by terrain differences and the complexity of the geological background, resulting in insufficient identification accuracy and blurred boundaries of the hazard hazard area, this solution creatively adopts a hazard identification and disaster coupling simulation method driven by regional feature learning. This method can deeply explore the multi-source feature differences and spatial correlations between the region and the adjacent geological units. While improving the hazard identification accuracy, it also enhances the modeling ability of the multi-hazard linkage mechanism in complex geological environments.

[0049] (2) In order to solve the technical problem that the traditional method of geological unit division is rough in the process of geological hazard feature analysis, resulting in insufficient unit expression dimension and weak spatial detail perception ability, this solution creatively adopts a lightweight multi-layer perception method combined with geological unit division enhancement to carry out regional feature learning. By introducing fine-scale landform zoning information and hierarchical structure recognition mechanism, the geological attribute factors such as lithology difference, fault structure, slope aspect and gradient change in the region are integrated and expressed at multiple scales, which improves the expression ability of feature learning and the discrimination between geological units. As a result, the identification of geological hazard risks is no longer limited to rough terrain indicators, but is more targeted and accurate.

[0050] (3) In view of the technical problems in the process of identifying geological hazard risks, the common graph neural network methods have insufficient ability to model complex coupling relationships and are difficult to effectively distinguish the response differences between similar geological units. This scheme creatively uses a graph convolutional network improved by density peak clustering for hazard identification, and uses a graph convolutional neural network combined with an improved geological response coupling index to perform disaster coupling simulation, so that the disaster coupling simulation process can more realistically reflect the inherent linkage and differential response mechanism between geological units, thereby improving the simulation credibility of the geological hazard mechanism and the modeling accuracy of the regional risk transmission path. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A schematic diagram of the structure of a geological disaster hazard identification system based on artificial intelligence provided by the present invention;

[0052] Figure 2 This is a flowchart of the multi-source data integration module;

[0053] Figure 3 This is a flowchart of the regional feature learning module;

[0054] Figure 4 This is a flowchart of the hidden danger identification module;

[0055] Figure 5 This is a flow chart of the disaster coupling simulation module.

[0056] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0058] In the description of the present invention, it should be understood that terms such as "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0059] Example 1, see Figure 1 The present invention provides an artificial intelligence-based geological disaster hazard identification system, which includes a multi-source data integration module, a regional feature learning module, a hazard identification module, a disaster coupling simulation module, and an emergency strategy generation module;

[0060] The multi-source data integration module is used for multi-source data integration, obtains multi-source geological raster data through multi-source data integration, and sends the multi-source geological raster data to the regional feature learning module and the hidden danger identification module;

[0061] The regional feature learning module is used for regional feature learning. It uses a lightweight multi-layer perception method combined with geological unit division enhancement to perform regional feature learning based on multi-source geological raster data to obtain geological hazard feature information, and sends the geological hazard feature information to the hazard identification module and the disaster coupling simulation module;

[0062] The hidden danger identification module is used to identify hidden dangers. Based on multi-source geological raster data and geological hidden danger feature information, it uses a graph convolutional network improved with density peak clustering to identify hidden dangers, obtain geological hidden danger risk information, and send the geological hidden danger risk information to the disaster coupling simulation module and the emergency strategy generation module;

[0063] The disaster coupling simulation module is used for disaster coupling simulation. Based on geological hazard characteristic information and geological hazard risk information, a graph convolutional neural network combined with an improved geological response coupling index is used to perform disaster coupling simulation, obtain disaster coupling simulation reference information, and send the disaster coupling simulation reference information to the emergency strategy generation module.

[0064] The emergency strategy generation module is used to generate emergency strategies and obtain emergency strategy support reference information through emergency strategy generation;

[0065] By performing the above operations, in order to address the technical problem that in the process of geological hazard identification, there is a tendency to be affected by terrain differences and the complexity of the geological background, resulting in insufficient identification accuracy and thus blurred boundaries of the hazard area, this solution creatively adopts a hazard identification and disaster coupling simulation method driven by regional feature learning. It can deeply explore the multi-source feature differences and spatial correlations between the internal region and the adjacent geological units, while improving the hazard identification accuracy, and enhancing the modeling ability of the multi-hazard linkage mechanism in complex geological environments.

[0066] Example 2, see Figure 1 and Figure 2 ,This embodiment is based on the above embodiment, and the multi-source data integration ,includes the following steps: multi-source data acquisition, data conversion, ,geological text data collection and multi-source raster data construction;

[0067] The multi-source data collection specifically involves collecting remote sensing image data, geological survey data, terrain elevation data, meteorological and hydrological data, and engineering activity record data to obtain original geological data;

[0068] The data conversion is specifically to obtain geological raster data by performing coordinate registration, semantic rasterization, spatial resampling and channel coding operations on the original geological data;

[0069] The coordinate registration is specifically to project all data in the original geological data into the same geographic coordinate system to ensure the consistency of multi-source data in spatial dimension;

[0070] The semantic rasterization is specifically to convert vector data in the original geological data into regular raster data for retaining spatial semantic features;

[0071] The spatial resampling is specifically to resample the different resolution data in the original geological data into a unified regular grid;

[0072] The channel encoding is specifically encoding different types of data in the original geological data into a unified multi-channel raster tensor;

[0073] The geological text data collection specifically involves collecting geological survey reports and geological knowledge bases, and performing text preprocessing to obtain geological text data;

[0074] The text preprocessing includes word segmentation, irrelevant information filtering and stop word removal;

[0075] The multi-source raster data construction is specifically to introduce spatial semantic information in geological text data on the basis of geological raster data, structure the text information in the geological text data through geological entity extraction, spatial position association and semantic variable rasterization, and map it to the corresponding spatial position to obtain multi-source geological raster data.

[0076] Example 3, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. The regional feature learning is specifically based on multi-source geological raster data and adopts a lightweight multi-layer perception method combined with geological unit division enhancement to perform regional feature learning to obtain geological hazard feature information, including the following steps: geological unit division, geological unit boundary enhancement, spatial coupling enhancement, feature mapping, sensitivity scoring, and generating regional feature learning results;

[0077] The geological unit division is specifically to calculate the terrain gradient of each grid point based on multi-source geological grid data, and automatically identify the geological unit boundaries through the watershed algorithm, perform geological unit division, and obtain a geological unit boundary map;

[0078] The calculation formula for calculating the terrain gradient of each grid point is:

[0079] ;

[0080] Where G(x,y) is the terrain gradient at the grid point (x,y), which is used to represent the slope intensity at the grid point (x,y), x is the horizontal coordinate index, y is the vertical coordinate index, and h is the elevation function, which is used to represent the surface height value at the two-dimensional space (x,y). It is the partial derivative of the elevation function along the x-axis, which is used to express the rate at which the surface height changes with the horizontal direction. It is the partial derivative of the elevation function along the y-axis, which is used to express the rate at which the surface height changes in the vertical direction;

[0081] The geological unit boundary enhancement specifically comprises extracting the geological attribute factors of each geological unit from the geological unit boundary map, calculating the boundary consistency index between each pair of adjacent geological units, fusing the boundaries of each pair of geological units whose boundary consistency index is less than a set threshold, merging and generating a new geological unit, and obtaining a geological unit boundary enhancement map; extracting the geological attribute factors of each geological unit from the geological unit boundary enhancement map, performing numerical coding on them, and obtaining geological unit enhancement features;

[0082] The geological attribute factors include slope, aspect, stratum lithology, fault distance, vegetation coverage and landform type index;

[0083] The calculation formula of the boundary consistency index is:

[0084] ;

[0085] Where C ij is the boundary consistency index between the i-th geological unit and the j-th geological unit, d is the total number of geological attribute factors, D is the geological attribute factor index, R i is the ith geological unit, R j is the jth geological unit, i is the first index of the geological unit, j is the second index of the geological unit, the second index of the geological unit is not equal to the first index of the geological unit, F d (R i ) is the value of the dth geological attribute factor on the i-th geological unit, F d (R j ) is the value of the dth geological attribute factor in the jth geological unit;

[0086] The spatial coupling enhancement is specifically performed by normalizing the geological unit enhancement features to obtain the geological unit standard features, and by encoding the geographical location of each geological unit center point in the geological unit boundary enhancement map to obtain the geographical location features. Then, the geological unit standard features and the geographical location features are spliced ​​to obtain the spatial coupling enhancement features. The calculation formula is:

[0087] ;

[0088] Where, is the spatial coupling enhancement feature of the i-th geological unit, Concat(·) is the splicing operation, Nor(·) is the normalization process, X i is the geological unit enhancement feature of the i-th geological unit, PE(·) is the geographic location code, lon i is the latitude of the center point of the i-th geological unit, lat i is the longitude of the center point of the i-th geological unit, elev iis the elevation of the center point of the i-th geological unit;

[0089] The feature mapping is specifically to introduce geologically inspired residual connections and sparse activation control mechanisms into a lightweight multilayer perceptron to construct an improved lightweight multilayer perceptron; through the improved lightweight multilayer perceptron, feature mapping is performed on the spatial coupling enhancement features to obtain geological semantic coding features, and the calculation formula is:

[0090] ;

[0091] ;

[0092] Where, It is the output feature of the 0th layer, specifically the output feature of the 0th layer corresponding to the i-th geological unit in the improved lightweight multilayer perceptron. is the output feature of the Lth layer, Sig(·) is the Sigmoid activation function, and W (L) is the weight of the Lth layer, is the output feature of the L-1 layer, B (L) is the bias term of the Lth layer, is a control parameter used to adjust the influence of the geological heuristic residual term, Mask(·) is a sparse mask function used to selectively activate key features;

[0093] The sensitivity score is used to estimate the catastrophic sensitivity of geological units, specifically by calculating the sensitivity score of geological semantic coding features through linear mapping and converting the sensitivity score into a sensitivity level;

[0094] Generating regional feature learning results, specifically obtaining geological hazard feature information through the geological unit division, the geological unit boundary enhancement, the spatial coupling enhancement, the feature mapping and the sensitivity score, wherein the geological hazard feature information includes geological semantic coding features, sensitivity scores and sensitivity levels;

[0095] By performing the above operations, in order to address the technical problem that in the process of geological hazard characteristic analysis, traditional methods roughly divide geological units, resulting in insufficient unit expression dimensions and weak spatial detail perception, this solution creatively adopts a lightweight multi-layer perception method combined with geological unit division enhancement to conduct regional feature learning. By introducing fine-scale landform zoning information and hierarchical structure recognition mechanism, multi-scale integrated expression of geological attribute factors such as lithology differences, fault structures, and slope aspect and gradient changes in the region is performed, which improves the expression ability of feature learning and the discrimination between geological units, making the identification of geological hazard risks no longer limited to rough terrain indicators, but more targeted and accurate.

[0096] Example 4, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. The hidden danger identification is used to identify geological hazard hidden dangers. Specifically, based on multi-source geological raster data and geological hazard feature information, a graph convolutional network improved by density peak clustering is used to identify hidden dangers to obtain geological hazard risk information. The method includes the following steps: risk sequence construction, mutation factor extraction, hidden danger evolution modeling, and hidden danger level classification.

[0097] The risk sequence construction specifically involves extracting time series dynamic factors from multi-source geological raster data, constructing multi-dimensional risk features based on the time series dynamic factors and geological hazard characteristic information, and constructing a risk sequence matrix;

[0098] The time series dynamic factors include rainfall, surface displacement, groundwater level, soil moisture content and engineering disturbance intensity index;

[0099] The calculation formula of the multidimensional risk characteristics is:

[0100] ;

[0101] Where, is the multidimensional risk characteristic of the ith geological unit at the tth time step, t is the time step index, is the rainfall of the ith geological unit at the tth time step, is the surface displacement of the ith geological unit at the tth time step, is the groundwater level of the ith geological unit at time step t, is the soil moisture content of the ith geological unit at the tth time step, is the engineering disturbance intensity index of the i-th geological unit at the t-th time step, geo i is the geological semantic coding feature of the i-th geological unit, sco i is the sensitivity score of the ith geological unit, lev i is the sensitivity level of the ith geological unit;

[0102] The calculation formula of the multidimensional risk feature sequence matrix is:

[0103] ;

[0104] Where S i is the risk sequence matrix of the ith geological unit, is the multidimensional risk characteristic of the ith geological unit at the first time step, is the multidimensional risk characteristic of the ith geological unit at the second time step, is the multidimensional risk characteristic of the ith geological unit at the Tth time step, where T is the maximum time step;

[0105] The mutation factor extraction is used to extract mutation factors that have a leading effect on the evolution of hidden dangers. Specifically, a mutation sensitivity index is constructed; a factor similarity graph is constructed based on the similarity between temporal dynamic factors; factor embedding features are extracted by performing graph convolution on the factor similarity graph, and then the factor embedding density is calculated based on the factor embedding features; an improved density peak clustering scoring function is constructed by combining the factor embedding density and the mutation sensitivity index, and a joint modeling score is generated. The N temporal dynamic factors with the highest joint modeling scores are selected as mutation factors;

[0106] The nodes of the factor similarity graph are used to represent time series dynamic factors;

[0107] The edges of the factor similarity graph are used to represent the similarity between temporal dynamic factors;

[0108] The mutation sensitivity index is used to indicate the degree of fluctuation of the time series dynamic factor in the time series. The calculation formula is:

[0109] ;

[0110] Where, is the mutation sensitivity of the kth temporal dynamic factor, k is the first index of the temporal dynamic factor, f k (t) is the value of the kth time series dynamic factor at the tth time step, f k (t-1) is the value of the kth time series dynamic factor at the t-1th time step, is a minimum constant term used to avoid the denominator being zero;

[0111] The calculation formula for the similarity between the temporal dynamic factors is:

[0112] ;

[0113] Where w k,b is the similarity between the kth time series dynamic factor and the bth time series dynamic factor, b is the second index of the time series dynamic factor, the second index of the time series dynamic factor is not equal to the first index of the time series dynamic factor, sim(·) is the Gaussian kernel similarity function, exp(·) is the exponential function, ||·||2 is the L2 norm operator, f k is the complete time series vector of the kth time series dynamic factor, f b is the complete time series vector of the bth time series dynamic factor, It is a decay adjustment item, used to adjust the decay speed of similarity;

[0114] The factor embedding density is used to indicate the degree of aggregation of temporal dynamic factors in the feature space, and the calculation formula is:

[0115] ;

[0116] Where, is the factor embedding density corresponding to the kth temporal dynamic factor, K is the number of temporal dynamic factors, e k is the factor embedding feature corresponding to the kth temporal dynamic factor, e b is the factor embedding feature corresponding to the b-th temporal dynamic factor;

[0117] The improved density peak clustering scoring function is used to measure the leading role of time series dynamic factors in the evolution of hidden dangers. The calculation formula is:

[0118] ;

[0119] In the formula, score k is the joint modeling score of the kth temporal dynamic factor. The higher the joint modeling score, the greater the probability that the corresponding temporal dynamic factor will promote the evolution of hidden dangers. is the sensitivity weight, which is used to control the influence of mutation sensitivity on the joint modeling score;

[0120] The hidden danger evolution modeling is used to construct a joint graph structure of the geological unit spatial adjacency graph and the factor similarity graph, and to mine the hidden danger evolution characteristics. Specifically, the geological adjacency graph is constructed by combining the geological unit spatial proximity and the mutation factor similarity. By constructing a time-series graph convolutional network, the geological adjacency graph is subjected to feature aggregation to obtain the hidden danger evolution characteristics.

[0121] The nodes of the geological adjacency graph are used to represent geological units;

[0122] The edges of the geological adjacency graph are used to represent the strength of association between geological units;

[0123] The correlation strength between the geological units is used to represent the spatial proximity of the geological units and the similarity of the mutation factors. The calculation formula is:

[0124] ;

[0125] Where A ij is the correlation strength between the i-th geological unit and the j-th geological unit, dis ij is the spatial distance between the i-th geological unit and the j-th geological unit, is the distance weight, which is used to control the influence of spatial distance on the association strength. i is the temporal characteristics of the mutation factor of the i-th geological unit, mut j is the temporal characteristics of the mutation factor of the jth geological unit, Used to measure the similarity of the mutation factors between the i-th geological unit and the j-th geological unit, is the similarity weight, which is used to control the influence of mutation factor similarity on the association strength;

[0126] The hazard level classification is used to predict the future risk level and evolution trend of geological units. Specifically, based on the hazard evolution characteristics, a fully connected classification network is constructed to perform hazard level classification to obtain geological hazard risk information. The geological hazard risk information specifically refers to the geological hazard risk level.

[0127] Example 5, see Figure 1 and Figure 5 This embodiment is based on the above embodiment. The disaster coupling simulation is specifically based on geological hazard characteristic information and geological hazard risk information, and adopts a graph convolutional neural network combined with an improved geological response coupling index to perform disaster coupling simulation to obtain disaster coupling simulation reference information, including the following steps: coupling graph construction, geological response coupling index construction, graph convolution disaster propagation simulation and disaster coupling risk assessment;

[0128] The coupling map is constructed by adopting a standard graph structure construction method based on the spatial proximity of geological units and geological semantic similarity to construct a disaster coupling map to obtain a disaster coupling map;

[0129] The nodes of the hazard coupling diagram are used to represent geological units;

[0130] The edges of the hazard coupling graph are used to represent the spatial proximity and geological semantic similarity of geological units;

[0131] A weighted adjacency matrix of the hazard coupling graph is used to represent a geological response coupling index;

[0132] The geological response coupling index is constructed by introducing geological structure coupling terms and hidden danger risk response coupling terms to construct the coupling index of geological units, and using the geological response index as a weighted adjacency matrix of the disaster coupling graph;

[0133] The calculation formula of the geological response coupling index is:

[0134] ;

[0135] Where R mn is the geological response coupling index, is the geological semantic similarity weight, The whole is a geological structure coupling term, which is used to express geological semantic similarity. cos (·) is the cosine similarity calculation function, geo m is the geological semantic coding feature of the mth geological unit node, geo nis the geological semantic coding feature of the nth geological unit neighbor node, m is the geological unit node index, which is used to represent the node of the hazard coupling graph, and n is the geological unit neighbor node index. is the risk level weight of geological unit hidden danger, is the hidden danger risk response coupling term, which is used to express the hidden danger risk of the geological unit, r m is the geological hazard risk level of the mth geological unit node, r n is the geological hazard risk level of the neighboring node of the nth geological unit;

[0136] The graph convolution disaster propagation simulation is specifically based on the disaster coupling graph, by constructing a standard graph convolution network, performing graph convolution disaster propagation simulation, and obtaining disaster coupling feature data;

[0137] The disaster coupling risk assessment is used to assess the probability of hidden dangers becoming disasters. Specifically, based on the disaster coupling characteristic data, the disaster coupling risk assessment is performed by introducing a standard linear mapping and a softmax classifier function to obtain disaster coupling simulation reference data. The disaster coupling simulation reference data specifically refers to the probability of hidden dangers becoming disasters in geological units.

[0138] By performing the above operations, in order to address the technical problems in the process of geological hazard identification, that is, the common graph neural network methods have insufficient ability to model complex coupling relationships and are difficult to effectively distinguish the response differences between similar geological units, this solution creatively adopts a graph convolutional network improved with density peak clustering for hazard identification, and adopts a graph convolutional neural network combined with an improved geological response coupling index for disaster coupling simulation, so that the disaster coupling simulation process can more realistically reflect the inherent linkage and differential response mechanism between geological units, thereby improving the simulation credibility of the geological hazard mechanism and the modeling accuracy of the regional risk transmission path.

[0139] Example 6, see Figure 1 This embodiment is based on the above embodiment. Specifically, the emergency strategy is generated by generating an emergency response strategy based on the geological hazard risk information and the disaster coupling simulation reference data through a rule-matching method, and obtaining emergency strategy support reference information by constructing an early warning trigger threshold mechanism.

[0140] The emergency strategy support reference information includes risk area classification data, emergency resource deployment reference data and emergency strategy comprehensive report data.

[0141] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0142] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

[0143] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. A geological hazard identification system based on artificial intelligence, characterized by: It includes multi-source data integration module, regional feature learning module, hidden danger identification module, disaster coupling simulation module and emergency strategy generation module; The multi-source data integration module is used for multi-source data integration, obtains multi-source geological raster data through multi-source data integration, and sends the multi-source geological raster data to the regional feature learning module and the hidden danger identification module; The regional feature learning module is used for regional feature learning. It uses a lightweight multi-layer perception method combined with geological unit division enhancement to perform regional feature learning based on multi-source geological raster data to obtain geological hazard feature information, and sends the geological hazard feature information to the hazard identification module and the disaster coupling simulation module; The hidden danger identification module is used to identify hidden dangers. Based on multi-source geological raster data and geological hidden danger feature information, it uses a graph convolutional network improved with density peak clustering to identify hidden dangers, obtain geological hidden danger risk information, and send the geological hidden danger risk information to the disaster coupling simulation module and the emergency strategy generation module; The disaster coupling simulation module is used for disaster coupling simulation. Based on geological hazard characteristic information and geological hazard risk information, a graph convolutional neural network combined with an improved geological response coupling index is used to perform disaster coupling simulation, obtain disaster coupling simulation reference information, and send the disaster coupling simulation reference information to the emergency strategy generation module. The disaster coupling simulation includes the following steps: coupling graph construction, geological response coupling index construction, graph convolution disaster propagation simulation and disaster coupling risk assessment; The coupling map is constructed by adopting a standard graph structure construction method based on the spatial proximity of geological units and geological semantic similarity to construct a disaster coupling map to obtain a disaster coupling map; The nodes of the hazard coupling diagram are used to represent geological units; The edges of the hazard coupling graph are used to represent the spatial proximity and geological semantic similarity of geological units; A weighted adjacency matrix of the hazard coupling graph is used to represent a geological response coupling index; The geological response coupling index is constructed by introducing geological structure coupling terms and hidden danger risk response coupling terms to construct the coupling index of geological units, and using the geological response index as a weighted adjacency matrix of the disaster coupling graph; The graph convolution disaster propagation simulation is specifically based on the disaster coupling graph, by constructing a standard graph convolution network, performing graph convolution disaster propagation simulation, and obtaining disaster coupling feature data; The disaster coupling risk assessment is used to assess the probability of hidden dangers becoming disasters. Specifically, based on the disaster coupling characteristic data, the disaster coupling risk assessment is performed by introducing a standard linear mapping and a softmax classifier function to obtain disaster coupling simulation reference data. The disaster coupling simulation reference data specifically refers to the probability of hidden dangers becoming disasters in geological units. The emergency strategy generation module is used for emergency strategy generation, and obtains emergency strategy support reference information through emergency strategy generation.

2. The artificial intelligence-based geological disaster hazard identification system according to claim 1, characterized in that: The regional feature learning is specifically based on multi-source geological raster data, using a lightweight multi-layer perception method combined with geological unit division enhancement to perform regional feature learning to obtain geological hazard feature information, including the following steps: geological unit division, geological unit boundary enhancement, spatial coupling enhancement, feature mapping, sensitivity scoring, and generating regional feature learning results; The geological unit division is specifically to calculate the terrain gradient of each grid point based on multi-source geological grid data, and automatically identify the geological unit boundaries through the watershed algorithm, perform geological unit division, and obtain a geological unit boundary map; The geological unit boundary enhancement specifically comprises extracting the geological attribute factors of each geological unit from the geological unit boundary map, calculating the boundary consistency index between each pair of adjacent geological units, fusing the boundaries of each pair of geological units whose boundary consistency index is less than a set threshold, merging and generating a new geological unit, and obtaining a geological unit boundary enhancement map; extracting the geological attribute factors of each geological unit from the geological unit boundary enhancement map, performing numerical coding on them, and obtaining geological unit enhancement features; The geological attribute factors include slope, aspect, stratum lithology, fault distance, vegetation coverage and landform type index; The spatial coupling enhancement is specifically performed by normalizing the geological unit enhancement features to obtain the geological unit standard features, and by encoding the geographical location of each geological unit center point in the geological unit boundary enhancement map to obtain the geographical location features, and then splicing the geological unit standard features and the geographical location features to obtain the spatial coupling enhancement features; The feature mapping is specifically to introduce geologically inspired residual connections and sparse activation control mechanisms into a lightweight multilayer perceptron to construct an improved lightweight multilayer perceptron; through the improved lightweight multilayer perceptron, feature mapping is performed on the spatial coupling enhancement features to obtain geological semantic coding features; The sensitivity score is used to estimate the catastrophic sensitivity of geological units, specifically by calculating the sensitivity score of geological semantic coding features through linear mapping and converting the sensitivity score into a sensitivity level; The generated regional feature learning results are specifically obtained by the geological unit division, the geological unit boundary enhancement, the spatial coupling enhancement, the feature mapping and the sensitivity scoring to obtain geological hazard feature information, wherein the geological hazard feature information includes geological semantic coding features, sensitivity scores and sensitivity levels.

3. The artificial intelligence-based geological disaster hazard identification system according to claim 2, characterized in that: The hidden danger identification is used to identify geological hazard hidden dangers. Specifically, based on multi-source geological raster data and geological hazard feature information, a graph convolutional network improved with density peak clustering is used to identify hidden dangers to obtain geological hazard risk information. The method includes the following steps: risk sequence construction, mutation factor extraction, hidden danger evolution modeling, and hidden danger level classification. The risk sequence construction specifically involves extracting time series dynamic factors from multi-source geological raster data, constructing multi-dimensional risk features based on the time series dynamic factors and geological hazard characteristic information, and constructing a risk sequence matrix; The time series dynamic factors include rainfall, surface displacement, groundwater level, soil moisture content and engineering disturbance intensity index; The mutation factor extraction is used to extract mutation factors that have a leading effect on the evolution of hidden dangers. Specifically, a mutation sensitivity index is constructed; a factor similarity graph is constructed based on the similarity between temporal dynamic factors; factor embedding features are extracted by performing graph convolution on the factor similarity graph, and then the factor embedding density is calculated based on the factor embedding features; an improved density peak clustering scoring function is constructed by combining the factor embedding density and the mutation sensitivity index, and a joint modeling score is generated. The N temporal dynamic factors with the highest joint modeling scores are selected as mutation factors; The hidden danger evolution modeling is used to construct a joint graph structure of the geological unit spatial adjacency graph and the factor similarity graph, and to mine the hidden danger evolution characteristics. Specifically, the geological adjacency graph is constructed by combining the geological unit spatial proximity and the mutation factor similarity. By constructing a time-series graph convolutional network, the geological adjacency graph is subjected to feature aggregation to obtain the hidden danger evolution characteristics. The hazard level classification is used to predict the future risk level and evolution trend of geological units. Specifically, based on the hazard evolution characteristics, a fully connected classification network is constructed to perform hazard level classification to obtain geological hazard risk information. The geological hazard risk information specifically refers to the geological hazard risk level.

4. The artificial intelligence-based geological disaster hazard identification system according to claim 3, characterized in that: In the mutation factor extraction, the nodes of the factor similarity graph are used to represent temporal dynamic factors; The edges of the factor similarity graph are used to represent the similarity between temporal dynamic factors; The mutation sensitivity index is used to indicate the degree of fluctuation of the time series dynamic factor in the time series; The factor embedding density is used to represent the degree of aggregation of temporal dynamic factors in the feature space; The improved density peak clustering scoring function is used to measure the leading role of time series dynamic factors in the evolution of hidden dangers.

5. The artificial intelligence-based geological disaster hazard identification system according to claim 4, characterized in that: In the hazard evolution modeling, the nodes of the geological adjacency graph are used to represent geological units; The edges of the geological adjacency graph are used to represent the strength of association between geological units; The strength of the association between the geological units is used to represent the spatial proximity of the geological units and the similarity of the mutation factors.

6. The artificial intelligence-based geological disaster hazard identification system according to claim 5, characterized in that: The emergency strategy generation is specifically to generate an emergency response strategy based on the geological hazard risk information and the disaster coupling simulation reference data through a rule-matching method, and obtain emergency strategy support reference information by constructing an early warning trigger threshold mechanism.

7. The artificial intelligence-based geological disaster hazard identification system according to claim 6, characterized in that: The multi-source data integration includes the following steps: multi-source data acquisition, data conversion, geological text data collection and multi-source raster data construction; The multi-source data collection specifically involves collecting remote sensing image data, geological survey data, terrain elevation data, meteorological and hydrological data, and engineering activity record data to obtain original geological data; The data conversion is specifically to obtain geological raster data by performing coordinate registration, semantic rasterization, spatial resampling and channel coding operations on the original geological data; The geological text data collection specifically involves collecting geological survey reports and geological knowledge bases, and performing text preprocessing to obtain geological text data; The multi-source raster data construction is specifically to introduce spatial semantic information in geological text data on the basis of geological raster data, structure the text information in the geological text data through geological entity extraction, spatial position association and semantic variable rasterization, and map it to the corresponding spatial position to obtain multi-source geological raster data.

Images