Temporary emergency shelter site selection method and system suitable for huge disaster scene

Abstract

The invention discloses a temporary emergency shelter site selection method and system suitable for a huge disaster scene, and relates to the technical field of natural disaster emergency management. According to the method, disaster data and remote sensing image data of a disaster area are processed, disaster grids and non-disaster grids in the disaster area are identified, the site selection range is narrowed, and the site selection precision and accuracy are further improved; risk evaluation, openness and traffic evaluation are carried out on the non-disaster grids, based on the evaluation result, the grey wolf algorithm is utilized to carry out site selection on the non-disaster grids, a temporary emergency shelter site selection scheme is obtained, through data driving, manual analysis dependence is reduced, the site selection automation degree is improved, the site investigation requirement is reduced, and efficiency is improved. And through integration and rapid analysis of multi-source data, dynamic response to change demands is realized, dynamics is enhanced, and a site selection strategy is updated through real-time data, so that emergency service coverage is ensured to be matched with actual demands.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of natural disaster emergency management, and particularly relates to a temporary emergency shelter site selection method and system suitable for a catastrophe scenario. BACKGROUND

[0002] After a catastrophe event, some original emergency shelters may be damaged and thus cannot be used. In this case, temporary emergency shelters need to be quickly selected to ensure the safe evacuation and short-term resettlement of disaster victims. The existing emergency shelter site selection technology has obvious deficiencies in dealing with catastrophe scenarios with widespread infrastructure damage and frequent secondary disasters. The site selection problem in this special situation has hardly been thoroughly researched. Meanwhile, the current emergency shelter site selection focuses more on long-term layout, and the selection of temporary shelters for quick response after a catastrophe is not considered. Since a catastrophe often causes widespread infrastructure damage and triggers secondary disasters, conventional shelters may not be available, and a more adaptable and faster temporary shelter site selection method is urgently needed to ensure the timeliness of safe evacuation and resettlement of personnel after a catastrophe.

[0003] In traditional emergency shelter site selection technology, site selection decisions are usually based on pre-disaster resource distribution and demand forecasts, such as material supply, traffic flow, medical resources, etc. However, when facing extreme disaster scenarios with widespread infrastructure damage and frequent secondary disasters, these traditional site selection methods are not up to the task. In this situation, due to the severe impact on resource and path availability, the layout and site selection of shelters must consider more dynamic and complex factors. However, this special situation of site selection has not been fully researched and explored, resulting in a lack of scientific guidance for effective shelter layout in extreme disaster scenarios.

[0004] The existing emergency shelter site selection usually focuses more on long-term layout, aiming to ensure the rational allocation of resources, transportation, and medical facilities in normal times. However, after a catastrophe, the demand for quick response surges, and traditional site selection methods often struggle to meet this demand. For temporary shelters that require quick response, site selection needs to consider not only safety and accessibility but also the flexibility of resource scheduling and the possibility of rapid deployment. However, the site selection of such temporary shelters is currently not well considered, lacking effective technical support and decision-making methods. SUMMARY

[0005] The present application aims to provide a temporary emergency shelter site selection method and system suitable for a catastrophe scenario to improve the above technical problems.

[0006] To achieve the above application purpose, the embodiments of the present application provide the following technical solutions:

[0007] A temporary emergency shelter site selection method suitable for a catastrophe scenario, comprising:

[0008] S1, collecting disaster data of a disaster area in real time and preprocessing;

[0009] S2, grid division is carried out on the preprocessed disaster data to generate corresponding grid data;

[0010] S3, obtaining remote sensing image data and classifying to obtain corresponding classification results; wherein, the remote sensing image data includes pre-disaster remote sensing image data and post-disaster remote sensing image data corresponding to the disaster area, and the classification results include disaster and non-disaster;

[0011] S4, based on the classification results, the grid data is removed to obtain the removed grid data, that is, the non-disaster grid data;

[0012] S5, risk assessment is carried out on the non-disaster grid data to obtain corresponding risk assessment value;

[0013] S6, screening the non-disaster grid data based on the risk assessment value;

[0014] S7, open degree and traffic assessment is carried out on the screened grid data to obtain the evaluated grid data;

[0015] S8, using multi-objective grey wolf algorithm to process the evaluated grid data to obtain the temporary emergency shelter site selection scheme.

[0016] Further, the disaster data includes topographic data, geological data, road data, building data, weather condition data; the topographic data includes slope, slope direction, elevation, river channel; the geological data includes soil type, rock structure, fault distribution, underground water level; the road data includes road network information, road bearing capacity information, road slope, road width, road traffic state, bridge and tunnel distribution information; the building data includes building distribution information, hospital location information, emergency reserve warehouse location information; the weather condition data includes rainfall, temperature, humidity, wind speed, air pressure.

[0017] Further, the grid division is carried out on the preprocessed disaster data to generate corresponding grid data, comprising:

[0018] determining a coordinate system; based on the coordinate system, using the fishnet analysis of ArcGIS software to process the preprocessed disaster data to generate corresponding initial grid data; using the spatial connection function of ArcGIS software to connect the initial grid data, giving the initial grid data corresponding grid space data information, to obtain the grid data.

[0019] Further, the remote sensing image data is acquired and classified to obtain a corresponding classification result, comprising:

[0020] S3-1, collecting disaster area corresponding pre-disaster remote sensing image data and post-disaster remote sensing image data through remote sensing satellite;

[0021] S3-2, calibrating and denoising the pre-disaster remote sensing image data and the post-disaster remote sensing image data to obtain processed pre-disaster remote sensing image data and processed post-disaster remote sensing image data;

[0022] S3-3, based on the grid data, the processed pre-disaster remote sensing image data and the processed post-disaster remote sensing image data are divided to obtain corresponding pre-disaster patch data block and post-disaster patch data block;

[0023] S3-4, inputting the pre-disaster patch data block and the post-disaster patch data block into the Transformer model for classification, and outputting to obtain a corresponding classification result.

[0024] Further, the grid risk model is used to evaluate the risk of the non-disaster grid data; the grid risk model comprises an output layer, a convolutional neural network, a long short-term memory network, an attention layer, a fully connected layer and an output layer connected in turn;

[0025] The training process of the grid risk model comprises the following steps:

[0026] S5-1, obtaining historical disaster data and preprocessing to obtain preprocessed historical disaster data;

[0027] S5-2, using the same method as S2 to S4 to process the preprocessed historical disaster data to obtain corresponding historical non-disaster grid data;

[0028] S5-3, inputting the historical non-disaster grid data into the convolutional neural network to output corresponding historical disaster spatial feature data;

[0029] S5-4, inputting the historical disaster spatial feature data into the long short-term memory network to output historical disaster spatio-temporal feature data;

[0030] S5-5, inputting the historical disaster spatio-temporal feature data into the attention layer and the fully connected layer to output corresponding disaster attention score;

[0031] S5-6, calculating the disaster attention score using a softmax function to obtain normalized disaster attention weight;

[0032] S5-7, based on the normalized disaster attention weight, performing weighted summation on the historical disaster spatio-temporal feature data to obtain weighted summation disaster feature data;

[0033] S5-8, inputting the weighted summation disaster feature data into the output layer to output a corresponding historical risk assessment value; wherein the output layer adopts a softmax function;

[0034] S5-9, calculating a cross-loss function based on the historical risk assessment value;

[0035] S5-10, based on the cross-loss function, performing gradient descent optimization on the grid risk model using Adam optimization to adjust the weight parameters corresponding to the grid risk model to obtain a trained grid risk model.

[0036] Further, the screening of the non-disaster grid data based on the risk assessment value comprises:

[0037] According to the elimination condition:

[0038]

[0039] processing the non-disaster grid data to obtain an availability value S i of the i-th non-disaster grid data, deleting the non-disaster grid data with the availability value S i of 0, retaining the non-disaster grid data with the availability value S i of 1, and obtaining screened grid data; wherein R i,j represents a risk assessment value of the j-th risk factor in the i-th non-disaster grid data, and threshold represents a risk threshold.

[0040] Further, the openness and traffic assessment of the screened grid data to obtain a corresponding assessment result comprises:

[0041] S7-1, calculating the openness of the screened grid data;

[0042] S7-2, sorting the openness according to a descending order, and selecting the grid corresponding to the top 40% openness to obtain twice-screened grid data;

[0043] S7-3, calculating the traffic accessibility of the twice-screened grid data;

[0044] S7-4, sorting the traffic accessibility according to a descending order, and selecting the grid corresponding to the top 40% traffic accessibility to obtain the evaluated grid data.

[0045] Furthermore, based on the evaluation results, the filtered raster data is processed using the multi-objective gray wolf algorithm to obtain a temporary emergency shelter site selection scheme, including:

[0046] S8-1. Set three objectives; wherein, the three objectives include the ease of material allocation, the maximization of rescue coverage, and the openness of the site;

[0047] S8-2. Randomly select the filtered raster data to generate a gray wolf population; wherein, each gray wolf in the gray wolf population is a candidate refuge site; the position of each gray wolf is a combination of several filtered raster cells;

[0048] S8-3. Based on the aforementioned three objectives, calculate the resource allocation convenience score, rescue coverage score, and site openness score corresponding to the gray wolf population.

[0049] S8-4. Based on the resource allocation convenience score, the rescue coverage score, and the site openness score, the gray wolf population is selected using random perturbation to obtain α wolf, β wolf, and δ wolf.

[0050] S8-5. Randomly select a gray wolf individual from the gray wolf population;

[0051] S8-6. Based on the positions of the α wolf, β wolf, and δ wolf, calculate the distance between the selected gray wolf individual and the α wolf, β wolf, and δ wolf;

[0052] S8-7. Based on the distance, update the center position of the selected gray wolf individual;

[0053] S8-8. Based on the intermediate position, update the position of the corresponding selected gray wolf individual;

[0054] S8-9, repeat S8-2 to S8-8 until the number of iterations is reached to obtain the optimal grid, which is the temporary emergency shelter site selection scheme.

[0055] Furthermore, the formula for the resource allocation convenience score corresponding to the gray wolf population is:

[0056] SA i =W1·C dts +W2·C dttn ;

[0057] Among them, SA i This represents the resource allocation convenience score for the i-th individual gray wolf. W1 and W2 represent the weights of the filtered raster data and resource storage points, and the weights of the filtered raster data and transportation nodes, respectively. C dtsrepresents the distance score of the screened grid data corresponding to the i-th gray wolf individual and the material reserve point, C dttn represents the distance score of the screened grid data corresponding to the i-th gray wolf individual and the traffic node, C

[0058] The formula of the rescue coverage score is:

[0059] RC i = W3·C pd + W4·C dth + W5·C dtec

[0060] Wherein, RC i represents the rescue coverage score of the i-th gray wolf individual, W3, W4, W5 respectively represent the rescue coverage weight, the population density weight and the weight of the screened grid data and the medical facility, C pd represents the rescue coverage score of the disaster area, C dth represents the population density score of the disaster area, C dtec represents the distance score of the screened grid data corresponding to the i-th gray wolf individual and the medical facility, C

[0061] The formula of the site openness score is:

[0062]

[0063] Wherein, E i represents the site openness score of the i-th gray wolf individual, A usable,i represents the available open area in the screened grid data corresponding to the i-th gray wolf individual, A i represents the total area of the screened grid data corresponding to the i-th gray wolf individual, S i represents the terrain flatness score of the screened grid data corresponding to the i-th gray wolf individual.

[0064] A temporary emergency shelter site selection system suitable for a disaster scenario, comprising a disaster data acquisition and processing module, a grid division module, an image classification module, a screening module, an evaluation module, a temporary emergency shelter site selection module; wherein:

[0065] The disaster data acquisition and processing module is used for real-time acquisition and preprocessing of disaster data in the disaster area;

[0066] The grid division module is used for grid division of the preprocessed disaster data to generate corresponding grid data;

[0067] The image classification module is used for obtaining remote sensing image data and classifying to obtain corresponding classification results;

[0068] The screening module is configured to perform elimination on the grid data based on the classification result to obtain eliminated grid data, i.e., non-affected grid data, and perform screening on the non-affected grid data based on the risk evaluation value;

[0069] The evaluation module is configured to perform risk evaluation on the non-affected grid data to obtain a corresponding risk evaluation value, and perform openness and traffic evaluation on the screened grid data to obtain evaluated grid data.

[0070] The temporary emergency shelter site selection module is configured to process the evaluated grid data by using a multi-objective grey wolf algorithm to obtain a temporary emergency shelter site selection scheme.

[0071] The present application has the following advantages:

[0072] The present method processes disaster data and remote sensing image data of an affected area, identifies affected grids and non-affected grids in the affected area, narrows the site selection range, and further improves site selection accuracy and precision. The present method performs risk evaluation, openness and traffic evaluation on the non-affected grids, performs site selection on the non-affected grids based on the evaluation result by using a grey wolf algorithm, obtains a temporary emergency shelter site selection scheme, reduces dependence on manual analysis through data driving, improves site selection automation degree, improves efficiency, further ensures the safety of residents, dynamically responds to changing demands through integration and rapid analysis of multi-source data, enhances dynamicity, and ensures that emergency service coverage matches actual demand by updating site selection strategies in real time.

[0073] The present system has a simple structure, can identify and evaluate disaster data and remote sensing image data of an affected area, can reduce cost, improve site selection accuracy, ensure reasonable distribution of temporary emergency shelters, and reduce resource waste and uneven service problems. BRIEF DESCRIPTION OF DRAWINGS

[0074] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope, and other related drawings can also be obtained by those skilled in the art without creative labor on the premise of not paying.

[0075] Figure 1 The present application has the following advantages:

[0076] Figure 2 The present application has the following advantages: DETAILED DESCRIPTION

[0077] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. The components of the embodiments of the present application described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0078] Referring to Figure 1 The embodiment provides a temporary emergency shelter site selection method suitable for a catastrophe scenario, which comprises the following steps:

[0079] S1, collecting disaster data of a disaster area in real time and preprocessing;

[0080] The disaster data comprises terrain data, geological data, road data, building data and weather condition data; the terrain data comprises slope, slope direction, elevation and river channel; the geological data comprises soil type, rock structure, fault distribution and underground water level; the road data comprises road network information, road bearing capacity information, road slope, road width, road traffic state and bridge and tunnel distribution information; the building data comprises building distribution information, hospital location information and emergency reserve warehouse location information; the weather condition data comprises rainfall, temperature, humidity, wind speed and air pressure.

[0081] The preprocessing comprises format arrangement, missing value interpolation, abnormal value processing and data coding. The preprocessed disaster data can be connected to a geographic information system to realize visualization.

[0082] Mountain area A is selected as an experimental object. Mountain area A is affected by a tropical low pressure, and has continuous rainstorms for 3 days, with a cumulative rainfall of 450mm, far exceeding the average value (200mm) of the same period in many years. The strong rainfall causes landslides at multiple points. Among them, the scale of part of the landslides exceeds 100,000 cubic meters, directly destroys 120 village houses, blocks 2 main roads, 3 bridges, interrupts communication facilities, and about 30,000 people are affected.

[0083] S2, performing grid division on the preprocessed disaster data to generate corresponding grid data;

[0084] The grid division on the preprocessed disaster data to generate corresponding grid data comprises:

[0085] Determine a coordinate system; based on the coordinate system, process the pre-processed disaster data using a fishing net analysis of ArcGIS software to generate corresponding initial grid data; use the spatial connection function of the ArcGIS software to connect the data of the initial grid data, assign the corresponding grid spatial data information to the initial grid data, and obtain the grid data. Wherein, the coordinate system can be WGS1984World Mercator.

[0086] S3, obtain remote sensing image data and classify to obtain corresponding classification results; wherein, the remote sensing image data includes pre-disaster remote sensing image data and post-disaster remote sensing image data corresponding to the disaster area, and the classification results include disaster and non-disaster;

[0087] The remote sensing image data is obtained and classified to obtain the corresponding classification results, comprising:

[0088] S3-1, collect the pre-disaster remote sensing image data and the post-disaster remote sensing image data corresponding to the disaster area through remote sensing satellites (such as Sentinel-2, Landsat, etc.);

[0089] S3-2, calibrate and denoise the pre-disaster remote sensing image data and the post-disaster remote sensing image data to obtain pre-disaster processed remote sensing image data and post-disaster processed remote sensing image data;

[0090] S3-3, based on the grid data, divide the pre-disaster processed remote sensing image data and the post-disaster processed remote sensing image data to obtain corresponding pre-disaster patch data blocks and post-disaster patch data blocks; each pre-disaster patch data block or post-disaster patch data block is mapped to a feature vector, and the corresponding expression is:

[0091] x i =Embedding(patch i );

[0092] Wherein, x i represents the feature vector corresponding to the i-th pre-disaster patch data block or post-disaster patch data block, Embedding(·) represents an image embedding function, and patch i represents the i-th pre-disaster patch data block or post-disaster patch data block.

[0093] S3-4, input the pre-disaster patch data block and the post-disaster patch data block into the Transformer model for classification, and output to obtain the corresponding classification results.

[0094] Before the pre-disaster patch data block and the post-disaster patch data block are input into the Transformer model, the pre-disaster patch data block and the post-disaster patch data block need to be positionally encoded to retain the corresponding spatial information.

[0095] The Transformer model includes an encoder and a decoder, and the encoder and the decoder adopt a multi-head attention mechanism. The pre-disaster patch data block and the post-disaster patch data block after position encoding are input into the encoder, and the significant area of the surface change is detected through the multi-head self-attention mechanism, that is, each feature vector x i corresponding query vector, key vector and value vector; based on the query vector, key vector and value vector, each feature vector x i corresponding attention score, and the corresponding expression is:

[0096]

[0097] wherein Q, K and V represent the query vector, the key vector and the value vector respectively, W Q , W K and W V represent the weight matrix corresponding to the query vector Q, the key vector K and the value vector K respectively, softmax(·) represents an activation function, and Attention(Q, K, V) represents the attention score corresponding to the feature vector x i , d k and d T represent the dimension of the key vector K, and K T represents the transpose matrix of the key vector K.

[0098] Based on the attention score, the feature vector x i is encoded to obtain the encoded feature data.

[0099] The encoded feature data is input into the decoder for decoding. The same multi-head self-attention mechanism as the encoder is adopted to process the encoded feature data to obtain decoded feature data; the decoded feature data is input into an output layer to obtain a classification result; when the classification result is 1, it indicates that the grid is not affected; when the classification result is 0, it indicates that the grid is affected.

[0100] S4, based on the classification result, the grid data is removed to obtain removed grid data, that is, the unaffected grid data; the grid data corresponding to the classification result of 0 is deleted, and the grid data corresponding to the classification result of 1 is retained to obtain the unaffected grid data.

[0101] Real-time acquisition of disaster area topography, geology, traffic, building, weather data, all data are converted into GeoTIFF format, and the pretreated disaster data of the disaster area are formed after pretreatment. A unified coordinate system (such as WGS1984WorldMercator) is specified, and a grid (500m x 500m) is constructed using the fishnet analysis of ArcGIS software, and the grid spatial data information is given through the spatial connection function, and the grid data are obtained. The grid data are shown in Table 1.

[0102] Table 1

[0103] Grid ID Average elevation / m ... Maximum elevation / m Minimum elevation / m Slope / ° Soil type 001 100 ... 200 73 12.5 Sandy soil 002 150 ... 158 147 20.3 Sandy soil 003 80 ... 178 20 8.7 Sandy soil ... ... ... ... ... ... ...

[0104] Disaster area pre-disaster remote sensing image data and post-disaster remote sensing image data are obtained from remote sensing satellites, and calibration and denoising processing are performed. Based on the deep Transformer model, each image is divided into 100 small blocks (patches) in the grid, and position encoding is added to retain spatial information. Using the Transformer encoder layer, the significant area of the ground surface change is detected through the multi-head self-attention mechanism, and the output layer is classified as "1: not affected" and "0: affected", as shown in Table 2, and the grid data with more than 20% of the affected patches are removed.

[0105] Table 2

[0106] Grid ID Number of affected plots Number of unaffected plots Grid status 001 80 20 Affected 002 15 85 Unaffected 003 10 90 Unaffected … … … …

[0107] S5, risk assessment is performed on the non-affected grid data to obtain corresponding risk assessment values;

[0108] The grid risk model is used to perform risk assessment on the non-affected grid data; the grid risk model comprises an output layer, a convolutional neural network, a long short-term memory network, an attention layer, a fully connected layer and an output layer connected in sequence;

[0109] The training process of the grid risk model comprises the following steps:

[0110] S5-1, historical disaster data are obtained from government, meteorological bureau, geological monitoring agency and other channels and pretreated to obtain pretreated historical disaster data;

[0111] S5-2, the pretreated historical disaster data are processed using the same method as S2 to S4 to obtain corresponding historical non-affected grid data;

[0112] S5-3, the historical non-affected grid data are input into the convolutional neural network, and the corresponding historical disaster spatial feature data are output;

[0113] S5-4, input the historical disaster spatial feature data into the long short-term memory network, and output historical disaster spatio-temporal feature data;

[0114] S5-5, input the historical disaster spatio-temporal feature data into the attention layer and the fully connected layer, and output corresponding disaster attention scores; the fully connected layer adopts a Tanh activation function;

[0115] S5-6, calculate the disaster attention scores to obtain normalized disaster attention weights;

[0116] S5-7, based on the normalized disaster attention weights, perform weighted summation on the historical disaster spatio-temporal feature data to obtain weighted summation disaster feature data;

[0117] S5-8, input the weighted summation disaster feature data into the output layer, and output corresponding historical risk assessment values; wherein the output layer adopts a softmax function;

[0118] The formulas corresponding to S5-5 to S5-8 are as follows:

[0119] e t =tanh(W a *h t +b a );

[0120]

[0121] μ=softmax(w o *z+b o );

[0122] wherein e t represents a disaster attention score corresponding to a time step t of historical disaster spatio-temporal feature data, W a represents a weight matrix of the attention layer, and is a weight given to a hidden state at different time steps, h t represents a hidden state of a time step t of historical disaster spatio-temporal feature data, b a represents a bias term, tanh(·) represents an activation function, a t represents a normalized disaster attention weight, T' represents a total number of historical disaster spatio-temporal feature data, exp(·) represents a logarithmic function with a natural constant e as a base, e i represents an i-th disaster attention score, z represents an attention weighted output, and is a weighted average of important information in a time sequence, μ represents a historical risk assessment value, w o represents a weight matrix of the output layer, and b o represents a bias term of the output layer.

[0123] S5-9, calculate a cross-entropy loss function based on the historical risk assessment values, and the corresponding formula is:

[0124]

[0125] wherein L represents the historical risk assessment value calculated cross-entropy loss function value, C represents the total number of risk levels, μ' i represents the i-th historical risk assessment value output by the grid risk model, μ i represents the i-th historical risk assessment value μ' i corresponding to the actual label of the i-th class value, and log(·) represents a logarithmic function.

[0126] S5-10, based on the cross-entropy loss function, using Adam optimization zone to gradient descent optimization of the grid risk model, adjust the corresponding weight parameters of the grid risk model, that is, adjust W a , b a , w o , b o , to obtain the trained grid risk model.

[0127] Collect the historical disaster data corresponding to mountainous area A, and divide it according to 7:2:1 to obtain the training set, the validation set and the test set.

[0128] Input the training set into the grid risk model. Input the training set (slope, elevation, soil type, river distribution, etc.) into the convolutional neural network (CNN), output the spatial feature map of each time step, that is, the historical disaster spatial feature data, and the corresponding resolution is 100m x 100m; input the historical disaster spatial feature data extracted by the CNN into the LSTM network, output the landslide risk level prediction (low, medium, high); the attention mechanism weight is distributed to the time step of the rainfall peak period, and the high slope and the area near the fault zone are weighted, to improve the prediction ability of the model to the key period; wherein the initial learning rate of the grid risk model is set to 0.001, the cosine annealing scheduler is used to gradually reduce the learning rate, the Adam optimizer is used for gradient descent, and the prediction accuracy is improved by iteration optimization. When the AUC value of the validation set is stable at 0.95 or more, the corresponding grid risk model is packaged.

[0129] S6, screening the non-disaster grid data based on the risk assessment value;

[0130] The screening of the non-disaster grid data based on the risk assessment value comprises:

[0131] According to the elimination condition:

[0132]

[0133] Processing the non-affected grid data to obtain an availability value S corresponding to the i th non-affected grid data i ; when the availability value S i is 1, it means that the score of any risk factor does not exceed the threshold, that is, R i,j < threshold for all j, the corresponding non-affected grid data is retained; when the availability value S i is 0, it means that the score of any risk factor reaches or exceeds the threshold, that is, R i,j ≥ threshold for all j, the corresponding non-affected grid data is deleted.

[0134] Delete the non-affected grid data with the availability value S i 0, retain the non-affected grid data with the availability value S i 1, to obtain the screened grid data; wherein, R i,j represents the risk evaluation value of the j th risk factor in the i th non-affected grid data, and threshold represents the risk threshold.

[0135] S7, open and traffic evaluation is carried out on the screened grid data to obtain the evaluated grid data;

[0136] The open and traffic evaluation on the screened grid data obtains the corresponding evaluation result, which includes:

[0137] S7-1, the open of the screened grid data is calculated, and the corresponding formula is:

[0138]

[0139]

[0140] Among them, represents the open of the i th screened grid data, represents the available open area (i.e. the area excluding buildings, trees, water bodies, etc.) in the i th screened grid data, represents the total area of the i th screened grid data, represents the terrain flatness score of the i th screened grid data, represents the average slope value of the screened grid data, represents the maximum slope value of the screened grid data;

[0141] S7-2, according to the order from large to small, the open is sorted, and the grid corresponding to the top 40% open is selected to obtain the twice screened grid data;

[0142] S7-3, calculate the traffic accessibility of the second screened grid data, and the corresponding formula is:

[0143]

[0144] wherein, T n represents the traffic accessibility of the nth second screened grid data, w1 and w2 represent weights, N represents the number of roads in the second screened grid data, represents the connection score of the nth road, represents the score of the distance of the nth road from the traffic length.

[0145] Connection score and score The corresponding formula is:

[0146]

[0147] wherein, m represents the total number of edges of the road network, represents the distance of the nth road from the traffic length.

[0148] S7-4, according to the order from large to small, the traffic accessibility is sorted, and the grid corresponding to the top 40% traffic accessibility is selected to obtain the evaluated grid data.

[0149] S8, using a multi-objective grey wolf algorithm to process the evaluated grid data to obtain a temporary emergency shelter site selection scheme.

[0150] The temporary emergency shelter site selection scheme is obtained by using a multi-objective grey wolf algorithm to process the screened grid data based on the evaluation result, comprising:

[0151] S8-1, setting three objectives; wherein the three objectives include material allocation convenience, maximum rescue coverage, and site openness; the material allocation convenience is to make the selected grid position convenient for rapid allocation and transportation of materials; the maximum rescue coverage is to ensure that the selected grid position covers more disaster-affected people and is close to key resources such as medical emergency facilities; the site openness ensures that the selected grid position has enough open space and flat terrain to accommodate a large number of refugees and reduce secondary disaster risks.

[0152] S8-2, randomly selecting the screened grid data to generate a grey wolf population; wherein each grey wolf individual in the grey wolf population is a candidate emergency shelter scheme; the position of each grey wolf individual is a combination of a plurality of screened grids;

[0153] S8-3, based on the three objectives, calculate the material allocation convenience score, the rescue coverage score, and the site openness score corresponding to the grey wolf population;

[0154] The formula of the material allocation convenience score corresponding to the gray wolf population is:

[0155] SA i = W1·C dts + W2·C dttn ;

[0156] Wherein, SA i represents the material allocation convenience score of the i-th gray wolf individual, W1 and W2 represent the weight of the screened grid data and the material reserve point, and the weight of the screened grid data and the traffic node respectively, C dts represents the distance score of the i-th gray wolf individual corresponding to the screened grid data and the material reserve point, the closer to the material reserve point, the higher the corresponding distance score C dts , C dttn represents the distance score of the i-th gray wolf individual corresponding to the screened grid data and the traffic node (main traffic node), the closer to the traffic node, the higher the corresponding distance score C dttn ;

[0157] The formula of the rescue coverage score is:

[0158] RC i = W3·C pd + W4·C dth + W5·C dtec ;

[0159] Wherein, RC i represents the rescue coverage score of the i-th gray wolf individual, W3, W4 and W5 represent the rescue coverage weight, the population density weight and the weight of the screened grid data and the medical facility respectively, C pd represents the rescue coverage score of the disaster area, C dth represents the population density score of the disaster area, C dtec represents the distance score of the i-th gray wolf individual corresponding to the screened grid data and the medical facility, the closer to the medical facility, the higher the corresponding distance score C dtec .

[0160] Wherein, the distance score C dts , the distance score C dttn , the distance score C dtec The corresponding formula is:

[0161]

[0162] Wherein, L dts represents the distance of the gray wolf individual corresponding to the screened grid data and the material reserve point, L dttnL represents the distance between the screened grid data corresponding to the gray wolf individual and the traffic node (main traffic node) dtec L represents the distance between the screened grid data corresponding to the gray wolf individual and the medical facility, both in kilometers.

[0163] Rescue coverage score C pd And the population density score C dth The corresponding formula is:

[0164] C pd = S / E;

[0165] C dth =1 / E;

[0166] Wherein, S represents the overlap of the disaster area surface and the emergency disaster area, E represents the service range of the emergency point with R (generally 5-10) kilometers as the radius.

[0167] The formula of the site openness score is:

[0168]

[0169] Wherein, E i represents the site openness score of the i-th gray wolf individual, A usable,i represents the available open area in the screened grid data corresponding to the i-th gray wolf individual, A i represents the total area of the screened grid data corresponding to the i-th gray wolf individual, S i represents the terrain flatness score of the screened grid data corresponding to the i-th gray wolf individual, slope represents the average slope value of the screened grid data corresponding to the i-th gray wolf individual, slope max represents the maximum slope value of the screened grid data corresponding to the i-th gray wolf individual.

[0170] S8-4, based on the material allocation convenience score, the rescue coverage score and the site openness score, the gray wolf population is selected by random disturbance, that is, the total score corresponding to the material allocation convenience score, the rescue coverage score and the site openness score is selected as the top three gray wolf individuals, and α wolf, β wolf and δ wolf are obtained. Wolf is the gray wolf individual corresponding to the highest total score, representing the current optimal site selection scheme, and its position is The corresponding formula is β wolf is the gray wolf individual corresponding to the second highest total score, representing the current suboptimal site selection scheme, and its position is The corresponding formula is δ wolf is the gray wolf individual corresponding to the third total score, representing the current third site selection scheme, and its position is The corresponding formula is wherein, represents the position of the alpha wolf before random disturbance, represents the position of the beta wolf before random disturbance, represents the position of the delta wolf before random disturbance, all represent random vectors, usually randomly generated within the range [-0.00001, 0.00001].

[0171] S8-5, randomly selecting a gray wolf individual in the gray wolf population;

[0172] S8-6, calculating the distance between the selected gray wolf individual and the alpha wolf, the beta wolf and the delta wolf according to the positions of the alpha wolf, the beta wolf and the delta wolf, and the corresponding formula is:

[0173]

[0174] wherein, respectively represent the distance vectors of the selected gray wolf individual and the alpha wolf, the beta wolf and the delta wolf, represents a random coefficient vector, respectively represent the positions of the alpha wolf, the beta wolf and the delta wolf, represents the position of the selected gray wolf individual; the random coefficient vector is used to control the range of update.

[0175] S8-7, updating the intermediate position of the selected gray wolf individual based on the distance, and the corresponding formula is:

[0176]

[0177] wherein, respectively represent the intermediate positions of the selected gray wolf individual and the alpha wolf, the beta wolf and the delta wolf, represents an adjustment factor vector, represents; the intermediate position is the intermediate position of the selected gray wolf individual and the alpha wolf, the beta wolf and the delta wolf. Generally, the adjustment factor vector will gradually decrease with the increase of the number of iterations.

[0178] S8-8, updating the position of the corresponding selected gray wolf individual based on the intermediate position;

[0179] S8-9, repeating S8-2 to S8-8 until the number of iterations is reached, and obtaining the optimal grid, i.e. obtaining the temporary emergency shelter site selection scheme.

[0180] In the multi-objective optimization process, the grey wolf algorithm is used to generate a set of candidate solutions in each iteration, and the non-dominated solution is selected from the candidate solutions, that is, the non-dominated solution is the optimal solution in this set of candidate solutions. The non-dominated solution in the current iteration is the Pareto front solution. With the iteration, the non-dominated solution in the solution set will constantly approach the real Pareto front, and the final solution of the Pareto front is the optimal non-dominated solution in all iteration results, that is, the temporary emergency shelter site selection scheme. The use of Pareto front solution can make the selected grid scheme balanced optimization in multiple objectives such as material allocation convenience, maximum rescue coverage and site openness.

[0181] As shown in Figure 2 A temporary emergency shelter site selection system suitable for catastrophe scenarios includes a disaster data acquisition and processing module, a grid division module, an image classification module, a screening module, an evaluation module, and a temporary emergency shelter site selection module. Wherein:

[0182] The disaster data acquisition and processing module is used for real-time acquisition and preprocessing of disaster data in the disaster area.

[0183] The grid division module is used for grid division of the preprocessed disaster data to generate corresponding grid data.

[0184] The image classification module is used to obtain remote sensing image data and classify it to obtain the corresponding classification result.

[0185] The screening module is used to exclude the grid data based on the classification result to obtain the excluded grid data, that is, the non-disaster grid data; and to screen the non-disaster grid data based on the risk evaluation value.

[0186] The evaluation module is used to evaluate the risk of non-disaster grid data to obtain the corresponding risk evaluation value; and to evaluate the openness and traffic of the screened grid data to obtain the evaluated grid data.

[0187] The temporary emergency shelter site selection module is used to process the evaluated grid data using a multi-objective grey wolf algorithm to obtain a temporary emergency shelter site selection scheme.

[0188] To sum up, the method of the present application processes disaster data and remote sensing image data of the disaster area, identifies disaster grids and non-disaster grids in the disaster area, narrows down the site selection range, and further improves the site selection accuracy and precision; the non-disaster grids are evaluated in terms of risk, openness and traffic, based on the evaluation results, the grey wolf algorithm is used to select sites for the non-disaster grids, a site selection scheme for temporary emergency shelters is obtained, the dependence on manual analysis is reduced through data-driven, the automation degree of site selection is improved, the demand for on-site investigation is reduced, and the efficiency is improved, through the integration and rapid analysis of multi-source data, the dynamic response to changing needs is enhanced, the dynamic nature is enhanced, the site selection strategy is updated in real time through real-time data, and the emergency service coverage is ensured to match the actual demand. The system structure is simple, the disaster data and remote sensing image data of the disaster area are identified and evaluated, the cost can be reduced, the accuracy of site selection can be improved, the reasonable distribution of temporary emergency shelters can be ensured, and the problems of resource waste and uneven service can be reduced.

[0189] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for site selection of temporary emergency shelter suitable for catastrophe scenario, characterized in that, The method comprises the following steps: S1, collecting disaster data of a disaster area in real time and preprocessing the disaster data; S2, performing grid division on the preprocessed disaster data to generate corresponding grid data; S3, acquiring remote sensing image data and classifying the remote sensing image data to obtain corresponding classification results; wherein the remote sensing image data comprises pre-disaster remote sensing image data and post-disaster remote sensing image data corresponding to the disaster area, and the classification results comprise disaster-affected and non-disaster-affected; S4, based on the classification results, removing the grid data to obtain removed grid data, i.e. non-disaster-affected grid data; S5, performing risk assessment on the non-disaster-affected grid data to obtain corresponding risk assessment values; S6, screening the non-disaster-affected grid data based on the risk assessment values; S7, performing openness and traffic assessment on the screened grid data to obtain evaluated grid data; S8, processing the evaluated grid data by using a multi-objective grey wolf algorithm to obtain a temporary emergency shelter site selection scheme.

2. The method for site selection of temporary emergency shelter applicable in catastrophe scenario as claimed in claim 1, wherein, The disaster data comprises terrain data, geological data, road data, building data, and weather condition data; the terrain data comprises slope, slope direction, elevation, and river channel; the geological data comprises soil type, rock structure, fault distribution, and underground water level; the road data comprises road network information, road bearing capacity information, road slope, road width, road traffic state, and bridge and tunnel distribution information; the building data comprises building distribution information, hospital location information, and emergency reserve warehouse location information; the weather condition data comprises rainfall, temperature, humidity, wind speed, and air pressure.

3. The method for site selection of temporary emergency shelter applicable in catastrophe scenario as claimed in claim 1, wherein, The grid division on the preprocessed disaster data to generate corresponding grid data comprises the following steps: determining a coordinate system; based on the coordinate system, processing the preprocessed disaster data by using a fishnet analysis of ArcGIS software to generate corresponding initial grid data; and connecting the initial grid data by using a spatial connection function of the ArcGIS software to assign corresponding grid spatial data information to the initial grid data, thereby obtaining grid data.

4. The method for site selection of temporary emergency shelter applicable in catastrophe scenario as claimed in claim 1, wherein, The acquisition of remote sensing image data and the classification of the remote sensing image data to obtain corresponding classification results comprise the following steps: S3-1, acquiring pre-disaster remote sensing image data and post-disaster remote sensing image data corresponding to the disaster area by using a remote sensing satellite; S3-2, calibrating and denoising the pre-disaster remote sensing image data and the post-disaster remote sensing image data to obtain processed pre-disaster remote sensing image data and processed post-disaster remote sensing image data; S3-3, based on the grid data, dividing the processed pre-disaster remote sensing image data and the processed post-disaster remote sensing image data to obtain corresponding pre-disaster patch data blocks and post-disaster patch data blocks; S3-4, inputting the pre-disaster patch data blocks and the post-disaster patch data blocks into a Transformer model for classification to output corresponding classification results.

5. The method for site selection of temporary emergency shelter applicable in catastrophe scenario as claimed in claim 1, wherein, The non-disaster-affected grid data is assessed by using a grid risk model; the grid risk model comprises output layers, a convolutional neural network, a long short-term memory network, an attention layer, a fully connected layer, and output layers connected in sequence; The training process of the grid risk model comprises the following steps: S5-1, historical disaster data is obtained and preprocessed to obtain preprocessed historical disaster data; S5-2, the preprocessed historical disaster data is processed by using the same method as S2 to S4 to obtain corresponding historical non-disaster grid data; S5-3, the historical non-disaster grid data is input into the convolutional neural network to output corresponding historical disaster spatial feature data; S5-4, the historical disaster spatial feature data is input into the long short-term memory network to output historical disaster spatio-temporal feature data; S5-5, the historical disaster spatio-temporal feature data is input into the attention layer and the fully connected layer to output corresponding disaster attention scores; S5-6, the disaster attention scores are calculated by using a softmax function to obtain normalized disaster attention weights; S5-7, based on the normalized disaster attention weights, the historical disaster spatio-temporal feature data is weighted and summed to obtain weighted and summed disaster feature data; S5-8, the weighted and summed disaster feature data is input into the output layer to output corresponding historical risk assessment values; wherein the output layer uses a softmax function; S5-9, a cross-entropy loss function is calculated based on the historical risk assessment values; S5-10, based on the cross-entropy loss function, the grid risk model is optimized by gradient descent using Adam optimization to adjust the corresponding weight parameters of the grid risk model to obtain a trained grid risk model.

6. The method for site selection of temporary emergency shelter applicable in catastrophe scenario as claimed in claim 1, wherein, The screening of non-disaster grid data based on the risk assessment values comprises: According to the elimination condition: The non-disaster grid data is processed to obtain an availability value Si corresponding to the i th non-disaster grid data; the non-disaster grid data with an availability value Si of 0 is deleted, and the non-disaster grid data with an availability value Si of 1 is retained to obtain screened grid data; wherein Ri,j represents the risk assessment value of the j th risk factor in the i th non-disaster grid data, and threshold represents a risk threshold.

7. The method for site selection of temporary emergency shelter applicable in catastrophe scenario as claimed in claim 1, wherein, The open degree and traffic assessment of the screened grid data to obtain the corresponding assessment results comprises: S7-1, the open degree of the screened grid data is calculated; S7-2, the open degree is sorted in descending order, and the grid corresponding to the top 40% open degree is selected to obtain secondary screened grid data; S7-3, the traffic accessibility of the secondary screened grid data is calculated; S7-4, the traffic accessibility is sorted in descending order, and the grid corresponding to the top 40% traffic accessibility is selected to obtain the evaluated grid data.

8. The method for site selection of temporary emergency shelter applicable in catastrophe scenario as claimed in claim 1, wherein, Based on the evaluation results, the screened grid data is processed by using a multi-objective grey wolf algorithm to obtain a temporary emergency shelter site selection scheme, comprising: S8-1, three objectives are set; wherein the three objectives include material allocation convenience, maximum rescue coverage, and site openness; S8-2, randomly select the screened grid data to generate a gray wolf population; wherein each gray wolf individual in the gray wolf population is taken as a candidate refuge site scheme; and the position of each gray wolf individual is a combination of a plurality of screened grid data; S8-3, based on the triple target, calculate the material allocation convenience score, the rescue coverage score and the site openness score corresponding to the gray wolf population; S8-4, based on the material allocation convenience score, the rescue coverage score and the site openness score, the gray wolf population is selected by random disturbance to obtain alpha wolf, beta wolf and delta wolf; S8-5, randomly select a gray wolf individual in the gray wolf population; S8-6, according to the positions of the alpha wolf, the beta wolf and the delta wolf, calculate the distance between the selected gray wolf individual and the alpha wolf, the beta wolf and the delta wolf; S8-7, based on the distance, update the intermediate position of the selected gray wolf individual; S8-8, based on the intermediate position, update the position of the corresponding selected gray wolf individual; S8-9, repeat S8-2 to S8-8 until the iteration number is reached to obtain the optimal grid, that is, the temporary emergency refuge site selection scheme.

9. The method for site selection of temporary emergency shelter applicable in catastrophe scenario as claimed in claim 8, wherein, The formula of the material allocation convenience score corresponding to the gray wolf population is: SAi=W1·Cdts+W2·Cdttn; Wherein, SAi represents the material allocation convenience score of the i-th gray wolf individual, W1 and W2 represent the weight of the screened grid data and the material reserve point, the weight of the screened grid data and the traffic node respectively, Cdts represents the distance score of the i-th gray wolf individual corresponding to the screened grid data and the material reserve point, Cdttn represents the distance score of the i-th gray wolf individual corresponding to the screened grid data and the traffic node; The formula of the rescue coverage score is: RCi=W3·Cpd+W4·Cdth+W5·Cdtec; Wherein, RCi represents the rescue coverage score of the i-th gray wolf individual, W3, W4 and W5 represent the rescue coverage weight, the population density weight and the weight of the screened grid data and the medical facility respectively, Cpd represents the rescue coverage score of the disaster area, Cdth represents the population density score of the disaster area, and Cdtec represents the distance score of the i-th gray wolf individual corresponding to the screened grid data and the medical facility; The formula of the site openness score is: Wherein, Ei represents the site openness score of the i-th gray wolf individual, Ausable,i represents the available open area in the screened grid data corresponding to the i-th gray wolf individual, Ai represents the total area of the screened grid data corresponding to the i-th gray wolf individual, and Si represents the terrain flatness score of the screened grid data corresponding to the i-th gray wolf individual.

10. A system for site selection of temporary emergency shelter for catastrophe scenario, for implementing the method for site selection of temporary emergency shelter for catastrophe scenario according to any one of claims 1 to 9, characterized in that, It comprises a disaster data collection and processing module, a grid division module, an image classification module, a screening module, an evaluation module and a temporary emergency refuge site selection module; wherein: The disaster data collection and processing module is used for real-time collection and preprocessing of disaster data in the disaster area. The grid division module is configured to perform grid division on the preprocessed disaster data to generate corresponding grid data. The image classification module is configured to acquire remote sensing image data and perform classification to obtain corresponding classification results. The screening module is configured to perform elimination on the grid data based on the classification results to obtain eliminated grid data, i.e., non-disaster grid data, and perform screening on the non-disaster grid data based on risk evaluation values. The evaluation module is configured to perform risk evaluation on the non-disaster grid data to obtain corresponding risk evaluation values. The grid data after screening is subjected to openness and traffic evaluation to obtain evaluated grid data. The temporary emergency shelter site selection module is configured to process the evaluated grid data by using a multi-objective grey wolf algorithm to obtain a temporary emergency shelter site selection scheme.

Images