A natural disaster early warning and tourism safety emergency response system

Abstract

The application discloses a natural disaster early warning and tourism safety emergency response system, comprising: a data layer constructs a space-time disaster causal network, a root node is configured with a disaster characteristic matrix containing an influence radius and a disaster-causing intensity, nodes and edges have geographic coordinates and time delay attributes, each edge has a time lag coefficient, an intermediate node is activated and needs an upstream weighted summation value to be greater than 0.7, a decision layer establishes disaster interaction edges and coupling operators when multiple root nodes are concurrent, forms a dynamic composite causal graph, outputs a nonlinear superposition effect of facility preservation and evacuation efficiency by a graph neural network, calls a real-time crowd simulation engine to generate a pareto optimal response scheme set, and an edge layer deploys the engine and a decision tree model obtained by knowledge distillation. An interaction layer converts the scheme into differentiated instructions and forcibly pushes the differentiated instructions to a guide and a tourist terminal through a three-level permission system, and the instructions are attached with digital signatures.

Description

Technical Field

[0001] This invention relates to the field of safety early warning technology, and in particular to a natural disaster early warning and tourism safety emergency response system. Background Technology

[0002] In the context of frequent natural disasters and the increasing popularity of tourism, the construction of a natural disaster early warning and tourism safety emergency response system is of great background need and practical significance. With the increasing complexity of global climate change and geological activity, various natural disasters such as earthquakes, floods, and typhoons pose an increasingly severe threat to the safety of tourists' lives and property. Traditional early warning and emergency response methods suffer from problems such as delayed information transmission and low response efficiency. This system aims to integrate advanced monitoring technologies, big data analysis, and efficient communication methods to achieve accurate early warning and rapid response to natural disasters. It can not only minimize the impact of natural disasters on the tourism industry and protect the lives of tourists during their travels, but also improve the safety management level and emergency response capabilities of tourist destinations, enhance tourists' trust in tourist destinations, and promote the healthy and sustainable development of the tourism industry.

[0003] In existing technologies, relying on manually defined superposition rules often fails to automatically learn the coupling patterns of concurrent disaster combinations from historical disaster data. Therefore, when unpredictable compound disasters occur, the assessment results will deviate significantly from the risks in the real physical world. Thus, a natural disaster early warning and tourism safety emergency response system is proposed. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a natural disaster early warning and tourism safety emergency response system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A natural disaster early warning and tourism safety emergency response system includes: Data Layer: Constructs a spatiotemporal disaster causal network. The root node of the spatiotemporal disaster causal network includes at least one of typhoon, wildfire, avalanche, and heat wave. Each root node is configured with a disaster feature matrix, which includes the influence radius and disaster intensity. The nodes and edges of the spatiotemporal disaster causal network have geographical coordinates and time delay attributes. Each edge is configured with a time delay coefficient. The activation condition of the intermediate node is that the result of the weighted sum of the disaster feature matrix value of the upstream node after applying a distance decay function and a time delay function is greater than 0.7. The distance decay function is the product function of the inverse of the influence radius in the disaster feature matrix and the distance. The time delay function depends on the time delay coefficient. The data layer embeds a spatiotemporal disaster script tree. When the cosine similarity between the real-time disaster type activation vector and any path in the spatiotemporal disaster script tree is greater than 0.9, the predicted secondary disaster type and the estimated arrival time are output. Decision layer: When two or more root nodes are identified concurrently, disaster interaction edges are established and disaster coupling operators are configured to form a dynamic composite disaster causal graph. The pre-trained graph neural network outputs the nonlinear superposition causal effect on facility preservation and tourist evacuation efficiency. Before starting optimization, the graph edit distance similarity between the current disaster subgraph and the composite disaster mode subgraph in the independently maintained extreme disaster chain long memory buffer is calculated. If it is greater than 0.85, the historical peak coupling probability of the mode is forced to cover the current coupling strength as a constraint condition. A multi-factor evolutionary algorithm is run, and a real-time crowd simulation engine is called to evaluate candidate solutions. The real-time crowd simulation engine outputs the minimum evacuation time and the minimum tourist behavior entropy based on the social force model and the real-time collected tourist density, flow rate and emotion parameters. The positions with tourist behavior entropy values ​​greater than 0.8 are marked as dynamic bottleneck points and a temporary constraint of "node passage status" is generated and added to the search space. The Pareto optimal response solution set is generated with the minimum evacuation time, the maximum material coverage rate and the minimum tourist behavior entropy as objectives. Edge layer: It includes the real-time crowd simulation engine and a decision tree model obtained by knowledge distillation of the Pareto optimal response scheme set. The decision tree model is pre-installed in the disaster information box or scenic area gate. Interaction Layer: The Pareto optimal response scheme set is transformed into differentiated instructions through a three-level role-based permission system and forcibly pushed to the tour guide terminal and the tourist mobile terminal. Each instruction is attached with a digital signature. The tour guide terminal and the tourist mobile terminal must return a confirmation of receipt within 30 seconds. Otherwise, the instruction content is automatically upgraded to a scenic area broadcast.

[0006] The above technical solution further includes: Furthermore, the distance decay function is specifically as follows: ,in, For spatial distance, The influence radius in the catastrophic feature matrix of the upstream node is defined as follows: the time delay function is specifically a step function, which is set to 1 when the actual delay time is greater than or equal to the time delay coefficient of the edge, and 0 otherwise. The weights of the weighted summation include a catastrophic intensity factor of 0.4 and a spatiotemporal coupling factor of 0.6.

[0007] Furthermore, when the decision layer constructs the dynamic composite disaster causal graph, the disaster interaction edge connects concurrent root nodes, the disaster coupling operator is the product normalization function of the intensity of the corresponding time period in the disaster feature matrix of each root node, and the pre-trained graph neural network contains three graph convolutional layers and one fully connected layer. It takes the adjacency matrix of the dynamic composite disaster causal graph and the disaster feature matrix of all nodes as input, and outputs the predicted values ​​of the facility preservation degree and the tourist evacuation efficiency.

[0008] Furthermore, the decision layer implements a dual-track rare event circuit breaker. When implementing the dual-track rare event circuit breaker, the method for extracting the structural features of the current disaster subgraph is as follows: constructing a minimum generating subgraph that only contains the currently activated disaster root node and the disaster interaction edge; calculating the Weisfeiler-Lehman graph kernel feature vector of the minimum generating subgraph; and calculating the graph edit distance similarity based on the normalized edit distance between the minimum generating subgraph and each composite disaster mode subgraph in the extreme disaster chain length memory buffer. When the maximum normalized edit distance similarity is greater than 0.85, the forced coverage is triggered.

[0009] Furthermore, the real-time crowd simulation engine divides the simulation area into a 15m × 15m grid, and counts the tourist density and movement direction unit vector in each grid every second. The tourist behavior entropy is defined as the information entropy of the movement direction in the grid. When the tourist behavior entropy value of a certain grid is greater than 0.8 and lasts for 5 seconds, it is marked as a dynamic bottleneck point and reported to the decision layer.

[0010] Furthermore, the temporary constraints on node traffic status generated by the decision layer include the grid coordinates of the dynamic bottleneck point, the restricted traffic direction, and the suggested detour node sequence. The lifespan of the temporary constraints is 30 seconds. If the visitor behavior entropy of the grid drops below 0.6 within 30 seconds, the constraints are automatically revoked; otherwise, they remain in effect and enter the next round of optimization.

[0011] Furthermore, the spatiotemporal disaster script tree embedded in the data layer module contains no less than 200 pre-deduced propagation paths. Each path consists of a disaster node sequence, a trigger time window, and a spatial propagation hop count. The cosine similarity calculation is performed on the real-time disaster type activation vector and the preset disaster type vector in the path. When the maximum cosine similarity is greater than 0.9, the secondary disaster node in the path that has not yet been activated but is within 30 minutes after the current time is output as the prediction result.

[0012] Furthermore, the decision-making layer includes a plan hot-switching unit, which receives the predicted secondary disaster type and estimated arrival time. When the estimated arrival time is less than 15 minutes, it automatically switches to the on-site evacuation plan. When the estimated arrival time is between 15 and 30 minutes, it automatically switches to the cableway evacuation plan and forcibly pushes the switched instruction to all terminals through the interaction layer within 5 seconds.

[0013] The present invention has the following beneficial effects: In this invention, a dynamic composite disaster causal graph containing disaster interaction edges and disaster coupling operators is dynamically generated, and a pre-trained graph neural network is used to extract global topological features in one go, directly outputting the nonlinear superposition causal effect on facility preservation and tourist evacuation efficiency. This method can automatically learn the enhancement or cancellation relationship between concurrent disasters, making risk assessment closer to the real physical world and avoiding the inaccuracy of assessment caused by simple superposition. Attached Figure Description

[0014] Figure 1 This is a system block diagram of a natural disaster early warning and tourism safety emergency response system proposed in this invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Please see Figure 1 As shown, this invention is a natural disaster early warning and tourism safety emergency response system, comprising: Data Layer: Constructs a spatiotemporal disaster causal network. The root node of the spatiotemporal disaster causal network includes at least one of typhoon, wildfire, avalanche, and heat wave. Each root node is configured with a disaster feature matrix, which includes the influence radius and disaster intensity. The nodes and edges of the spatiotemporal disaster causal network have geographical coordinates and time delay attributes. Each edge is configured with a time delay coefficient. The activation condition of the intermediate node is that the result of the weighted sum of the disaster feature matrix value of the upstream node after applying a distance decay function and a time delay function is greater than 0.7. The distance decay function is the product function of the inverse of the influence radius in the disaster feature matrix and the distance. The time delay function depends on the time delay coefficient. The data layer embeds a spatiotemporal disaster script tree. When the cosine similarity between the real-time disaster type activation vector and any path in the spatiotemporal disaster script tree is greater than 0.9, the predicted secondary disaster type and the estimated arrival time are output. Decision layer: When two or more root nodes are identified concurrently, disaster interaction edges are established and disaster coupling operators are configured to form a dynamic composite disaster causal graph. The pre-trained graph neural network outputs the nonlinear superposition causal effect on facility preservation and tourist evacuation efficiency. Before starting optimization, the graph edit distance similarity between the current disaster subgraph and the composite disaster mode subgraph in the independently maintained extreme disaster chain long memory buffer is calculated. If it is greater than 0.85, the historical peak coupling probability of the mode is forced to cover the current coupling strength as a constraint condition. A multi-factor evolutionary algorithm is run, and a real-time crowd simulation engine is called to evaluate candidate solutions. The real-time crowd simulation engine outputs the minimum evacuation time and the minimum tourist behavior entropy based on the social force model and the real-time collected tourist density, flow rate and emotion parameters. The positions with tourist behavior entropy values ​​greater than 0.8 are marked as dynamic bottleneck points and a temporary constraint of "node passage status" is generated and added to the search space. The Pareto optimal response solution set is generated with the minimum evacuation time, the maximum material coverage rate and the minimum tourist behavior entropy as objectives. Edge layer: It includes the real-time crowd simulation engine and a decision tree model obtained by knowledge distillation of the Pareto optimal response scheme set. The decision tree model is pre-installed in the disaster information box or scenic area gate. Interaction Layer: The Pareto optimal response scheme set is transformed into differentiated instructions through a three-level role-based permission system and forcibly pushed to the tour guide terminal and the tourist mobile terminal. Each instruction is attached with a digital signature. The tour guide terminal and the tourist mobile terminal must return a confirmation of receipt within 30 seconds. Otherwise, the instruction content is automatically upgraded to a scenic area broadcast.

[0017] In one embodiment, the distance decay function is specifically: ,in, For spatial distance, The influence radius in the catastrophic feature matrix of the upstream node is defined as follows: the time delay function is specifically a step function, which is set to 1 when the actual delay time is greater than or equal to the time delay coefficient of the edge, and 0 otherwise. The weights of the weighted summation include a catastrophic intensity factor of 0.4 and a spatiotemporal coupling factor of 0.6.

[0018] In one embodiment, when the decision layer constructs the dynamic composite disaster causal graph, the disaster interaction edge connects concurrent root nodes, the disaster coupling operator is a product normalization function of the corresponding time period intensity in the disaster feature matrix of each root node, and the pre-trained graph neural network contains three graph convolutional layers and one fully connected layer. It takes the adjacency matrix of the dynamic composite disaster causal graph and the disaster feature matrix of all nodes as input, and outputs the predicted values ​​of the facility preservation degree and the tourist evacuation efficiency.

[0019] In one embodiment, the decision layer executes a dual-track rare event circuit breaker. When executing the dual-track rare event circuit breaker, the method for extracting the structural features of the current disaster subgraph is as follows: constructing a minimum generated subgraph that only contains the currently activated disaster root node and the disaster interaction edge; calculating the Weisfeiler-Lehman graph kernel feature vector of the minimum generated subgraph; and calculating the graph edit distance similarity based on the normalized edit distance between the minimum generated subgraph and each composite disaster mode subgraph in the extreme disaster chain length memory buffer. When the maximum normalized edit distance similarity is greater than 0.85, the forced coverage is triggered.

[0020] In one embodiment, the real-time crowd simulation engine divides the simulation area into a 15m × 15m grid, and counts the tourist density and movement direction unit vector in each grid every second. The tourist behavior entropy is defined as the information entropy of the movement direction in the grid. When the tourist behavior entropy value of a certain grid is greater than 0.8 and lasts for 5 seconds, it is marked as a dynamic bottleneck point and reported to the decision layer.

[0021] In one embodiment, the temporary constraint on node traffic status generated by the decision layer includes the grid coordinates of the dynamic bottleneck point, the restricted traffic direction, and the suggested detour node sequence. The lifespan of the temporary constraint is 30 seconds. If the visitor behavior entropy of the grid drops below 0.6 within 30 seconds, the constraint is automatically revoked; otherwise, it remains in effect and enters the next round of optimization.

[0022] In one embodiment, the spatiotemporal disaster script tree embedded in the data layer module contains no less than 200 pre-deduced propagation paths. Each path consists of a disaster node sequence, a trigger time window, and a spatial propagation hop count. The cosine similarity calculation is performed on the real-time disaster type activation vector and the preset disaster type vector in the path. When the maximum cosine similarity is greater than 0.9, the secondary disaster node in the path that has not yet been activated but is within 30 minutes after the current time is output as the prediction result.

[0023] In one embodiment, the decision-making layer includes a plan hot-switching unit that receives the predicted secondary disaster type and estimated arrival time. When the estimated arrival time is less than 15 minutes, it automatically switches to the on-site evacuation plan. When the estimated arrival time is between 15 and 30 minutes, it automatically switches to the cableway evacuation plan and forcibly pushes the switched instruction to all terminals through the interaction layer within 5 seconds.

[0024] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A natural disaster early warning and tourism safety emergency response system, characterized in that, include: Data Layer: Constructs a spatiotemporal disaster causal network. Each root node is configured with a disaster feature matrix, which includes the influence radius and disaster intensity. The nodes and edges of the spatiotemporal disaster causal network have geographic coordinates and time delay attributes. Each edge is configured with a time delay coefficient. The activation condition of the intermediate node is that the result of the weighted sum of the disaster feature matrix value of the upstream node after the distance decay function and the time delay function is greater than 0.

7. The data layer embeds a spatiotemporal disaster script tree. When the cosine similarity between the real-time disaster type activation vector and any path in the spatiotemporal disaster script tree is greater than 0.9, the predicted secondary disaster type and the estimated arrival time are output. Decision layer: When two or more root nodes are identified concurrently, disaster interaction edges are established and disaster coupling operators are configured to form a dynamic composite disaster causal graph. The pre-trained graph neural network outputs the nonlinear superposition causal effect on facility preservation and tourist evacuation efficiency. Before starting optimization, the graph edit distance similarity between the current disaster subgraph and the composite disaster mode subgraph in the independently maintained extreme disaster chain long memory buffer is calculated. If it is greater than 0.85, the historical peak coupling probability of the mode is forced to cover the current coupling strength as a constraint condition. A multi-factor evolutionary algorithm is run, and a real-time crowd simulation engine is called to evaluate candidate solutions. The real-time crowd simulation engine generates a Pareto optimal response solution set with the objectives of minimum evacuation time, maximum material coverage, and minimum tourist behavior entropy. Edge layer: Includes the real-time crowd simulation engine and a decision tree model obtained by knowledge distillation of the Pareto optimal response scheme set; Interaction Layer: The Pareto optimal response scheme set is transformed into differentiated instructions through a three-level role-based permission system and forcibly pushed to the tour guide terminal and tourist mobile terminal. Each instruction is attached with a digital signature.

2. The natural disaster early warning and tourism safety emergency response system according to claim 1, characterized in that, The distance decay function is specifically as follows: ,in, For spatial distance, The radius of influence in the catastrophic feature matrix of the upstream node is defined as the time delay function, which is specifically a step function. It is set to 1 when the actual delay time is greater than or equal to the time delay coefficient of the edge, and otherwise to 0.

3. The natural disaster early warning and tourism safety emergency response system according to claim 1, characterized in that, When the decision layer constructs the dynamic composite disaster causal graph, the disaster interaction edge connects concurrent root nodes, the disaster coupling operator is the product normalization function of the intensity of the corresponding time period in the disaster feature matrix of each root node, the pre-trained graph neural network contains three graph convolutional layers and one fully connected layer, inputs the adjacency matrix of the dynamic composite disaster causal graph and the disaster feature matrix of all nodes, and outputs the predicted values ​​of the facility preservation degree and the tourist evacuation efficiency.

4. The natural disaster early warning and tourism safety emergency response system according to claim 3, characterized in that, The decision layer executes a dual-track rare event circuit breaker. When executing the dual-track rare event circuit breaker, the method for extracting the structural features of the current disaster subgraph is as follows: constructing a minimum generated subgraph that only contains the currently activated disaster root node and the disaster interaction edge; calculating the Weisfeiler-Lehman graph kernel feature vector of the minimum generated subgraph; and calculating the graph edit distance similarity based on the normalized edit distance between the minimum generated subgraph and each composite disaster mode subgraph in the extreme disaster chain length memory buffer. When the maximum normalized edit distance similarity is greater than 0.85, the forced coverage is triggered.

5. The natural disaster early warning and tourism safety emergency response system according to claim 1, characterized in that, The real-time crowd simulation engine divides the simulation area into grids and counts the tourist density and movement direction unit vector in each grid every second. The tourist behavior entropy is defined as the information entropy of the movement direction in the grid. When the tourist behavior entropy value of a certain grid is greater than 0.8 and lasts for 5 seconds, it is marked as a dynamic bottleneck point and reported to the decision layer.

6. The natural disaster early warning and tourism safety emergency response system according to claim 5, characterized in that, The temporary constraints on node traffic status generated by the decision layer include the grid coordinates of the dynamic bottleneck point, the restricted traffic direction, and the suggested detour node sequence. The lifespan of the temporary constraints is 30 seconds. If the visitor behavior entropy of the grid drops below 0.6 within 30 seconds, the constraints are automatically revoked; otherwise, they remain in effect and enter the next round of optimization.

7. The natural disaster early warning and tourism safety emergency response system according to claim 1, characterized in that, The spatiotemporal disaster script tree embedded in the data layer module contains no less than 200 pre-simulated propagation paths. Each path consists of a disaster node sequence, a trigger time window, and a spatial propagation hop count. The cosine similarity calculation is performed on the real-time disaster type activation vector and the preset disaster type vector in the path. When the maximum cosine similarity is greater than 0.9, the secondary disaster node in the path that has not yet been activated but is within 30 minutes after the current time is output as the prediction result.

8. The natural disaster early warning and tourism safety emergency response system according to claim 7, characterized in that, The decision-making layer includes a plan hot-switching unit, which receives the predicted secondary disaster type and estimated arrival time. When the estimated arrival time is less than 15 minutes, it automatically switches to the on-site evacuation plan. When the estimated arrival time is between 15 and 30 minutes, it automatically switches to the cableway evacuation plan and pushes the switched instructions to all terminals through the interaction layer.

Images