A natural disaster monitoring and early warning method and system based on emergency large model platform

Abstract

The present invention provides a natural disaster monitoring and early warning method and system based on an emergency large-scale model platform. The method comprises: step 1: acquiring multi-source data for natural disaster monitoring and early warning; step 2: constructing a disaster identification model; and step 3: utilizing the disaster identification model to respond to the multi-source data to identify disasters, and issuing early warning information via the emergency large-scale model platform. The present invention utilizes artificial intelligence technology to construct a disaster identification model, achieving intelligent and automated disaster identification. The model is connected to the emergency large-scale model platform. The platform acquires multi-source data by real-time monitoring multiple data nodes, then uses the model for intelligent analysis to output early warning information, thereby improving the real-time and accuracy of early warning information release.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to a natural disaster monitoring and early warning method and system based on an emergency large model platform. Background Art

[0002] Emergency big model platforms are systems that leverage artificial intelligence, particularly deep learning models, to improve the efficiency and effectiveness of emergency management. By integrating and analyzing large amounts of data, these platforms provide risk monitoring, early warning, and emergency decision support to address various emergencies.

[0003] At present, natural disaster monitoring and early warning are connected with various early warning results, lacking advance warning of natural disaster risks and intelligent and automated management of disaster warnings.

[0004] In view of this, there is an urgent need for a natural disaster monitoring and early warning method and system based on an emergency large model platform to at least solve the above-mentioned deficiencies. Summary of the Invention

[0005] One of the purposes of the present invention is to provide a natural disaster monitoring and early warning method and system based on an emergency large model platform, which uses artificial intelligence technology to construct a disaster identification model to realize the intelligence and automation of disaster identification. The model is connected to the emergency large model platform. The platform obtains multi-source data by real-time monitoring of multiple data nodes, and then uses the model to intelligently analyze and output early warning information, thereby improving the real-time and accuracy of early warning information release.

[0006] An embodiment of the present invention provides a natural disaster monitoring and early warning method based on an emergency large model platform, comprising:

[0007] Step 1: Obtain multi-source data for natural disaster monitoring and early warning;

[0008] Step 2: Build a disaster identification model;

[0009] Step 3: Use the disaster identification model to respond to multi-source data to identify disasters and issue early warning information through the emergency model platform.

[0010] Preferably, step 1: obtaining multi-source data for natural disaster monitoring and early warning includes:

[0011] Connect to multiple target data nodes; target data nodes include: national early warning information node, earthquake network center node, meteorological station node, satellite remote sensing image acquisition node, meteorological observation station node, geographic information node, social media node and historical disaster record node;

[0012] Based on the target data node, obtain multi-source data.

[0013] Preferably, step 2: constructing a disaster identification model includes:

[0014] Collect training data;

[0015] Get the pre-trained model;

[0016] Use the training data to train the pre-trained model and evaluate the performance of the trained model based on the validation data;

[0017] Optimize the training model based on the evaluation results of the performance indicators to obtain the optimized model;

[0018] The test data is used to verify the generalization ability of the optimization model, and the disaster identification model is obtained based on the verification results of the generalization ability.

[0019] Preferably, step 3: using the disaster identification model to respond to multi-source data to identify disasters and issuing warning information through the emergency large model platform includes:

[0020] Obtain disaster identification results;

[0021] Obtain disaster fluctuation curves based on disaster identification types;

[0022] Obtain geographic elements based on disaster identification results and disaster fluctuation curves;

[0023] Plan evacuation routes and safe shelters based on geographical factors;

[0024] The disaster identification results, evacuation routes and safe shelters are used as early warning information and sent to the platform early warning node based on the emergency big model platform.

[0025] Preferably, obtaining geographical elements based on the disaster identification results and the disaster fluctuation curve includes:

[0026] Obtain local disaster areas based on disaster identification results and disaster fluctuation curves;

[0027] Mark the local disaster areas on the target map that includes all local disaster areas;

[0028] Obtain the regional center distances between local disaster areas;

[0029] If the distance between the regional centers is less than a preset distance threshold, obtaining a first local target map including the corresponding local disaster area;

[0030] If the distance between the regional centers is greater than or equal to a preset distance threshold, determining whether the corresponding local disaster area is in any of the first local target maps; if not, obtaining a second local target map containing the corresponding local disaster area;

[0031] The geographic elements are acquired according to the first partial target map, the second partial target map and the geographic element extraction template corresponding to the target map.

[0032] Preferably, evacuation routes and safe shelters should be planned based on geographical factors, including:

[0033] Create a three-dimensional map of the disaster area based on geographic elements; the three-dimensional map of the disaster area will mark the local disaster area;

[0034] Determine the evacuation starting point in the three-dimensional map of the disaster area;

[0035] Determine non-local disaster areas in the three-dimensional map of disaster areas;

[0036] Obtaining the first associated path from the evacuation starting point to the non-local disaster area;

[0037] Extracting, based on a preset feature extraction template, first relationship features between the first associated path and the local disaster area in the three-dimensional map of the disaster area, the first relationship features including: associated building information of the first associated path, the total path length of the first associated path, the type of disaster in the local disaster area through which the first associated path passes, and the local path length of the first associated path through the local disaster area;

[0038] According to the first relationship feature, construct the first correlation path description factor;

[0039] Determining the best first associated path as the evacuation route based on the first associated path description factor and a preset evacuation route evaluation model;

[0040] Acquire a second associated path connected to the evacuation route;

[0041] Extracting, based on a preset feature extraction template, a second relational feature between the second associated path and the local disaster area in the three-dimensional map of the disaster area;

[0042] According to the second relationship feature, construct the second association path description factor;

[0043] Get the shelter search template;

[0044] According to the second associated path description factor and the refuge search template, the refuge spatially connected to the evacuation route is determined.

[0045] Preferably, obtaining a shelter search template includes:

[0046] Obtaining records of searches for shelters;

[0047] Parse the shelter search records and obtain the associated description semantics of the relationship feature type;

[0048] Constructing the first vector based on the association description semantics of different relationship feature types in the same shelter search record;

[0049] Performing cluster analysis on the first vector to obtain a second vector;

[0050] A shelter search template is constructed based on the second vector.

[0051] An embodiment of the present invention provides a natural disaster monitoring and early warning method based on an emergency large model platform, further comprising:

[0052] After the early warning information is issued, the movement status information of the crowd in the evacuation area is obtained, and the flight route of the target drone is planned based on the crowd movement status information.

[0053] Preferably, planning the flight route of the target UAV based on the crowd movement status information includes:

[0054] Plan the first flight path of the target UAV based on the evacuation route;

[0055] Control the target UAV to fly along the first route, and at the same time, monitor the movement status information of the crowd in real time;

[0056] Calculate the movement trajectory of the crowd based on the crowd movement status information;

[0057] Calculate the deviation between the crowd movement trajectory and the first driving route;

[0058] If the deviation is greater than or equal to the preset deviation threshold, the movement trajectory of the local crowd from the start of the deviation to the current moment is obtained;

[0059] Determine the conditions based on flight information and the movement trajectory of local people to determine the new flight information of the target drone;

[0060] Control the target UAV based on the newly added flight information while acquiring target images;

[0061] Acquire a comparison image corresponding to the target image when traveling along the local driving route from the time of starting deviation to the current time;

[0062] Calculate the image difference between the target image and the comparison image;

[0063] Predict sudden warnings based on image differences;

[0064] The first driving route is updated according to the emergency warning to obtain a second driving route, and the target UAV is controlled to fly according to the second driving route.

[0065] An embodiment of the present invention provides a natural disaster monitoring and early warning system based on an emergency large-scale model platform, comprising:

[0066] Data acquisition subsystem, used to obtain multi-source data for natural disaster monitoring and early warning;

[0067] Disaster identification model construction subsystem, used to build disaster identification model;

[0068] The early warning subsystem is used to identify disasters using disaster identification models in response to multi-source data, and to issue early warning information through the emergency large model platform.

[0069] The beneficial effects of the present invention are:

[0070] The present invention uses artificial intelligence technology to construct a disaster identification model to realize the intelligent and automated disaster identification. The model is connected to the emergency large model platform. The platform obtains multi-source data by real-time monitoring of multiple data nodes, and then uses the model for intelligent analysis to output warning information, thereby improving the real-time and accuracy of warning information release.

[0071] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in this application document.

[0072] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0074] Figure 1 Schematic diagram of a natural disaster monitoring and early warning method based on an emergency large model platform in an embodiment of the present invention;

[0075] Figure 2 Schematic diagram of a natural disaster monitoring and early warning system based on an emergency large model platform in an embodiment of the present invention. DETAILED DESCRIPTION

[0076] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0077] The embodiment of the present invention provides a natural disaster monitoring and early warning method based on an emergency large model platform, such as Figure 1 Shown, including:

[0078] Step 1: Obtain multi-source data for natural disaster monitoring and early warning;

[0079] Step 1: Acquire multi-source data for natural disaster monitoring and early warning, including:

[0080] Connect to multiple target data nodes; target data nodes include: national early warning information node, earthquake network center node, meteorological station node, satellite remote sensing image acquisition node, meteorological observation station node, geographic information node, social media node and historical disaster record node;

[0081] Based on the target data node, obtain multi-source data;

[0082] Step 2: Build a disaster identification model;

[0083] Among them, step 2: building a disaster identification model includes:

[0084] Collect training data; wherein the training data includes historical multi-source data and corresponding disaster identification results, and the historical multi-source data is obtained from the same source as the multi-source data;

[0085] Obtain a pre-trained model; a pre-trained model is a deep learning model used for disaster identification model training, such as SegNet U-Net, DeepLab, and CNN;

[0086] Use the training data to train the pre-trained model and evaluate the performance of the trained model based on the validation data;

[0087] Optimize the training model based on the evaluation results of the performance indicators to obtain the optimized model;

[0088] Use test data to verify the generalization ability of the optimization model, and obtain a disaster identification model based on the verification results of the generalization ability;

[0089] Step 3: Use the disaster identification model to respond to multi-source data to identify disasters and issue warning information through the emergency model platform. The emergency model platform is a comprehensive platform that integrates multiple emergency response and disaster management functions.

[0090] The working principle and beneficial effects of the above technical solution are:

[0091] The natural disaster monitoring and early warning method based on the emergency large-scale model platform specifically includes:

[0092] Step 1: Multi-source data fusion

[0093] Integrate data from various sources, including but not limited to the National Early Warning Information Center, the China Earthquake Networks Center, and the China Meteorological Administration, satellite remote sensing images, meteorological observation station data, geographic information system (GIS) information, social media data, and historical disaster records. The integration of these data will help to more fully understand the environmental changes before and after the disaster.

[0094] Step 2: Advanced Analytics and Predictive Modeling

[0095] Disaster identification models built using machine learning and deep learning techniques can be used to process and analyze the aforementioned multi-source data to identify patterns and trends in disaster occurrence. These models may include, but are not limited to, time series analysis, anomaly detection, classification, and regression algorithms to predict the likelihood, timing, and location of disasters. Model building is divided into the following seven steps:

[0096] 1. Collect data

[0097] Collect training data to train the model. Ensure the data is of high quality and contains enough information for the model to learn.

[0098] 2. Data Preprocessing

[0099] Cleaning data: removing or correcting incomplete, erroneous, or abnormal data.

[0100] Feature engineering: Selecting, transforming, or creating new features that help improve model performance.

[0101] Standardization / normalization: Scaling data to make data at different scales comparable.

[0102] Divide the dataset: usually into training set, validation set and test set.

[0103] 3. Select a model

[0104] Choose an appropriate model based on the nature of the problem. For machine learning, consider decision trees, random forests, support vector machines, etc.; for deep learning, consider convolutional neural networks (CNN), recurrent neural networks (RNN), long short-term memory networks (LSTM), etc.

[0105] 4. Training the Model

[0106] Use training data to train the model, including setting hyperparameters, choosing loss functions and optimization algorithms, etc.

[0107] For deep learning, configure the required computing resources (such as GPU) to accelerate the training process.

[0108] 5. Model Evaluation

[0109] Use the validation set to evaluate the model's performance and adjust hyperparameters to optimize the model. Common evaluation metrics (performance indicators) include accuracy, recall, F1 score, and mean squared error.

[0110] 6. Test the model

[0111] Evaluate the model on unseen test data to check its generalization ability.

[0112] 7. Deploy the model

[0113] When the model's performance on the test set meets the preset standard, it can be deployed to the production environment.

[0114] Step 3: Real-time monitoring and early warning system

[0115] The emergency large-scale model platform responds quickly to new data input and issues early warning information in a timely manner based on the analysis results of the model configured on the platform.

[0116] The present invention uses artificial intelligence technology to construct a disaster identification model to realize the intelligent and automated disaster identification. The model is connected to the emergency large model platform. The platform obtains multi-source data by real-time monitoring of multiple data nodes, and then uses the model for intelligent analysis to output warning information, thereby improving the real-time and accuracy of warning information release.

[0117] In one embodiment, step 3: utilizing the disaster identification model to respond to multi-source data to identify disasters and issuing warning information through the emergency model platform includes:

[0118] Obtain disaster identification results;

[0119] Obtain disaster fluctuation curves based on disaster identification types;

[0120] Obtain geographic elements based on disaster identification results and disaster fluctuation curves;

[0121] Plan evacuation routes and safe shelters based on geographical factors;

[0122] Disaster identification results, evacuation routes, and safe refuges are used as early warning information and sent to the platform's early warning nodes based on the emergency model platform;

[0123] Among them, according to the disaster identification results and disaster fluctuation curve, geographical elements are obtained, including:

[0124] According to the disaster identification results and the disaster fluctuation curve, a local disaster area is obtained; wherein the local disaster area is: the disaster range determined according to the disaster identification results and the disaster fluctuation curve;

[0125] Mark the local disaster areas on the target map that includes all local disaster areas;

[0126] Obtaining the distance between regional centers of local disaster areas; wherein the regional center determination rule is preset manually;

[0127] If the distance between the regional centers is less than a preset distance threshold, a first local target map including the corresponding local disaster area is obtained; wherein the preset distance threshold is manually pre-set, for example, 3 km; the first local target map is as small as possible while including the corresponding local disaster area;

[0128] If the distance between the regional centers is greater than or equal to a preset distance threshold, determining whether the corresponding local disaster area is in any of the first local target maps; if not, obtaining a second local target map containing the corresponding local disaster area;

[0129] Obtaining geographic elements based on the first partial target map, the second partial target map, and a geographic element extraction template corresponding to the target map; wherein the geographic element extraction template is used to extract geographic elements (elements contained in the natural form of the earth's surface, as well as elements formed by human transformation of the natural world during production activities) by reference to the target map;

[0130] Among them, planning evacuation routes and safe shelters based on geographical factors includes:

[0131] Create a three-dimensional map of the disaster area based on geographic elements; the three-dimensional map of the disaster area will mark the local disaster area;

[0132] Determine an evacuation starting point in a three-dimensional map of the disaster area; wherein the rules for determining the evacuation starting point are manually preset, such as a residential area;

[0133] Determine non-local disaster areas in the three-dimensional map of disaster areas;

[0134] Obtaining a first associated path from the evacuation starting point to the non-local disaster area; wherein the first associated path is: route information from the evacuation starting point to the non-local disaster area in the three-dimensional map of the disaster area, obtained based on path planning technology;

[0135] Based on a preset feature extraction template, a first relationship feature between the first associated path and the local disaster area in the three-dimensional map of the disaster area is extracted, the first relationship feature including: associated building information of the first associated path (for example, if the first associated path is a fire escape, the associated building information is related information of the corresponding fire escape), the total path length of the first associated path, the type of disaster in the local disaster area through which the first associated path passes, and the local path length of the first associated path through the local disaster area; wherein the preset feature extraction template is a template for extracting relationship features (the associated building information of the path, the total length of the path, the type of disaster in the local disaster area through which the path passes, and the local path length of the path through the local disaster area) by comparing the path with the three-dimensional map of the disaster area;

[0136] Constructing a first association path description factor according to the first relationship feature; wherein the first association path description factor is: a description vector of the first association path constructed according to the first relationship feature;

[0137] Based on the first associated path description factor and a preset evacuation route evaluation model, the optimal first associated path is determined as the evacuation route. The evacuation route evaluation model is an AI model that learns from the evaluation experience of manual evacuation route evaluation to evaluate evacuation routes. For example, the stronger the ability of the space where the path is located to resist the type of disaster in the local disaster area determined based on the associated building information of the path, the higher the evaluation level; the longer the total length of the evacuation route, the lower the evaluation level; and the longer the length of the local path of the evacuation route passing through the local disaster area, the lower the evaluation level.

[0138] Acquire a second associated path connected to the evacuation route; wherein the second associated path is: a branch road (bypass road) connected to the evacuation route;

[0139] Extracting, based on a preset feature extraction template, second relationship features between the second associated path and the local disaster area in the three-dimensional map of the disaster area; wherein the second relationship features include: associated building information of the second associated path, the type of disaster in the local disaster area through which the second associated path passes, and the local path length of the second associated path through the local disaster area;

[0140] According to the second relationship feature, a second association path description factor is constructed; wherein the second association path description factor is: a description vector of the second association path;

[0141] Get the shelter search template;

[0142] Determine the refuge places spatially connected to the evacuation route according to the second associated path description factor and the refuge place search template;

[0143] Get a template for finding a shelter, including:

[0144] Obtaining a shelter search record; wherein the shelter search record is a record of manually viewing the visual model presented by the emergency large model platform and searching for shelters therein;

[0145] Parsing the shelter search records to obtain the associated description semantics of the relationship feature type; wherein the associated description semantics is obtained by analyzing and extracting the semantics related to the relationship feature type in the shelter search records based on semantic analysis technology;

[0146] Constructing a first vector based on the associated description semantics of different relationship feature types in the same shelter search record; wherein the first vector is: a semantic description vector for shelter search constructed by extracting the associated description semantics based on the same shelter search record;

[0147] Performing cluster analysis on the first vector to obtain a second vector; wherein the second vector is: Cluster analysis is: clustering similar first vectors to obtain multiple cluster vector clusters, and determining the vector represented by each cluster vector cluster;

[0148] A shelter search template is constructed based on the second vector.

[0149] The working principle and beneficial effects of the above technical solution are:

[0150] The disaster fluctuation curve is a curve that describes the relationship between disaster intensity and possible losses. The disaster identification results are analyzed to determine the disaster identification type. The disaster scope (local disaster area) is deduced based on the disaster fluctuation curve corresponding to the disaster type (disaster identification type) and the disaster intensity obtained by analyzing the disaster identification results. The regional center distance between local disaster areas is obtained, and the corresponding local disaster areas with regional center distances less than the distance threshold are grouped together to obtain the first local target map corresponding to the grouped local disaster areas. The second local target map is separately obtained for the local disaster areas with regional center distances greater than or equal to the distance threshold and not in any of the first local target maps. The geographic element extraction template corresponding to the target map is introduced to compare the first local target map and the second local target map to obtain geographic elements. Taking into account the impact between similar local disaster areas, the maps between similar local disaster areas are obtained to extract geographic elements, thereby improving the rationality of subsequent evacuation route planning.

[0151] When planning evacuation routes and safe refuges, a three-dimensional map of the disaster area is first created based on geographic elements. An evacuation starting point is determined in the three-dimensional map of the disaster area according to a rule for determining an evacuation starting point. Non-local disaster areas other than the local disaster area are determined in the three-dimensional map of the disaster area. A first associated path from the evacuation starting point to the non-local disaster area is planned based on path planning technology. A feature extraction template is introduced to extract first relationship features, including: associated building information of the first associated path, the total path length of the first associated path, the type of disaster in the local disaster area passed by the first associated path, and the local path length of the first associated path passing through the local disaster area, and a first associated path description factor is constructed. An evacuation route evaluation model is obtained by learning from manual evaluation experience of evacuation routes based on machine learning technology, and is used to intelligently output the best-evaluated first associated path (evacuation route) based on the first associated path description factor. A second associated path connected to the evacuation route is obtained, a second relationship feature is extracted based on the feature extraction template, and a description vector (second associated path description factor) of the second associated path is constructed based on the second relationship feature. A refuge search template is introduced to determine refuges spatially connected to the evacuation route based on the second associated path description factor, thereby improving the rationality of refuge.

[0152] The present invention creates a three-dimensional map of the disaster area based on the geographical elements of the disaster area, marks the disaster range on the three-dimensional map of the disaster area to obtain a three-dimensional map of the disaster area containing the disaster area; introduces path planning technology to plan evacuation routes and safe refuge places in the marked map, and sends them to the platform warning node associated with the emergency large model platform, thereby improving the timeliness of the warning and being more humane.

[0153] An embodiment of the present invention provides a natural disaster monitoring and early warning method based on an emergency large model platform, further comprising:

[0154] After the warning information is issued, the crowd movement status information in the evacuation area is obtained, and the flight route of the target drone is planned based on the crowd movement status information; the crowd movement status information includes: the movement direction and speed of the crowd in the evacuation area;

[0155] Among them, according to the crowd movement status information, the flight route of the target drone is planned, including:

[0156] Planning a first flight path for the target UAV based on the evacuation route; wherein the first flight path is: a flight path for the target UAV determined based on the evacuation route initially planned by the emergency large-scale model platform, such as an airspace route parallel to the evacuation route;

[0157] Control the target UAV to fly along the first route, and at the same time, monitor the movement status information of the crowd in real time;

[0158] Calculate the movement trajectory of the crowd based on the crowd movement status information;

[0159] Calculate the deviation between the crowd movement trajectory and the first driving route;

[0160] If the deviation is greater than or equal to a preset deviation threshold, the local crowd movement trajectory from the start of the deviation to the current moment is obtained; wherein the preset deviation threshold is manually preset, for example: 0.9;

[0161] Determine the conditions based on flight information and the movement trajectory of local people to determine the new flight information of the target drone;

[0162] Controlling the target UAV to fly according to the newly added flight information while acquiring a target image; wherein the target image is: an image acquired when the target UAV flies based on the newly added flight information;

[0163] Obtaining a comparison image corresponding to the target image when traveling along the local driving route from the time of the start of the deviation to the current time; wherein the comparison image is: when generating the evacuation route, a three-dimensional map of the local disaster area within a preset range (e.g., within 10 km) around the local driving route corresponding to the time of the start of the deviation to the current time;

[0164] Calculate the image difference between the target image and the comparison image; the image difference is the dissimilar portion of the image matching between the target image and the comparison image, for example, building A is not shown on fire in the 3D map of the local disaster area, but is shown on fire in the target image.

[0165] Predict sudden warnings based on image differences; sudden warnings are new warnings obtained through image difference analysis, such as a fire in building A.

[0166] updating the first driving route to obtain a second driving route based on the emergency warning, and controlling the target UAV to fly along the second driving route; wherein, when updating the first driving route, re-deducing the personnel evacuation route based on the emergency warning, and determining the second driving route based on the re-deduced personnel evacuation route;

[0167] The flight information confirmation conditions include:

[0168] After the current moment, there are n target drones above the local crowd movement trajectory, M≤n<N, the quotient of the local crowd movement trajectory S and M is less than or equal to the preset quotient threshold, and the quotient of the local crowd movement trajectory S and (M-1) is greater than the quotient threshold. The quotient threshold is a preset value, such as 0.5 km, and n, N, and M are all positive integers;

[0169] The extension lines of the fuselages of n target drones are parallel to each other, and Δ(d i+2 -d i+1 )≥Δ(di+1 -d i ), d i is the distance between the extended direction of the fuselage of the i-th target UAV and the extended direction of the fuselage of the i-1-th target UAV, 2≤i≤n-3, and the order of the extended direction of the fuselage is from the first point corresponding to the movement trajectory of the local crowd at the time of the deviation to the second point corresponding to the movement trajectory of the local crowd at the current time;

[0170] The extension lines of the fuselage's direction all intersect with the local driving routes.

[0171] The working principle and beneficial effects of the above technical solution are:

[0172] Although the emergency model platform can quickly process data and provide early warning feedback, its data acquisition relies heavily on data acquisition devices such as terminal sensors. When the terminal sensor equipment is damaged, the early warning is not timely. In addition, a disaster can easily cause subsequent disasters (for example, the spread of fire as the wind direction changes). When encountering subsequent dangers after the initial evacuation, people will also change their routes based on their own judgment or the auxiliary judgment of the evacuation personnel. The target drone was originally used to track the evacuation of personnel. It is easy to lose track of the evacuated people if it has been flying based on the initial evacuation route. The embodiment of the present invention can solve this problem well. It calculates the deviation between the movement trajectory of the crowd and the first driving route in real time. When the deviation is greater than the deviation threshold, the flight information determination condition is introduced to add new flight information of the target drone. The flight information determination condition includes:

[0173] Condition 1: n target drones are dispatched to conduct inspections above the local crowd movement trajectory, M≤n<N. Setting the upper limit condition N ensures that there are target drones that continuously monitor based on the initial route, while calling some target drones to obtain target images. Setting the lower limit condition M ensures that the number of target images obtained is more compatible with the local crowd movement trajectory S and the target images are obtained more comprehensively.

[0174] Condition 2: The extension lines of the target drone’s fuselage are parallel to each other and pass through the initial local driving route. The target image is acquired in an orderly manner, and the source direction is the crowd avoidance direction. The constraint Δ(d i+2 -d i+1 )≥Δ(d i+1 -d i ), the closer to the deviation point, the denser the target UAVs are set for image acquisition, which improves the rationality of the target image acquisition process;

[0175] Based on the newly added flight information, the target image is obtained, and the comparison image of the target image is determined and compared. If the corresponding dangerous situation does not exist in the comparison image, an emergency warning is predicted based on the image difference. Based on the emergency warning update, a second driving route is obtained, and the target UAV is controlled to fly according to the second driving route, which improves the rationality of the evacuation personnel tracking strategy.

[0176] The embodiment of the present invention provides a natural disaster monitoring and early warning system based on an emergency large model platform, such as Figure 2 As shown, including:

[0177] Data acquisition subsystem 1, used to acquire multi-source data for natural disaster monitoring and early warning;

[0178] Disaster identification model construction subsystem 2, used to construct a disaster identification model;

[0179] The early warning subsystem 3 is used to identify disasters using a disaster identification model in response to multi-source data, and to issue early warning information through the emergency large model platform.

[0180] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A natural disaster monitoring and early warning method based on an emergency large model platform, characterized in that: include: Step 1: Obtain multi-source data for natural disaster monitoring and early warning; Step 2: Build a disaster identification model; Step 3: Use the disaster identification model to respond to multi-source data to identify disasters and issue early warning information through the emergency model platform; After the warning information is issued, the movement status information of the crowd in the evacuation area is obtained, and the flight route of the target drone is planned based on the crowd movement status information; Among them, according to the crowd movement status information, the flight route of the target drone is planned, including: Plan the first flight path of the target UAV based on the evacuation route; Control the target UAV to fly along the first route, and at the same time, monitor the movement status information of the crowd in real time; Calculate the movement trajectory of the crowd based on the crowd movement status information; Calculate the deviation between the crowd movement trajectory and the first driving route; If the deviation is greater than or equal to the preset deviation threshold, the movement trajectory of the local crowd from the start of the deviation to the current moment is obtained; Determine the conditions based on flight information and the movement trajectory of local people to determine the new flight information of the target drone; Control the target UAV based on the newly added flight information while acquiring target images; Acquire a comparison image corresponding to the target image when traveling along the local driving route from the time of starting deviation to the current time; Calculate the image difference between the target image and the comparison image; Predict sudden warnings based on image differences; The first driving route is updated according to the emergency warning to obtain a second driving route, and the target UAV is controlled to fly according to the second driving route.

2. A natural disaster monitoring and early warning method based on an emergency large model platform as claimed in claim 1, characterized in that: Step 1: Obtain multi-source data for natural disaster monitoring and early warning, including: Connect to multiple target data nodes; target data nodes include: national early warning information node, earthquake network center node, meteorological station node, satellite remote sensing image acquisition node, meteorological observation station node, geographic information node, social media node and historical disaster record node; Based on the target data node, obtain multi-source data.

3. A natural disaster monitoring and early warning method based on an emergency large model platform as claimed in claim 1, characterized in that: Step 2: Build a disaster identification model, including: Collect training data; Get the pre-trained model; Use the training data to train the pre-trained model and evaluate the performance of the trained model based on the validation data; Optimize the training model based on the evaluation results of the performance indicators to obtain the optimized model; The test data is used to verify the generalization ability of the optimization model, and the disaster identification model is obtained based on the verification results of the generalization ability.

4. A natural disaster monitoring and early warning method based on an emergency large model platform as claimed in claim 1, characterized in that: Step 3: Use the disaster identification model to respond to multi-source data to identify disasters and issue early warning information through the emergency model platform, including: Obtain disaster identification results; Obtain disaster fluctuation curves based on disaster identification types; Obtain geographic elements based on disaster identification results and disaster fluctuation curves; Plan evacuation routes and safe shelters based on geographical factors; The disaster identification results, evacuation routes and safe shelters are used as early warning information and sent to the platform early warning node based on the emergency big model platform.

5. A natural disaster monitoring and early warning method based on an emergency large model platform as claimed in claim 4, characterized in that: Based on the disaster identification results and disaster fluctuation curve, obtain geographic elements, including: Obtain local disaster areas based on disaster identification results and disaster fluctuation curves; Mark the local disaster areas on the target map that includes all local disaster areas; Obtain the regional center distances between local disaster areas; If the distance between the regional centers is less than a preset distance threshold, obtaining a first local target map including the corresponding local disaster area; If the distance between the regional centers is greater than or equal to a preset distance threshold, determining whether the corresponding local disaster area is in any of the first local target maps; if not, obtaining a second local target map containing the corresponding local disaster area; The geographic elements are acquired according to the first partial target map, the second partial target map and the geographic element extraction template corresponding to the target map.

6. A natural disaster monitoring and early warning method based on an emergency large model platform as claimed in claim 4, characterized in that: Plan evacuation routes and safe shelters based on geographical factors, including: Create a three-dimensional map of the disaster area based on geographic elements; the three-dimensional map of the disaster area will mark the local disaster area; Determine the evacuation starting point in the three-dimensional map of the disaster area; Determine non-local disaster areas in the three-dimensional map of disaster areas; Obtaining the first associated path from the evacuation starting point to the non-local disaster area; Extracting, based on a preset feature extraction template, first relationship features between the first associated path and the local disaster area in the three-dimensional map of the disaster area, the first relationship features including: associated building information of the first associated path, the total path length of the first associated path, the type of disaster in the local disaster area through which the first associated path passes, and the local path length of the first associated path through the local disaster area; According to the first relationship feature, construct the first correlation path description factor; Determining the best first associated path as the evacuation route based on the first associated path description factor and a preset evacuation route evaluation model; Acquire a second associated path connected to the evacuation route; Extracting, based on a preset feature extraction template, a second relational feature between the second associated path and the local disaster area in the three-dimensional map of the disaster area; According to the second relationship feature, construct the second association path description factor; Get the shelter search template; According to the second associated path description factor and the refuge search template, the refuge spatially connected to the evacuation route is determined.

7. A natural disaster monitoring and early warning method based on an emergency large model platform as claimed in claim 6, characterized in that: Get a template for finding a shelter, including: Obtaining records of searches for shelters; Parse the shelter search records and obtain the associated description semantics of the relationship feature type; Constructing the first vector based on the association description semantics of different relationship feature types in the same shelter search record; Performing cluster analysis on the first vector to obtain a second vector; A shelter search template is constructed based on the second vector.

8. A natural disaster monitoring and early warning system based on an emergency large-scale model platform, characterized in that: include: Data acquisition subsystem, used to obtain multi-source data for natural disaster monitoring and early warning; Disaster identification model construction subsystem, used to build disaster identification model; The early warning subsystem is used to identify disasters using a disaster identification model in response to multi-source data and issue early warning information through the emergency large model platform; The natural disaster monitoring and early warning system based on the emergency large model platform also performs the following operations: After the warning information is issued, the movement status information of the crowd in the evacuation area is obtained, and the flight route of the target drone is planned based on the crowd movement status information; Among them, according to the crowd movement status information, the flight route of the target drone is planned, including: Plan the first flight path of the target UAV based on the evacuation route; Control the target UAV to fly along the first route, and at the same time, monitor the movement status information of the crowd in real time; Calculate the movement trajectory of the crowd based on the crowd movement status information; Calculate the deviation between the crowd movement trajectory and the first driving route; If the deviation is greater than or equal to the preset deviation threshold, the movement trajectory of the local crowd from the start of the deviation to the current moment is obtained; Determine the conditions based on flight information and the movement trajectory of local people to determine the new flight information of the target drone; Control the target UAV based on the newly added flight information while acquiring target images; Acquire a comparison image corresponding to the target image when traveling along the local driving route from the time of starting deviation to the current time; Calculate the image difference between the target image and the comparison image; Predict sudden warnings based on image differences; The first driving route is updated according to the emergency warning to obtain a second driving route, and the target UAV is controlled to fly according to the second driving route.

Images