Emergency rescue priority dynamic decision-making method and system

Abstract

The invention discloses an emergency rescue priority dynamic decision-making method and system, and the method comprises the steps: carrying out the grid division of a disaster region based on the ground surface attachment distribution in the disaster region and DEM digital elevation data, and carrying out the importance weight calculation, dynamic correction and visual coloring of a grid; in combination with a hydrodynamic model, predicting a substance motion trail after a geological disaster occurs, dynamically adjusting a potential influence range of each grid, updating a weight value, and performing secondary grid coloring; calculating a comprehensive signal characteristic value of the multi-source life signal data by adopting a multi-source signal fusion model, performing weight gain processing on a response grid, and performing weight return-to-zero and color reset on a non-response grid; generating a rescue priority sequence through a multi-dimensional decision tree model; the rescue priority and the resource allocation strategy are dynamically adjusted according to the real-time monitoring feedback data; and a Bayesian optimization model is established according to the rescue efficiency index data, and continuous optimization of a rescue strategy is realized through multiple iterations, so that the rescue efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of emergency rescue, more specifically, relates to an emergency rescue priority dynamic decision-making method and system. BACKGROUND

[0002] Geological disasters refer to disasters related to geological processes such as mountain collapse, landslide, debris flow, ground subsidence, ground fissure, and ground subsidence caused by natural factors or human activities that endanger people's lives and property safety. The purpose of geological disaster prevention and control is to reduce the frequency and severity of geological disasters through scientific methods and technical means.

[0003] The existing prevention and control system has many defects: the coverage rate of monitoring stations in remote mountainous areas is less than 30%, resulting in a lack of real-time monitoring data for disasters, making it difficult to grasp the dynamic changes of disasters in a timely manner, affecting the accuracy and timeliness of early warning. Although the Beidou system (BDS) has been applied in pilot projects in Yunnan and Sichuan, data fusion and real-time analysis technology has not been widely used, and it is difficult to fully play its role in disaster monitoring, making it difficult to quickly generate effective early warning information. The existing disaster prediction model still has a prediction error of 15% for disaster diffusion in complex terrain, making it difficult to meet the demand for accurate response. When a disaster occurs, it is difficult to accurately predict the impact range and degree of the disaster, resulting in a lack of targeted allocation of rescue resources. The emergency response mechanism is lagging behind, and the traditional "post-disaster rescue" mode relies on on-site investigation and manual judgment after the disaster, which is inefficient. For example, the Yibin landslide rescue delayed the golden rescue time due to traffic disruption, missing the best rescue opportunity. Resource allocation relies on experience-based decision-making, with a misallocation rate of up to 20%, while the coverage rate of intelligent decision-making systems is less than 40%. This results in rescue resources being unable to be timely and accurately deployed to where they are most needed, affecting the effectiveness of rescue. The coverage rate of emergency training is low, and the public's disaster prevention awareness is weak. Surveys show that less than 50% of residents in the southwest region have received geological disaster emergency training, resulting in insufficient self-rescue capabilities. In the event of a disaster, the public is unable to effectively take self-rescue and mutual rescue measures, increasing the risk of injury and death. There is a lack of interdisciplinary research in geology, climatology, and artificial intelligence, making it difficult to develop comprehensive solutions. The data silo problem between disciplines is serious, making it difficult to effectively share and integrate information, affecting the overall effectiveness of disaster prevention and control.

[0004] Landslide and debris flow disaster prevention and control mainly rely on single engineering control, lacking a whole-cycle defense line from "pre-disaster prevention and control-disaster response-post-disaster optimization". Before the disaster occurs, there is a lack of effective prevention measures; when the disaster occurs, the response measures are not timely and accurate; after the disaster occurs, there is a lack of summary and optimization of the rescue process. SUMMARY

[0005] In view of the above defects or improvement needs of the prior art, the present application provides an emergency rescue priority dynamic decision method and system, which significantly improves the accuracy and timeliness of disaster warning by constructing a geological disaster prevention and control system that integrates multi-source data, dynamic grid division, real-time monitoring and intelligent decision-making, and improves the warning accuracy to more than 90%. The system uses advanced data fusion and real-time analysis technology to solve the problem of insufficient coverage of remote mountain monitoring sites, and through intelligent resource allocation and emergency response mechanism, reduces the misallocation rate to within 10%, while shortening the response time and seizing the golden opportunity for rescue. The construction of the decision tree is optimized by using the CART algorithm, realizing intelligent sorting and dynamic adjustment of the rescue task, effectively solving the data island problem, promoting interdisciplinary collaboration and information sharing. In addition, the system promotes the accuracy of resource allocation and the success rate of rescue by iterating and adjusting key parameters through the promotion of the Bayesian optimization closed-loop system; a comprehensive "pre-disaster prevention and control-disaster response-post-disaster optimization" full-cycle defense line is constructed, providing a full-cycle, multi-dimensional solution from passive response to active prevention and control for disaster prevention, so as to better protect people's life and property safety and social stability under the double pressure of climate change and human activities.

[0006] In order to achieve the above-mentioned purpose, one aspect of the present application provides an emergency rescue priority dynamic decision method, comprising the following steps:

[0007] S1: Based on the distribution of surface attachments in the disaster area and DEM digital elevation data, the disaster area grid is divided, and a four-color hierarchical model is used to calculate, dynamically correct and visually color the grid importance weight;

[0008] S2: According to the movement partition characteristics of landslides or debris flows, combined with the fluid dynamics model to predict the material movement trajectory after the occurrence of geological disasters, dynamically adjust the potential influence range of each grid, update the weight value and perform secondary grid coloring;

[0009] S3: Obtain multi-source life signal data of trapped personnel through multi-source life detection equipment, calculate the comprehensive signal characteristic value of the multi-source life signal data by using a multi-source signal fusion model, determine the effective signal according to the signal strength, perform weight gain processing on the response grid, and perform weight zero and color reset on the non-response grid, to form the third grid coloring;

[0010] S4: Construct a multi-dimensional decision tree model, generate a rescue priority sequence by combining weight values, coloring levels and traffic accessibility parameters;

[0011] S5: Dynamically adjust the rescue priority and resource allocation strategy according to the real-time monitoring and feedback of data related to disaster response and emergency rescue and priority transition rules;

[0012] S6: Collecting rescue efficiency index data, establishing a Bayesian optimization model according to the rescue efficiency index data, applying parameter safety constraints and dynamic parameter adjustment rules to iterate the rescue efficiency index data, and realizing continuous optimization of the rescue strategy.

[0013] Further, step S1 comprises:

[0014] S11: Collecting ground surface attachment distribution information in the disaster area;

[0015] S12: Obtaining DEM digital elevation data of the disaster area;

[0016] S13: Collecting population distribution data of the disaster area;

[0017] S14: Determining the size of the grid according to the range and accuracy requirements of the disaster area; using GIS software or programming tools to divide the disaster area into regular grids; for each grid, extracting its terrain features and ground surface attachment information according to the DEM digital elevation data and ground surface attachment distribution information;

[0018] S15: Using a linear weighted model to calculate the comprehensive weight of each grid;

[0019] S16: Determining the classification standard according to the comprehensive weight value;

[0020] S17: Using GIS software or programming tools to color each grid according to the classification standard; displaying the color of each grid on the map to form an intuitive visual effect;

[0021] S18: Adjusting the grid weight according to the disaster development and real-time monitoring; increasing the weight when the grid is in the main flow line of the debris flow, reducing the facility weight when the slope is greater than 25°; setting the weight to the highest level when detecting vital signs, and reducing the weight when there is no signal for 2 hours.

[0022] Further, the comprehensive weight W in step S15 is calculated by formula (1):

[0023] W = α × W 人口 + β × W 设施 (1)

[0024] Wherein, α is the population weight correction coefficient, β is the facility weight correction coefficient, α + β = 1, α = 0.7, β = 0.3; W 人口 is the population density weight; W 设施 is the ground surface attachment weight;

[0025] Step S16 comprises: dividing the grid into four levels according to the comprehensive weight value, and assigning corresponding color identifiers;

[0026] The comprehensive weight interval is ≥0.8, the personnel density quantization value is >100 people / km 2 (or single building ≥50 people), the lifeline engineering is classified as I level, indicated by red, and the emergency response requirement is: immediate response, priority to put heavy rescue equipment;

[0027] The comprehensive weight interval is 0.6-0.79, the personnel density quantization value is 50-100 people / km 2 , the dangerous source facility is classified as II level, indicated by yellow, and the emergency response requirement is: 2 hours to deploy professional rescue team;

[0028] The comprehensive weight interval is 0.4-0.59, the personnel density quantization value is 10-50 people / km 2 , the general residential area is classified as I level, indicated by blue, and the emergency response requirement is: 4 hours to start basic rescue;

[0029] The comprehensive weight interval is <0.4, the personnel density quantization value is <10 people / km 2 , the ecological protection area is classified as I level, indicated by green, and the emergency response requirement is: monitoring and early warning, dynamic intervention according to disaster situation.

[0030] Further, in step S2, according to the movement zoning characteristics of the landslide or debris flow, the trajectory of the material after the occurrence of the geological disaster is predicted by combining the fluid dynamics model, the potential influence range of each grid is dynamically adjusted, the weight value is updated, and secondary coloring is performed; including:

[0031] S21: selecting the movement distance calculation formula of the landslide and debris flow two types of disasters;

[0032] S22: generating a disaster impact probability distribution map through spatial discretization, motion simulation and probability coverage;

[0033] S23: dynamically correcting the grid weight and color according to the disaster impact probability distribution map;

[0034] S24: correcting the parameters in the movement distance formula according to the real-time topographic feedback, triggering local grid topology reconstruction, and generating a new disaster impact probability distribution map;

[0035] S25: formulating technical implementation standards according to the new disaster impact probability distribution map;

[0036] Step S22 includes the following steps:

[0037] S221: spatial discretization, dividing the DEM topographic data of the study area into a regular grid system; extracting necessary topographic features for each grid;

[0038] S222: motion simulation: select the appropriate motion model according to the type of disaster; set the model parameters according to the existing disaster characteristics and terrain conditions; run the model to simulate the motion path and range of disaster materials on the terrain, and get the potential impact of each grid;

[0039] S223: probability coverage: use Monte Carlo method to estimate the probability of each grid being affected by disaster through a large number of random simulations; according to the simulation results, generate a disaster impact probability distribution map, in which the color or value of each grid represents the probability of its being affected by disaster;

[0040] Step S23 specifically includes: when the grid is in the main flow line of the debris flow, the probability is greater than or equal to 70%, the weight is increased by 0.2, the upper limit is 1.0, and the grid color is upgraded from red to deep red flashing warning; when the grid is located in the motion edge, 30%≤probability<70%, the weight is increased by 0.1, the color is changed from yellow to orange, and blue to light blue; when the grid is out of the predicted range, the probability is less than 30%, the weight is reduced to 50% of the original value, the lower limit is 0.2, and the green grid is restored to the background color;

[0041] The technical implementation standard in step S25 includes:

[0042] When the landslide motion distance L slide and the debris flow motion distance L debris satisfy:

[0043] L slide ≥300m or L debris ≥500m, it is a red alert area;

[0044] When the landslide motion distance and the debris flow motion distance satisfy:

[0045] 150m≤L slide <300m or 200m≤L debris <500m, it is a yellow warning area.

[0046] Further, in step S3, the multi-source life signal data of the trapped personnel is obtained by the multi-source life detection equipment, the comprehensive signal characteristic value of the multi-source life signal data is calculated by using the multi-source signal fusion model, the effective signal is judged according to the signal intensity, the response grid is processed by weight gain, the non-response grid is executed by weight zero and color reset, and the third grid coloring is formed; including:

[0047] S31: collect the vital signs signal, thermal imaging signal and motion signal of the trapped personnel by the life detection instrument, thermal imaging equipment and radar detector;

[0048] S32: the vital signs signal, thermal imaging signal and motion signal of the trapped personnel are fused by using the multi-source signal fusion model to obtain the comprehensive signal characteristic value;

[0049] S33: According to the comprehensive signal characteristic value, an effective signal determination standard is formulated; the comprehensive signal characteristic value S≥0.7 is determined as a strong response signal, which is marked with red flashing; the comprehensive signal characteristic value 0.4≤S<0.7 is determined as a weak response signal, which is marked with yellow pulse; and the comprehensive signal characteristic value S<0.4 is determined as an invalid signal, which does not trigger weight adjustment;

[0050] S34: The grid weight and color are corrected according to the effective signal determination standard; specifically, the weight gain processing is performed on the responsive grid, the weight is reset to zero, the color is reset, and the data is traced on the non-responsive grid.

[0051] Further, the multi-source signal fusion model in step S32 is represented by formula (4):

[0052] S = λ1·S 生命 + λ2·S 热力 + λ3·S 运动 (4)

[0053] Wherein, S is the comprehensive signal characteristic value; S 生命 is the vital sign signal characteristic value; S 热力 is the thermal imaging signal characteristic value; S 运动 is the motion signal characteristic value; λ1=0.5, λ2=0.3, λ3=0.2, and the weight is allocated according to the device accuracy level;

[0054] The weight gain processing in step S34 includes:

[0055] The gain weight W new is calculated, which is represented by formula (8):

[0056] W new = min(1.0, W 原 + ΔW·log2(1+S)) (8)

[0057] Wherein, ΔW is the gain coefficient, 0.3 for strong response and 0.15 for weak response; W 原 is the original weight value;

[0058] According to the time decay mechanism, the gain value is attenuated by 50% every 15 minutes without updating the signal; and the gain effect is cleared after 3 times of continuous non-updating;

[0059] According to the color synchronization rule, the grid area with the gain weight W new ≥0.8 is displayed with a deep red background and a gold pulse frame; the system triggers an audible and light alarm, and sends an emergency notification to the relevant personnel through the alarm mode of short message push;

[0060] The gain weight 0.6≤Wnew Grid area with a value less than 0.8 is displayed with an orange gradient animation; the user is reminded to pay attention to this area through a pop-up window prompt in the system;

[0061] The gain after weight W new Grid area with a value less than 0.6 maintains the original color, indicating that the priority of this area is low; the system only records the information of this area in the system log, and does not perform additional reminders or alarms;

[0062] The trigger condition for zeroing the weight of the non-responsive grid in step S34 is: first-level determination, no valid signal for 2 consecutive detection periods; second-level determination, confirmed by manual review that there is no sign of life;

[0063] The color reset rule is: red / yellow grid, weight zero, restore pre-disaster color, semi-transparent superimposed "X" mark; blue / green grid, weight zero, retain original color but transparency increased to 70%, displayed as a dashed boundary;

[0064] The data tracing mechanism is: the weight zero grid automatically generates a signal disappearance report, and the historical data is retained for 72 hours for manual review.

[0065] Further, in step S4, a multi-dimensional decision tree model is constructed to generate a rescue priority sequence by integrating weight value, coloring level, and traffic accessibility parameters; including:

[0066] S41: define the feature dimension of the decision tree; including three categories and six dimensions of decision features; three categories include disaster emergency degree, resource accessibility, and rescue efficiency; six dimensions include grid color level and comprehensive weight value for evaluating disaster emergency degree, road traffic index and air drop feasibility for evaluating resource accessibility, and historical success rate and equipment matching degree for evaluating rescue efficiency;

[0067] S50: for the feature dimension of the decision tree, use the CART algorithm to split the decision tree nodes, construct a binary tree, select the feature and split point that can reduce the Gini impurity to the greatest extent, and divide the rescue task into different feature priorities;

[0068] S43: dynamically generate an initial rescue sequence according to the feature priority ranking and the calculation results of the decision tree model.

[0069] Further, the rescue sequence in step S43 includes:

[0070] Priority 1, trigger condition combination: red grid + weight ≥ 0.8 + air drop available; resource deployment strategy: immediately dispatch a helicopter + unmanned aerial vehicle cluster delivery

[0071] Priority 2, trigger condition combination: red grid + weight ≥ 0.8 + no empty drop; resource deployment strategy: heavy machinery open road + satellite communication support assault team;

[0072] Priority 3: trigger condition combination: red grid + 0.6 ≤ weight < 0.8 + sufficient equipment; resource deployment strategy: standard rescue team double formation cooperative advance;

[0073] Priority 4: trigger condition combination: red grid + 0.6 ≤ weight < 0.8 + insufficient equipment; resource deployment strategy: two-stage rescue after emergency allocation of materials;

[0074] Priority 5: trigger condition combination: non-red grid + road unobstructed + historical high success rate; resource deployment strategy: light rescue team rapid response;

[0075] Priority 6: trigger condition combination: non-red grid + road unobstructed + historical low success rate; resource deployment strategy: expert command group with team;

[0076] Priority 7: trigger condition combination: yellow grid + road not passable; resource deployment strategy: advance reconnaissance team to establish temporary supply point;

[0077] Priority 8: trigger condition combination: blue / green grid; resource deployment strategy: not in action for the time being;

[0078] Step S41 also includes:

[0079] For disaster emergency degree:

[0080] According to the dynamic grid processing engine, the grid area is divided into four levels according to the severity of the disaster, and different values are assigned: red-I level: value is 4, indicating the most serious disaster, which needs the most priority rescue; yellow-II level: value is 3, indicating the disaster is relatively serious; blue-III level: value is 2, indicating the disaster is general; green-IV level: value is 1, indicating the disaster is relatively light;

[0081] Based on the four-dimensional weight calculation unit, the comprehensive weight value of each grid is calculated, and these weight values are normalized to the interval of 0-1;

[0082] For resource accessibility:

[0083] Based on the traffic real-time big data platform, the road traffic index is obtained;

[0084] The road traffic index is represented by formula (9):

[0085] R road =e -0.1t (9),

[0086] wherein t represents the time required to reach the grid area;

[0087] The feasibility of air drop is determined by the UAV reconnaissance system. The feasibility of air drop is represented by a 0-1 Boolean value, indicating whether the grid area can be air dropped for rescue. If the terrain flatness is greater than 5° and there is no high-voltage line obstruction, the air drop is feasible, and the Boolean value is 1. Otherwise, the air drop is not feasible, and the Boolean value is 0.

[0088] For rescue efficiency:

[0089] Based on the adaptive learning database, the historical rescue success rate of similar scenarios is obtained, and multiplied by 0.01 as the value of this indicator.

[0090] Based on the material management information system, the equipment matching degree is calculated according to the matching degree of rescue demand and existing rescue equipment.

[0091] Further, step S5 includes:

[0092] S51: Data collection, collecting data from multi-source life detection equipment; collecting road traffic index and terrain change data through traffic monitoring system, satellite navigation, and UAV reconnaissance system; collecting material consumption and equipment matching degree data from material management information system;

[0093] S52: Data analysis, calculating the comprehensive signal characteristic value according to the detection equipment data, and evaluating the intensity of life signs; updating the road traffic index every 5 minutes according to the traffic data; when the material consumption reaches 50%, recalculating the equipment matching degree;

[0094] S53: Priority evaluation according to priority transition rules, obtaining new priority data;

[0095] S54: Updating the feature dimension of the decision tree according to the new priority data, using the CART algorithm, updating the node splitting rule of the decision tree based on the Gini impurity minimization principle, and obtaining the updated priority sequence;

[0096] S55: Optimizing the scheduling and allocation of rescue resources according to the updated priority sequence, dynamically adjusting the resource allocation plan according to real-time feedback and priority transition rules;

[0097] The priority transition rules in step S53 are: if the grid color level is upgraded, the rescue priority of the related area is increased by 3 levels; when the road is suddenly interrupted and the road traffic index decreases by more than 0.3, the priority of the affected area is reduced by 2 levels; when new life signs are detected, the priority of the related area is increased to the current highest level + 1.

[0098] Further, step S6 includes:

[0099] S61: Collect and record grid rescue time efficiency, resource consumption and success metric indicators;

[0100] S62: Define optimization objective function according to grid rescue time efficiency, resource consumption and success metric indicators; optimization objective is to minimize rescue time and resource consumption, while maximizing rescue success rate and key facility preservation rate;

[0101] S63: Determine decision variable space;

[0102] S64: Build Bayesian optimization model;

[0103] S65: Clean and process collected rescue efficiency indicator data, construct features conducive to model training; based on Bayesian optimization model, calculate posterior distribution of rescue strategy key parameters; use acquisition function to guide optimization search of rescue strategy key parameters; based on digital twin platform, perform Monte Carlo simulation to evaluate the effect of the optimized strategy;

[0104] S66: Implement the optimized rescue strategy in actual rescue operations; collect rescue feedback data after implementation to evaluate the effect of the optimized strategy in actual rescue;

[0105] S67: Feedback new rescue data and evaluation results to the Bayesian optimization model, continuously update and iterate the model; adjust model parameters and optimization strategy according to new rescue data and evaluation results;

[0106] The collection parameters of the time efficiency indicators in step S61 include response delay time and task duration; response delay time is recorded when the event is triggered; task duration is updated when the task is completed;

[0107] The collection parameters of the resource consumption indicators include equipment usage intensity and manpower input equivalent; equipment usage intensity is calculated according to the usage of rescue equipment per hour; manpower input equivalent is summarized at the end of the task;

[0108] The collection parameters of the success metric indicators include life rescue success rate P life and key facility preservation rate P facility ; life rescue success rate P life is evaluated at the completion of the task; key facility preservation rate P facility is evaluated within 48 hours after the disaster;

[0109] The optimization objective function in step S62 is represented by formula (10):

[0110] θ=max(ω1P life +ω2P facility -ω3T total -ω4Q) (10)

[0111] wherein the weights are set according to the Multi-Objective Optimization Weight Distribution Specification:

[0112] ω1=0.4, ω2=0.3, ω3=0.2, ω4=0.1; PP life is the survival rate of the living body; P facility is the survival rate of the living body; P total is the total rescue time; Q is the resource consumption;

[0113] In step S63, the decision variable space includes adjustable parameters and their definition domains and physical meanings; specifically as follows:

[0114] The definition domain of the weight coefficient a is [0.5, 0.9], indicating the proportion of the personnel density in the comprehensive weight;

[0115] The definition domain of the road passing threshold R crt is [0.4, 0.8], which is the minimum passing index triggering road repair operations;

[0116] The definition domain of the air-drop safety angle θ safe is [5°, 25°], indicating the maximum terrain inclination angle allowed for unmanned aerial vehicle air-drop;

[0117] The definition domain of the gain decay period τ is [10, 60] minutes, indicating the half-life period of the detection signal gain effect.

[0118] Further, the dynamic parameter adjustment rule in step S6 is: level I response, online real-time optimization, iteration every 15 minutes, using cloud computing cluster for calculation; level II response, offline batch optimization, iteration every 2 hours, using edge server for calculation; level III response, historical pattern matching, updated every 6 hours, using local workstation for calculation;

[0119] The parameter safety constraint is that the weight coefficient a cannot be lower than the lower limit specified in the geological disaster prevention regulation, i.e. a≥0.55; the air-drop safety angle adjustment range cannot exceed 5° at a time to prevent drastic fluctuations.

[0120] The second aspect of the present application provides an emergency rescue priority dynamic decision system for implementing the emergency rescue priority dynamic decision method, comprising:

[0121] A multi-source data fusion module for integrating data from satellite remote sensing data, unmanned aerial vehicle surveying and mapping data, and Internet of Things sensors;

[0122] A dynamic grid processing engine for identifying real-time changes in the terrain and topologically reconstructing the grid according to these changes;

[0123] A four-dimensional weight calculation unit is used to calculate the weight of each grid by combining the four parameters of terrain slope, population density, infrastructure density and ecological sensitivity, to evaluate the importance and priority of different regions and guide resource allocation and rescue operations;

[0124] A visual decision interface is used to provide three-dimensional situation deduction and dynamic planning of rescue paths, enabling decision makers to intuitively understand and analyze disaster situations and optimize rescue paths and strategies;

[0125] An adaptive learning database is used to store historical disaster patterns and rescue efficiency comparison data, to improve future rescue decisions by learning from historical data and adapting to new rescue scenarios.

[0126] Overall, compared with the prior art, the above technical solutions of the present application can achieve the following beneficial effects:

[0127] (1) The emergency rescue priority dynamic decision method and system of the present application, for landslides or debris flow disasters, on the basis of initial surface attachments, forms a grid initial division according to the obtained DEM data, assigns weights to the grid according to the importance according to the weight calculation model of the red, yellow, blue and green four-color grid, and colors the grid; according to the movement partition characteristics of landslides or debris flow, the potential movement distance of the above-mentioned surface attachments is corrected, and the above-mentioned weight assignment and grid coloring method is used to form a secondary correction; on the basis of the above-mentioned grid partition, a priority classification of detection is formed, for the detection with signal response, the weight is re-assigned and the grid is colored, for the weight without signal reflection, the weight is zeroed and the color is returned to the background color, forming a third grid coloring; on this basis, according to the importance principle, guidance is provided for the deployment of ground search and rescue resources, according to the resource allocation decision tree method, an emergency rescue priority dynamic decision method is formed; finally, the rescue data of each grid is recorded to provide a big data analysis basis for the next rescue optimization; the present application effectively solves the key pain points in traditional disaster rescue through technical innovation, improves the efficiency and effectiveness of rescue, and provides strong technical support for disaster emergency management.

[0128] (2) The emergency rescue priority dynamic decision-making method and system of the present application, by fusing fluid dynamics model and real-time remote sensing data, constructs a three-dimensional geological disaster deduction system, realizes the innovation of dynamic risk assessment and early warning technology, and improves the disaster warning accuracy to more than 90%; based on satellite, unmanned aerial vehicle and IoT sensor and other multi-source data, dynamic grid division and weight correction are realized, rapid response is realized, and emergency response time is shortened, which wins valuable time for rescue operation; using Bayesian optimization closed loop system, through historical rescue data iterative adjustment of key parameters, the accuracy of resource allocation is improved, resource waste is reduced, and the effectiveness and safety of rescue operation are ensured; the intelligent emergency decision-making model of man-machine cooperation is constructed, the machine learning prediction result is combined with the expert experience rule, the high-risk area of disaster is quickly identified, the resource allocation and rescue operation are optimized; through multi-stage dynamic correction mechanism and intelligent decision-making model, the priority misjudgment rate and resource misallocation rate can be significantly reduced; the success rate of life rescue, resource reuse rate and secondary disaster warning accuracy are effectively improved; through the cross research of geology, climatology and artificial intelligence, the data island problem is solved, the information sharing and resource integration are promoted, the cross-disciplinary cooperation and policy support are realized, the scientificity and accuracy of disaster prediction and response are improved; the disaster prevention and control is changed from single engineering management to comprehensive system of "scientific and technological early warning-intelligent decision-making-ecological restoration", the global monitoring network coverage and community resilience construction are carried out from the technical level, the passive response to active prevention and control is realized; facing the double pressure of climate change and human activity, through interdisciplinary innovation, policy strengthening and public participation, the whole cycle defense line of "pre-disaster prevention and control-disaster response-post-disaster optimization" is constructed; through the intelligent early warning and decision-making system, the efficiency of disaster management is improved, the resource waste is reduced, and the speed and accuracy of disaster response are improved.

[0129] (3) The emergency rescue priority dynamic decision-making method and system of the present invention integrates multiple data sources, such as DEM data, UAV mapping, laser rangefinder, etc., to achieve high-precision surface deformation monitoring and grid management; by connecting with the traffic management platform and material management system, it can achieve accurate delivery of rescue resources, thereby improving rescue efficiency and safety; through the three stages of initial weight calculation, motion model correction and real-time detection feedback, it can realize dynamic weight adjustment of disaster-affected areas, significantly reducing the priority misjudgment rate (reduced by 50%); combining the static parameters of the terrain and the dynamic propagation characteristics of the disaster, Using drones and other equipment to update terrain data every 30 minutes, local grids can be reconstructed in real time, enabling a more accurate assessment of the impact of disasters. A decision-making model tree is constructed based on six dimensions, including disaster urgency, resource accessibility, and rescue effectiveness, to achieve minute-level policy updates and control the resource mismatch rate within 10%. A Bayesian optimization closed-loop system is used to continuously learn historical rescue data and adaptively tune parameters to improve rescue efficiency and safety. Compared with traditional methods, the present invention has effectively improved response delay, life rescue success rate, resource reuse rate, and secondary disaster warning accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0130] Figure 1 A flowchart of a method for dynamic decision-making of emergency rescue priority according to an embodiment of the present invention is shown;

[0131] Figure 2 This is a schematic diagram of the structure of a dynamic decision-making system for emergency rescue priority according to an embodiment of the present invention;

[0132] Figure 3 The figure is a schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0133] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to illustrate the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0134] like Figure 1 As shown, one aspect of the present invention provides a method for dynamic decision-making of emergency rescue priority, comprising the following steps:

[0135] S1: Based on the distribution of surface attachments in the disaster area and DEM digital elevation data, the disaster area is divided into grids, and the importance weight of the grid is calculated, dynamically corrected, and visualized using a four-color grading model (yellow / blue / green / red);

[0136] S2: According to the movement zoning characteristics of landslides or debris flows, the trajectory of the material movement after the occurrence of geological disasters is predicted by combining fluid dynamics models, the potential impact range of each grid is dynamically adjusted, the weight values are updated, and secondary grid coloring is performed;

[0137] S3: Obtain multi-source life signal data of trapped personnel through multi-source life detection equipment, calculate the comprehensive signal characteristic value of the multi-source life signal data by using a multi-source signal fusion model, determine the effective signal according to the signal strength, perform weight gain processing on the response grid, perform weight zero and color reset on the non-response grid, and form the third grid coloring;

[0138] S4: Construct a multi-dimensional decision tree model to generate a rescue priority sequence by integrating weight values, coloring levels, and traffic accessibility parameters;

[0139] S5: Dynamically adjust the rescue priority and resource allocation strategy according to real-time monitoring and feedback data related to disaster response and emergency rescue, and priority transition rules;

[0140] S6: Collect rescue efficiency index data, establish a Bayesian optimization model based on the rescue efficiency index data, iterate the rescue efficiency index data, and realize continuous optimization of the rescue strategy.

[0141] Further, in step S1, the disaster area grid is divided based on the distribution of surface attachments in the disaster area and DEM digital elevation data, and a four-color classification model is used for importance weight calculation, dynamic correction, and visual coloring of the grid; including:

[0142] S11: Collect the distribution information of surface attachments in the disaster area, including the distribution data of buildings, roads, schools, hospitals, substations, communication base stations, farmland, forest land, etc. These data can be obtained through geographic information system (GIS) database, satellite remote sensing image, unmanned aerial vehicle surveying and mapping, etc.

[0143] S12: Obtain DEM (Digital Elevation Model) data of the disaster area for analyzing terrain features such as slope and elevation. DEM digital elevation data can be obtained through satellite surveying and mapping, aerial photogrammetry, laser radar (LiDAR), etc.

[0144] S13: Collect population distribution data in the disaster area, including population density, resident location information, etc. These data can be obtained through population census data, GIS database, etc.

[0145] S14: Determine the size of the grid according to the range and accuracy requirements of the disaster area; divide the disaster area into regular grids using GIS software or programming tools; for each grid, extract its terrain features (such as elevation, slope) and surface attachments information (such as building density, population density, facility type, etc.) according to DEM digital elevation data and surface attachments distribution information;

[0146] S15: Calculate the comprehensive weight of each grid using a linear weighting model; the comprehensive weight of each grid includes personnel density weight and surface attachments weight;

[0147] S16: Determine the classification standard according to the comprehensive weight value;

[0148] S17: Use GIS software or programming tools to color each grid according to the classification standard; display the color of each grid on the map to form an intuitive visual effect;

[0149] S18: Adjust the grid weight according to the disaster development and real-time monitoring; increase the weight when the grid is in the main flow line of the debris flow, and reduce the facility weight when the slope is greater than 25°; set the weight to the highest level when detecting vital signs, and reduce the weight when there is no signal for 2 hours;

[0150] Further, step S15 includes:

[0151] S151: Determine the weight calculation model, and calculate the comprehensive weight of each grid using a linear weighting model; the comprehensive weight value W is calculated by formula (1):

[0152] W = α × W 人口 + β × W 设施 (1)

[0153] Wherein, α is the population weight correction coefficient, β is the facility weight correction coefficient, α + β = 1, according to the "Technical Code for Geological Disaster Emergency Rescue", α = 0.7, β = 0.3; W 人口 is the personnel density weight; W 设施 is the surface attachments weight;

[0154] S152: Calculate the personnel density weight W 人口 of each grid according to the personnel density classification; the quantization value > 100 people / km 2 , the weight coefficient is 1.0; the quantization value is 50-100 people / km 2 , the weight coefficient is 0.8; the quantization value is 10-50 people / km 2 , the weight coefficient is 0.6; the quantization value < 10 people / km 2 , the weight coefficient is 0.3; as shown in Table 1;

[0155] Table 1 Personnel density weight table

[0156] Personnel density classification quantified values (people / km 2 )]]> Weighting factor Very high density (Class I) >100 1.0 High density (Class II) 50～100 0.8 Medium density (Class III) 10～50 0.6 Low density (Class IV) <10 0.3

[0157] S153: According to the facility type, calculate the ground surface attachment weight W of each grid 设施 ; Lifeline engineering (such as hospitals, schools, government agencies, transportation hubs): the weight coefficient is 1.0; Hazardous facilities (such as transformer substations, communication base stations, chemical plants): the weight coefficient is 0.8; Ordinary civil buildings (such as residential communities, commercial blocks): the weight coefficient is 0.6; Non-sensitive ecological areas (such as farmland, forest land, unpopulated facility areas): the weight coefficient is 0.2; See Table 2;

[0158] Table 2 Ground surface attachment weight table

[0159]

[0160]

[0161] S154: According to the personnel density weight and the ground surface attachment weight, calculate the comprehensive weight of each grid by formula (1);

[0162] Further, step S16 includes: according to the comprehensive weight value, dividing the grid into four levels and giving corresponding color identification; the comprehensive weight interval ≥0.8, the personnel density quantization value is >100 people / km 2 (or single building ≥50 people), lifeline engineering is classified as level I, represented by red, representing high-risk life gathering area, and the emergency response requirement is: immediate response, priority to put heavy rescue equipment; the comprehensive weight interval is 0.6-0.79, the personnel density quantization value is 50-100 people / km 2 , dangerous source facilities are classified as level II, represented by yellow, representing secondary disaster risk area, and the emergency response requirement is: 2 hours to deploy professional rescue team; the comprehensive weight interval is 0.4-0.59, the personnel density quantization value is 10-50 people / km 2 , general residential area is classified as level I, represented by blue, representing the key node of infrastructure, and the emergency response requirement is: 4 hours to start basic rescue; the comprehensive weight interval is <0.4, the personnel density quantization value is <10 people / km 2 , ecological protection area is classified as level I, represented by green, representing ecological protection area, and the emergency response requirement is: monitoring and early warning, dynamic intervention according to disaster situation; see Table 3;

[0163] Table 3 Four-level classification of emergency rescue and corresponding color identification division standard

[0164]

[0165]

[0166] Further, the dynamic correction rule of the grid weight in step S18 meets the Technical Guidelines for Real-time Monitoring of Geological Disasters, the weight coefficient threshold matches the Code for Emergency Investigation of Geological Disasters GB / T 36112, the color classification is compatible with the Measures for the Operation and Management of the National Emergency Warning Information Release System, and the dynamic correction mechanism meets the requirements of the Guidelines for the Construction of Intelligent Geological Disaster Monitoring Systems, as follows:

[0167] ①Disaster superposition correction: when the grid is in the main flow line of the debris flow, the weight is increased by 0.2 (not more than 1.0), and the facility weight is reduced by 0.1 in the area with a slope of > 25°.

[0168] ②Real-time feedback correction: when vital signs are detected, the weight is forcibly set to level I (red), and if there is no signal for 2 hours, the weight is reduced to 80% of the original level.

[0169] According to the measured data, the grid weight dynamic correction method of the present application realizes the objectivity and standardization of disaster classification through the deep coupling of quantitative indicators and legal norms, and improves the rescue efficiency by 41% compared with the traditional experience judgment method.

[0170] Further, in step S2, according to the movement partition characteristics of landslides or debris flows, the motion trajectory of geological disasters after the occurrence is predicted by combining the fluid dynamics model, the potential influence range of each grid is dynamically adjusted, the weight value is updated and secondary coloring is performed; including:

[0171] S21: selecting the movement distance calculation formula of landslides and debris flows;

[0172] S22: generating a disaster impact probability distribution map through spatial discretization, motion simulation and probability coverage;

[0173] S23: dynamically correcting the grid weight and color according to the disaster impact probability distribution map;

[0174] S24: correcting the parameters in the motion distance formula according to the real-time terrain feedback, triggering local grid topology reconstruction, and generating a new disaster impact probability distribution map;

[0175] S25: formulating technical implementation standards according to the new disaster impact probability distribution map;

[0176] Further, in step S21, the movement distance L of the landslide slide The formula (2) is calculated as follows:

[0177] L slide = K·H·V·tanφ (2)

[0178] Wherein, K is the friction correction coefficient of rock-soil body (0.8-1.2), H is the vertical height of landslide (m), V is the volume of landslide (10 3 m 3 ), φ is the friction angle of sliding surface (°);

[0179] The movement distance L of debris flow debris is calculated by formula (3):

[0180]

[0181] Wherein, wherein, ρ is the fluid density (kg / m 3 ), h is the flow depth (m), θ is the channel slope (°), η is the viscosity coefficient (Pa·s), t flow is the flow duration (s); v0 is the volume of debris flow.

[0182] Further, the step S22 comprises the following steps:

[0183] S221: Spatial discretization, divide the DEM terrain data of the study area into a regular grid system, each grid has the same size, such as 10m×10m; extract the necessary terrain features for each grid, including elevation, slope, aspect, lithology, etc.

[0184] S222: Motion simulation: select the appropriate motion model according to the disaster type (such as landslide, debris flow, etc.), such as cellular automata model, fluid dynamics model, etc.; according to the existing disaster characteristics and terrain conditions, set the model parameters, such as friction coefficient, density, viscosity, etc.; run the model, simulate the motion path and range of disaster materials on the terrain, and get the potential impact of each grid;

[0185] S223: Probability coverage: use the Monte Carlo method (1000 iterations) to estimate the probability of each grid being affected by the disaster through a large number of random simulations; according to the simulation results, generate a disaster impact probability distribution map, the color or value of each grid in the map represents the probability of its being affected by the disaster;

[0186] Further, the step S23 specifically comprises: when the grid is in the main flow line of debris flow (probability≥70%), the weight is increased by 0.2 (upper limit 1.0), and the grid color is upgraded from red to deep red flashing warning; when the grid is located in the motion edge (30%≤probability<70%), the weight is increased by 0.1, and the color is changed from yellow to orange, and blue to light blue; when the grid is out of the prediction range (probability<30%), the weight is reduced to 50% of the original value (lower limit 0.2), and the green grid returns to the background color; see Table 4 for details.

[0187] Table 4 Grid weight dynamic correction method and standard division table

[0188]

[0189] Further, step S24 specifically includes the following steps:

[0190] S241: Topography monitoring, using laser range finder, unmanned aerial vehicle and other equipment to detect the working area DEM elevation data in real time and further monitor the topographic changes;

[0191] S250: Parameter correction, when detecting topographic mutation (elevation change > 1m), correcting the parameters in the motion model according to the new topographic data, such as elevation, slope, etc.

[0192] S243: Trigger local grid topology reconstruction, update the topology structure of the affected grid (such as the grid within the influence radius of 500m), including re-dividing the grid, adjusting the grid features, etc.

[0193] S244: Re-run the motion model using the updated grid data to get the new disaster impact probability distribution map;

[0194] Specifically, the working area DEM elevation data is detected in real time by laser range finder and unmanned aerial vehicle; the conditions for triggering local grid topology reconstruction are set, such as the topographic change exceeding a certain threshold (e.g. elevation change > 1m); when detecting topographic mutation (elevation change > 1m), the relevant parameters in the motion distance calculation formula of the two types of disasters, landslide and debris flow, are recalculated, and under the triggering condition, the topology structure of the affected grid (such as the grid within the influence radius of 500m) is updated, including re-dividing the grid, adjusting the grid features, etc., to ensure that the assessment of disaster impact range is more accurate and timely; the motion model is re-run using the updated grid data to get the new disaster impact probability distribution map.

[0195] Further, the technical implementation standards in step S25 include:

[0196] When the landslide motion distance and the debris flow motion distance meet:

[0197] L slide ≥300m or L debris ≥500m, it is a red alert zone;

[0198] When the landslide motion distance and the debris flow motion distance meet:

[0199] 150m≤L slide <300m or 200m≤L debris <500m, it is a yellow warning zone;

[0200] The model verification requirement of the movement distance calculation formula of landslide and debris flow disasters is that the prediction error of the movement distance is less than or equal to 15%; the time consumption of grid reconstruction is less than or equal to 5 minutes (meeting the timeliness of emergency response). The parameter values of the formula meet the “Code for Geotechnical Engineering Investigation” (GB 50021); the dynamic adjustment mechanism meets the secondary response timeliness requirement of the “Natural Disaster Relief Emergency Plan”; the probability coverage model is verified by the National Geological Disaster Data Center (with a case matching rate of more than 90%), and the identification accuracy of the rescue key area is improved by 50% (compared with the traditional experience method) by coupling the physical model with real-time data, thereby providing a scientific basis for accurate allocation of emergency resources.

[0201] Further, in step S3, the multi-source life signal data of the trapped personnel is acquired by the multi-source life detection device, a multi-source signal fusion model is used to calculate the comprehensive signal characteristic value of the multi-source life signal data, the effective signal is determined according to the signal intensity, the weight gain processing is performed on the response grid, the weight is reset to zero and the color is reset for the non-response grid, and the third grid coloring is formed; comprising:

[0202] S31: Collecting the vital sign signals, thermal imaging signals and motion signals of the trapped personnel by the life detector, thermal imaging device and radar detector;

[0203] S32: Fusing the vital sign signals, thermal imaging signals and motion signals of the trapped personnel by using a multi-source signal fusion model to obtain a comprehensive signal characteristic value;

[0204] S33: Formulating an effective signal determination standard according to the comprehensive signal characteristic value; setting the comprehensive signal characteristic value S≥0.7 as a strong response signal, which is marked with red flashing; setting the comprehensive signal characteristic value 0.4≤S<0.7 as a weak response signal, which is marked with yellow pulse; and setting the comprehensive signal characteristic value S<0.4 as an invalid signal, which does not trigger weight adjustment;

[0205] S34: Correcting the grid weight and color according to the effective signal determination standard; specifically including: performing weight gain processing on the response grid, performing weight reset, color reset and data tracing processing on the non-response grid;

[0206] Further, the multi-source signal fusion model in step S32 is represented by formula (4):

[0207] S = λ1·S 生命 + λ2·S 热力 + λ3·S 运动 (4)

[0208] Wherein, D is the comprehensive signal characteristic value; S 生命 is the vital sign signal characteristic value; S 热力 is the thermal imaging signal characteristic value; and S 运动The motion signal characteristic value; λ1=0.5, λ2=0.3, λ3=0.2, the weight is allocated according to the equipment accuracy level;

[0209] The vital sign signal characteristic value S 生命 It is expressed by formula (5):

[0210]

[0211] Wherein, t is the detection delay minutes, the quantization range is 0~1.2;

[0212] The thermal imaging signal characteristic value S 热力 It is expressed by formula (6):

[0213]

[0214] Wherein, θ is the angle between the device and the heat source; the quantization range is 0~0.8;

[0215] The motion signal characteristic value S 运动 It is expressed by formula (7):

[0216]

[0217] Wherein, the quantization range is 0~1.0;

[0218] Further, the weight gain processing in step S34 includes:

[0219] The weight W after gain new It is calculated and expressed by formula (8):

[0220] W new = min(1.0, W 原 +ΔW·log2(1+S)) (8)

[0221] Wherein, ΔW is the gain coefficient, the strong response takes 0.3, and the weak response takes 0.15; W 原 Is the original weight value;

[0222] According to the time decay mechanism, the gain value is attenuated by 50% every 15 minutes without updating the signal, and the gain effect is cleared after 3 times of continuous updating;

[0223] According to the color synchronization rule, the grid area with the weight W new ≥0.8 after gain is highlighted with a dark red background and a gold color pulse frame (the frequency of the pulse is 2Hz, that is, flashing twice per second); the system triggers an audible and light alarm, and sends an emergency notification to the relevant personnel (on-site personnel or monitoring personnel) through the alarm mode of short message push;

[0224] The weight 0.6≤W newThe grid area with a value less than 0.8 is displayed with an orange gradient animation (similar to the breathing light effect); a pop-up window will be displayed to remind the user to pay attention to the area.

[0225] The weight W after the gain new Grid areas with a value less than 0.6 retain their original color, indicating that the area has a lower priority. The system only records information about this area in the system log and does not issue additional reminders or alarms.

[0226] In this way, the system can calculate the new weight value W new Dynamically adjust the display and reminder methods of grid areas to help users quickly identify and respond to areas of different priorities; see Table 5 for details;

[0227] Table 5 Color synchronization rules

[0228]

[0229]

[0230] Furthermore, the triggering conditions for resetting the weight of the unresponsive grid to zero in step S34 are: first-level judgment, no valid signal (S<0.4) for two consecutive detection cycles (30 minutes); second-level judgment, no signs of life confirmed by manual review;

[0231] The color reset rules are as follows: Red / yellow grid: weight is reset to zero, the pre-disaster background color (gray) is restored, and a semi-transparent "X" mark is superimposed; Blue / green grid: weight is reset to zero, the original color is retained but the transparency is increased to 70%, and it is displayed as a dotted border; see Table 6;

[0232] Table 6 Color reset rules

[0233]

[0234] The data traceability mechanism is as follows: the weighted zero grid automatically generates a "Signal Loss Report", and historical data is retained for 72 hours for manual review;

[0235] The signal characteristic value algorithm in step S3 passes the "Technical Specification for Certification of Fire Emergency Rescue Equipment" (XF / T 3015); the weight adjustment threshold complies with the "General Technical Requirements for Emergency Command Systems" (GB / T 37228); and the color coding is compatible with the "Colors of Public Safety Emergency Signs" (GB / T 2893.4). Measured data shows that the present invention can achieve minute-by-minute conversion of detection signals into rescue decisions (average response time in minutes), with a false alarm rate below 10%.

[0236] Further, the decision tree in step S4 is a tree structure used for classification or regression of data. Each internal node represents a feature (attribute), each branch represents a possible value of the feature, and each leaf node represents a class or output value. The decision tree maps input data to output results through a series of decision rules; in step S4, a multi-dimensional decision tree model is constructed to generate a rescue priority sequence based on the weight value, coloring level, and traffic accessibility parameter; including:

[0237] S41: defining the feature dimension of the decision tree;

[0238] Specifically, it includes establishing three types of six-dimensional decision features (see Table 7); the three types include disaster emergency degree, resource accessibility, and rescue efficiency; the six dimensions include grid color level and comprehensive weight value for evaluating disaster emergency degree, road traffic index and air drop feasibility for evaluating resource accessibility, and historical success rate and equipment matching degree for evaluating rescue efficiency;

[0239] For disaster emergency degree:

[0240] According to the dynamic grid processing engine, the grid area is divided into four levels according to the severity of the disaster, and different values are assigned: red (level I): value 4, indicating the most serious disaster and the most urgent rescue; yellow (level II): value 3, indicating a relatively serious disaster; blue (level III): value 2, indicating a general disaster; green (level IV): value 1, indicating a relatively light disaster;

[0241] Based on the four-dimensional weight calculation unit, the comprehensive weight value of each grid is calculated and normalized to the interval of 0-1. The comprehensive weight value takes into account population density, importance of surface attachments, and other factors, reflecting the overall importance of the grid area;

[0242] For resource accessibility:

[0243] Based on the traffic real-time big data platform, the road traffic index is obtained; the closer the road traffic index value is to 1, the smoother the road is and the better the accessibility is;

[0244] The road traffic index is represented by formula (9):

[0245] R road =e -0.1t (9),

[0246] Where t represents the time (in minutes) required to reach the grid area;

[0247] The feasibility of air drop is determined by the UAV reconnaissance system. The feasibility of air drop is represented by a 0-1 Boolean value, indicating whether the grid area can be air dropped for rescue. If the terrain flatness is greater than 5° and there is no high-voltage line obstruction, the air drop is feasible, and the Boolean value is 1. Otherwise, the air drop is not feasible, and the Boolean value is 0.

[0248] For rescue efficiency:

[0249] Based on the adaptive learning database, the historical rescue success rate of similar scenarios is obtained, and multiplied by 0.01 as the value of this indicator. The historical success rate reflects the probability of successful rescue operation under similar disaster conditions.

[0250] Based on the material management information system, the equipment matching degree is calculated according to the matching degree of rescue demand and existing rescue equipment, reflecting whether the existing equipment can meet the demand of rescue task. The equipment matching degree is calculated by formula (10):

[0251]

[0252] Where M is the equipment matching degree, indicating the degree to which the existing rescue equipment meets the demand of rescue task, which is a value between 0 and 1, the higher the value, the better the matching degree. i is the equipment type, n is the total number of equipment types, indicating the total number of different types of equipment involved in the rescue task.

[0253] By integrating the six evaluation indicators of the above three types of features, each grid area can be quantitatively scored to determine the priority of the rescue task. These indicators provide a scientific basis for rescue decision-making, helping rescue personnel to allocate resources reasonably and improve rescue efficiency and success rate.

[0254] Table 7 Decision-making dimension division table

[0255]

[0256]

[0257] S50: For the feature dimension of decision tree, CART algorithm is used to split the decision tree nodes to construct a binary tree and determine the best split point. The decision tree node splitting rule is to minimize the Gini impurity. The feature and split point that can maximize the reduction of Gini impurity are selected to optimize the classification effect of the decision tree. Multiple dimensions are considered, including disaster urgency, resource accessibility, and rescue efficiency. According to the output of the decision tree, the rescue task is divided into different feature priorities to achieve fast and accurate emergency response and optimal allocation of resources.

[0258] The feature priority ranking is as follows:

[0259] The decision tree of the present application is based on multi-condition judgment, which is used to determine the priority of a certain task (from priority 1 to priority 8). The following is the textual description of the decision tree:

[0260] First, check if the color level is red:

[0261] If it is red:

[0262] Check if the weight value is ≥ 0.8:

[0263] If yes, further check if the air drop feasibility is 1:

[0264] If yes, the task priority is priority 1.

[0265] If no, the task priority is priority 2.

[0266] If the weight value < 0.8, check if the equipment matching degree is > 0.6:

[0267] If yes, the task priority is priority 3.

[0268] If no, the task priority is priority 4.

[0269] If it is not red:

[0270] Check if the road pass index is > 0.7:

[0271] If yes, further check if the historical success rate is > 75%:

[0272] If yes, the task priority is priority 5.

[0273] If no, the task priority is priority 6.

[0274] If the road pass index ≤ 0.7, check if the color level is yellow:

[0275] If yes, the task priority is priority 7.

[0276] If no (i.e. the color level is neither red nor yellow), the task priority is priority 8.

[0277] The decision tree determines the task to be assigned to one of the eight different priorities by gradually judging the conditions of color level, weight value, air drop feasibility, equipment matching degree, road pass index, and historical success rate, etc.

[0278] S43: According to the feature priority ranking and the calculation results of the decision tree model, dynamically generate an initial rescue sequence; see Table 8; consider the data monitored and fed back in real time, such as road traffic conditions, new signs of life, etc.

[0279] Priority 1, trigger condition combination: red grid + weight ≥ 0.8 + can be air dropped; resource allocation strategy: immediately dispatch helicopters + unmanned aerial vehicle cluster delivery

[0280] Priority 2, trigger condition combination: red grid + weight ≥ 0.8 + cannot be air dropped; resource allocation strategy: heavy machinery to open up the road + satellite communication to support the assault team;

[0281] Priority 3: trigger condition combination: red grid + 0.6 ≤ weight < 0.8 + sufficient equipment; resource allocation strategy: standard rescue team double formation to jointly advance;

[0282] Priority 4: trigger condition combination: red grid + 0.6 ≤ weight < 0.8 + insufficient equipment; resource allocation strategy: implement two-stage rescue after emergency allocation of materials;

[0283] Priority 5: trigger condition combination: non-red grid + road unobstructed + historical high success rate; resource allocation strategy: light rescue team rapid response;

[0284] Priority 6: trigger condition combination: non-red grid + road unobstructed + historical low success rate; resource allocation strategy: expert command group to strengthen the team;

[0285] Priority 7: trigger condition combination: yellow grid + road not passable; resource allocation strategy: reconnaissance team to establish a temporary supply point;

[0286] Priority 8: trigger condition combination: blue / green grid; resource allocation strategy: not to act within the monitoring range;

[0287] Table 8 Priority Classification Table

[0288]

[0289] Further, in step S5, the rescue priority and resource allocation strategy are dynamically adjusted according to the real-time monitoring and feedback of the data related to disaster response and emergency rescue and the priority transition rules; including:

[0290] S51: Data collection, collect data from multi-source life detection devices such as life detectors, thermal imaging devices, and radar detectors; collect road traffic index and terrain change data through traffic monitoring system, satellite navigation, and unmanned aerial vehicle reconnaissance system; collect material consumption and equipment matching degree data from material management information system;

[0291] S52: Data analysis, calculate the comprehensive signal characteristic value according to the detection device data, and evaluate the intensity of life signs; update the road traffic index every 5 minutes according to the traffic data; when the material consumption reaches 50%, recalculate the equipment matching degree;

[0292] S53: Priority evaluation according to priority transition rules to obtain new priority data;

[0293] The priority transition rules are: grid color level upgrade (such as yellow→red), which increases the rescue priority of the related area by 3 levels; when the road disruption occurs and the road traffic index decreases by more than 0.3, the priority of the affected area is reduced by 2 levels; when new life signs are detected, the priority of the related area is increased to the current highest level +1;

[0294] S54: Update the feature dimension of the decision tree according to the new priority data, update the node splitting rule of the decision tree based on the Gini impurity minimization principle using the CART algorithm, and obtain the updated priority sequence;

[0295] S55: According to the updated priority sequence, optimize the scheduling and allocation of rescue resources, and dynamically adjust the resource allocation plan according to real-time feedback and priority transition rules;

[0296] The intelligent decision system of the application defines the node splitting rule and priority sequence generation standard, uses the CART algorithm and Gini impurity minimization method to construct the decision tree, dynamically generates and updates the rescue priority sequence. The system automatically triggers the priority transition according to the real-time monitoring data and specific conditions such as road conditions, life sign detection results, etc., to optimize the allocation of rescue resources and ensure the rapid response and flexible adjustment of rescue operations, thereby improving the efficiency and success rate of rescue.

[0297] The decision tree feature selection in steps S4 and S5 meets the requirements of the "rescue decision" category in the "Emergency Management Information Standard System Framework"; the priority classification maintains a mapping relationship with the "National Emergency Response Classification Standard"; the dynamic adjustment mechanism is certified by the China Emergency Management Society Expert Committee (Certification No. CESM-ER-2023-017), which shortens the response time by 50% compared to the traditional manual decision-making method, and reduces the resource misallocation rate to within 10%.

[0298] Further, in step S6, rescue efficiency index data is collected, a Bayesian optimization model is established according to the rescue efficiency index data, and parameter safety constraints and dynamic parameter adjustment rules are applied to iterate the rescue efficiency index data to realize continuous optimization of the rescue strategy; including:

[0299] S61: Collect and record the grid rescue time efficiency, resource consumption and success measurement indicators;

[0300] S62: Define an optimization objective function according to the grid rescue time efficiency, resource consumption and success measurement indicators; the optimization objective is to minimize the rescue time and resource consumption, and maximize the rescue success rate and key facility preservation rate;

[0301] S63: Determine the decision variable space, which includes adjustable parameters, their domains, and physical meanings; specifically, as follows:

[0302] The domain of the weight coefficient α is [0.5, 0.9], which represents the proportion of personnel density in the comprehensive weight;

[0303] Road traffic threshold R crit The domain of is [0.4, 0.8], which is the minimum traffic index that triggers road repair work;

[0304] Airdrop safety angle θ safe The domain of is [5°, 25°], which represents the maximum terrain inclination angle that allows drone airdrops;

[0305] The gain decay period τ is defined in the range [10, 60] minutes, which represents the half-life of the gain effect of the detection signal;

[0306] S64: Build a Bayesian optimization model and select prior distributions for the Bayesian optimization model parameters; select a kernel function for the Gaussian process, such as the Matérn 5 / 2 kernel, to capture data correlation and uncertainty; and select an acquisition function, such as the improved EI (Expected Improvement) weighted entropy term, to guide model parameter search and optimization.

[0307] S65: Clean and process the collected rescue effectiveness indicator data to construct features that aid model training; calculate the posterior distribution of key rescue strategy parameters based on the Bayesian optimization model; use the acquisition function to guide the optimization search for key rescue strategy parameters; conduct 1,000 Monte Carlo simulations based on the digital twin platform to evaluate the effectiveness of the optimized strategy;

[0308] S66: Implement the optimized rescue strategy in actual rescue operations; collect rescue feedback data after implementation and evaluate the effectiveness of the optimized strategy in actual rescue operations, including indicators such as rescue duration, resource consumption, and success rate;

[0309] S67: Feed new rescue data and evaluation results back to the Bayesian optimization model to continuously update and iterate the model; adjust model parameters and optimization strategies based on the new rescue data and evaluation results.

[0310] Steps S65-S67 achieve continuous optimization of the rescue strategy through data cleaning, feature engineering, parameter posterior distribution calculation, acquisition function optimization, new strategy simulation and deduction, actual rescue verification and performance indicator evaluation.

[0311] Further, the collection parameters of the time efficiency indicator in step S61 include response delay duration and operation duration; the response delay duration is the time difference (minute) from receiving the alarm to the arrival of the first rescue team at the scene, which is recorded when the event is triggered; the operation duration is the time difference (minute) from the start of the rescue operation to the personnel being rescued or the task being terminated, which is updated when the task is completed;

[0312] The collection parameters of the resource consumption indicator include equipment usage intensity and manpower input equivalent; the equipment usage intensity is calculated according to the usage of rescue equipment per hour; the manpower input equivalent is calculated by multiplying the number of rescue personnel by the working hours (person·hour), which is summarized at the end of the task;

[0313] The collection parameters of the success measure indicator include the life rescue success rate P life and the key facility preservation rate P facility ; the calculation formula of the life rescue success rate P life is the number of rescued survivors divided by the total number of detected life bodies, then multiplied by 100%, which is evaluated when the task is completed in stages; the key facility preservation rate P facility is calculated by the percentage of undamaged facilities to the total number of facilities, which is evaluated within 48 hours after the disaster; see Table 9 for details;

[0314] Table 9 Data Dimension Standardization Table

[0315]

[0316]

[0317] Further, the optimization objective function in step S62 is represented by formula (11):

[0318] θ=max(ω1P life +ω2P facility -ω3T total -ω4Q) (11)

[0319] Wherein, according to the “Multi-objective Optimization Weight Distribution Specification”, it is set as:

[0320] ω1=0.4, ω2=0.3, ω3=0.2, ω4=0.1;

[0321] P life is the life rescue success rate; P facility is the key facility preservation rate; T total is the total rescue time; Q is the resource consumption;

[0322] Further, the cleaning and processing of the collected rescue efficiency index data in step S65 includes: eliminating abnormal values, such as non-typical data with a response delay of >24 hours; constructing features that facilitate model training, including constructing spatio-temporal composite features, such as “night rescue correction coefficient”, to reflect the impact of rescue operations at different times and in different environments;

[0323] Further, step S65 further includes: when calculating the posterior distribution of the rescue strategy key parameters, applying parameter safety constraints to ensure that the model parameters are within the safety range; and dynamically adjusting the parameters of the Bayesian optimization model according to the dynamic parameter adjustment rule;

[0324] When using the acquisition function (such as Expected Improvement) for parameter optimization, the parameter safety constraints are applied to ensure that the optimized parameters are within the safety range; the parameter search strategy is dynamically adjusted according to the dynamic parameter adjustment rule; when simulating and deducing based on the digital twin platform, the parameter safety constraints are applied to ensure the safety and reliability of the simulation strategy; and the simulation parameters are dynamically adjusted according to the dynamic parameter adjustment rule to evaluate the effects of different strategies;

[0325] Step S66 further includes: when implementing the optimized strategy in actual rescue operations, applying parameter safety constraints to ensure the safety and reliability of the rescue operations; and dynamically adjusting the rescue operation parameters according to the dynamic parameter adjustment rule to adapt to changes in the field situation;

[0326] The dynamic parameter adjustment rule is: level I response: online real-time optimization, iterated every 15 minutes, calculated by a cloud computing cluster; level II response: offline batch optimization, iterated every 2 hours, calculated by an edge server; level III response: historical pattern matching, updated every 6 hours, calculated by a local workstation;

[0327] The parameter safety constraints are: the weight coefficient α cannot be lower than the lower limit specified in the geological disaster prevention regulations, α≥0.55; the adjustment range of the air drop safety angle is not more than 5° at a time to prevent drastic fluctuations.

[0328] The following are the specific application results of the emergency rescue priority dynamic decision-making method in a debris flow disaster scenario, and the results show:

[0329] (1) The school area is identified as a red grid (weight 0.85) in the initial division stage

[0330] (2) The motion model prediction shows that the impact probability of this area is reduced to within 50%, and it is downgraded to a yellow grid

[0331] (3) Infrared detection finds signs of life, the weight is restored to 0.70 and marked as flashing red

[0332] (4) The decision tree prioritizes the delivery of medical kits by drones and simultaneously dispatches search and rescue teams

[0333] (5) The actual rescue time is recorded to be 20% longer than the predicted value, and the optimization model adjusts the terrain coefficient;

[0334] This invention effectively solves the technical pain points of poor adaptability and insufficient integration of multi-source information in traditional static assessment methods, and significantly improves the accuracy of rescue resource allocation through three dynamic corrections.

[0335] The core advantages of this invention are reflected in three dimensions: dynamic adaptability, multi-source collaboration, and intelligent decision-making. Through technological innovation, it solves the key pain points in traditional disaster relief. The specific advantages are as follows:

[0336] (1) Dynamic adaptability advantage

[0337] Three-level correction mechanism: initial weight calculation: a four-color grading model based on surface attachments and population density; motion model correction: coupled fluid dynamics prediction (landslide / mudslide motion formula); real-time detection feedback: a weight gain and zeroing mechanism driven by the characteristic value of life signals; effect: the priority misjudgment rate is reduced by less than 10%;

[0338] Spatiotemporal coupling analysis, real-time interaction between static terrain parameters (slope, lithology) and dynamic disaster propagation (speed, range), and the drone laser rangefinder updates terrain elevation data every 30 minutes to trigger local reconstruction.

[0339] (2) Multi-source synergy advantage

[0340] Data fusion capabilities, initial DEM data, large-scale surface deformation monitoring (sub-meter accuracy), support for initial gridding and four-color grading; drones, TEM spatial resolution reaches sub-meter level, laser rangefinder mm-level elevation measurement, thermal imaging life detection (centimeter-level resolution), microwave life detector, support for hierarchical progressive updates of grid four-color grading.

[0341] Cross-system collaboration, color-coded grading based on comprehensive detection, combined with docking with the traffic management platform to obtain real-time road conditions (delay < 5 seconds) and automatic triggering of equipment pre-allocation by the material management system (inventory matching increased by 30%), provides technical support for decision-making priority grading, ensures the precise deployment of ground rescue personnel and materials, and guarantees maximum rescue efficiency and safety.

[0342] (3) Advantages of intelligent decision-making

[0343] Through a multi-dimensional decision tree model, an eight-level priority level is constructed in a six-dimensional feature space (disaster situation, resources, and efficiency); minute-level policy updates are achieved through dynamic adjustment of rules. The result: ensuring that the resource misallocation rate is controlled within 10%.

[0344] Bayesian optimization closed loop, continuous learning based on historical rescue data (>1000 case library), key parameters (weight coefficient α, air-drop safety angle θ) adaptive optimization, ensure the order of database update and continuous learning, maximize the efficiency and safety of rescue.

[0345] (4) Technical compliance advantage

[0346] Standard compatibility, compliance with 7 national standards (GB / T 36112) of geology, engineering, emergency support and other industries, seamless docking with "National Emergency Response System Planning".

[0347] Safety controllability, data desensitization processing (compliance with GB / T 39477), parameter adjustment amplitude hard constraint (such as single α fluctuation ≤0.05), TEM detection geological safety provides bottom layer safety support, multi-level life detection guarantee accuracy, intelligent system guarantee decision priority, success and failure grid case database guarantee continuous learning optimization, provide multi-level safety support for safety.

[0348] (5) Performance improvement comparison, see Table 10;

[0349] Table 10 Performance improvement comparison

[0350]

[0351] As Figure 2 shown, the second aspect of the present application provides an emergency rescue priority dynamic decision system for realizing the above decision method, comprising:

[0352] Multi-source data fusion module, for integrating data from different sources, including satellite remote sensing data, unmanned aerial vehicle surveying and mapping data and Internet of Things (IoT) sensor data; Provide comprehensive, real-time data support to better understand disaster conditions and impact range;

[0353] Dynamic grid processing engine, for identifying real-time changes in terrain and topologically reconstructing the grid according to these changes; To ensure that rescue decisions are based on the latest terrain and disaster data, improve the accuracy and timeliness of decisions;

[0354] Four-dimensional weight calculation unit, for calculating the weight of each grid by combining terrain slope, population density, infrastructure density and ecological sensitivity, evaluating the importance and priority of different regions, guiding resource allocation and rescue operations;

[0355] Visual decision interface, for providing three-dimensional situation deduction and dynamic planning of rescue path functions, enabling decision makers to intuitively understand and analyze disaster conditions, optimize rescue path and strategy;

[0356] An adaptive learning database is used to store historical disaster patterns and rescue effectiveness comparison data, to improve future rescue decisions by learning from historical data, to learn from experience and adapt to new rescue scenarios;

[0357] The workflow of the system includes:

[0358] Data collection: Collect and integrate disaster-related data through a multi-source data fusion module;

[0359] Data processing: Process and analyze data using a dynamic grid processing engine and a four-dimensional weight calculation unit to calculate the weight of each grid;

[0360] Decision support: Display analysis results through a visual decision interface to support decision-makers in three-dimensional situation deduction and rescue path planning.

[0361] Learning and optimization: The adaptive learning database continuously optimizes rescue strategies and decision models based on historical data and rescue results.

[0362] The emergency decision system of the present application aims to improve the efficiency and effectiveness of disaster emergency response by integrating various advanced technologies and methods, and to reduce the loss caused by disasters.

[0363] It should be noted that the emergency rescue priority dynamic decision system provided in the embodiment can be a computer program (including program code) running in a computer device, for example, the emergency rescue priority dynamic decision system is an application software; the emergency rescue priority dynamic decision system can be used to execute the corresponding steps in the above-mentioned method provided by the embodiments of the present application.

[0364] In some possible implementation manners, the emergency rescue priority dynamic decision system provided in the embodiment can be implemented in a combination of software and hardware, for example, the emergency rescue priority dynamic decision system provided in the embodiment of the present application can be a processor in the form of a hardware decoding processor, which is programmed to execute the emergency rescue priority dynamic decision method provided in the embodiment of the present application, for example, the processor in the form of a hardware decoding processor can adopt one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs) or other electronic elements.

[0365] In some possible implementation manners, the emergency rescue priority dynamic decision system provided by the embodiment can be implemented in a software manner, which can be software in the form of programs and plug-ins, and includes a series of modules to implement the emergency rescue priority dynamic decision method provided by the embodiment.

[0366] The emergency rescue priority dynamic decision system provided by the embodiment integrates geology, climatology and artificial intelligence technology, and constructs an intelligent emergency decision system integrating multi-source data and dynamic grid division, which significantly improves the accuracy and management efficiency of disaster warning compared with the prior art. The construction of the decision tree is optimized by using the CART algorithm, the intelligent sorting and dynamic adjustment of the rescue task are realized, the data island problem is effectively solved, and cross-disciplinary collaboration and information sharing are promoted. The key parameters are iteratively adjusted through the Bayesian optimization closed-loop system, which further improves the accuracy of resource allocation and the success rate of rescue, and provides a full-cycle and multi-dimensional solution from passive response to active prevention and control for disaster prevention, so that people's life and property safety and social stability are better protected under the double pressure of climate change and human activities.

[0367] The third aspect of the application also provides an electronic device, Figure 3 is a structural schematic diagram of the electronic device of the embodiment, as Figure 3 shown, the electronic device 1000 in the embodiment can include a processor 1001, a network interface 1004 and a memory 1005, in addition, the above-mentioned electronic device 1000 can further include a user interface 1003, and at least one communication bus 1002. Wherein, the communication bus 1002 is used to realize the connection communication between these components. Wherein, the user interface 1003 can include a display (Display), a keyboard (Keyboard), and the optional user interface 1003 can further include a standard wired interface, a wireless interface. The network interface 1004 can optionally include a standard wired interface, a wireless interface (such as a WI-FI interface). The memory 1005 can be a high-speed RAM memory, or a non-volatile memory, for example, at least one disk memory. The memory 1005 can optionally be at least one storage device located away from the aforementioned processor 1001. As Figure 3 shown, the memory 1005 as a computer readable storage medium can include an operating system, a network communication module, a user interface module and a device control application program.

[0368] As Figure 3In the electronic device 1000 shown, the network interface 1004 can provide network communication functions; the user interface 1003 is mainly used to provide an interface for user input; and the processor 1001 can be used to call a device control application stored in the memory 1005 to implement each step of the above decision-making method.

[0369] It should be understood that in some possible implementations, the above processor 1001 can be a central processing unit (CPU), and the processor can also be other general-purpose processors, DSPs, ASICs, FPGAs or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The memory can include read-only memory and random access memory, and provide instructions and data for the processor. A part of the memory can also include non-volatile random access memory. For example, the memory can also store device type information.

[0370] In a specific implementation, the electronic device 1000 described above can execute the implementation manner provided by each step in the above Figure 1 , and specific implementation manners provided by each step are described above, which will not be repeated here.

[0371] The electronic device provided in the embodiment fuses geology, climatology and artificial intelligence technology, and constructs an intelligent emergency decision-making system integrating multi-source data and dynamic grid division. Compared with the prior art, the accuracy and management efficiency of disaster warning are significantly improved. The construction of the decision tree is optimized by using the CART algorithm, the intelligent sorting and dynamic adjustment of the rescue task are realized, the data island problem is effectively solved, and cross-disciplinary collaboration and information sharing are promoted. The key parameters are iteratively adjusted through the Bayesian optimization closed-loop system, and the accuracy of resource allocation and the success rate of rescue are further improved, providing a full-cycle, multi-dimensional solution from passive response to active prevention and control for disaster prevention, so that people's life and property safety and social stability are better protected under the double pressure of climate change and human activities.

[0372] The embodiment of the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method provided by each step in the above Figure 1 , and specific implementation manners provided by each step are described above, which will not be repeated here.

[0373] The computer readable storage medium provided by the embodiment fuses geology, climatology and artificial intelligence technology, and constructs an intelligent emergency decision system integrating multi-source data and dynamic grid division. Compared with the prior art, the accuracy of disaster warning and management efficiency are significantly improved. The construction of the decision tree is optimized by using the CART algorithm, the intelligent sorting and dynamic adjustment of the rescue task are realized, the data island problem is effectively solved, and the cross-disciplinary collaboration and information sharing are promoted. The key parameters are iteratively adjusted through the Bayesian optimization closed-loop system, and the accuracy of resource allocation and the success rate of rescue are further improved, providing a full-cycle and multi-dimensional solution from passive response to active prevention and control for disaster prevention, so that the safety of people's life and property and social stability can be better protected under the double pressure of climate change and human activities.

[0374] Any reference to storage, memory, database or other medium used herein includes non-volatile and / or volatile storage. Non-volatile storage can include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile storage can include random access memory (RAM), or external cache memory. By way of illustration, and not limitation, RAM is available in many forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), Rambus DRAM (RDRAM), direct Rambus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM).

[0375] Those skilled in the art can easily understand that the above description is only a preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A dynamic decision-making method for emergency rescue priority, characterized in that: The steps include: S1: Based on the distribution of surface attachments in the disaster area and DEM digital elevation data, the disaster area is divided into grids, and the importance weight of the grids is calculated, dynamically corrected, and visualized using a four-color grading model; S2: Based on the movement zoning characteristics of landslides or debris flows, combined with the fluid dynamics model, the material movement trajectory after the geological disaster occurs is predicted, the potential impact range of each grid is dynamically adjusted, the weight value is updated, and secondary grid coloring is performed; S3: Acquire multi-source life signal data of trapped persons through multi-source life detection equipment, calculate the comprehensive signal characteristic value of the multi-source life signal data using a multi-source signal fusion model, determine the valid signal according to the signal strength, perform weight gain processing on the response grid, perform weight zeroing and color reset on the unresponsive grid, and form the third grid coloring; S4: Construct a multi-dimensional decision tree model to generate a rescue priority sequence based on the weight value, coloring level, and traffic accessibility parameters; S5: Dynamically adjust rescue priorities and resource allocation strategies based on real-time monitoring and feedback of disaster response and emergency rescue-related data and priority transition rules; S6: Collect rescue effectiveness index data, establish a Bayesian optimization model based on the rescue effectiveness index data, apply parameter safety constraints and dynamic parameter adjustment rules to iterate the rescue effectiveness index data, and achieve continuous optimization of the rescue strategy.

2. The method for dynamic decision-making of emergency rescue priority according to claim 1, characterized in that: Step S1 includes: S11: Collect distribution information of surface attachments in the disaster area; S12: Obtain DEM digital elevation data of the disaster area; S13: Collect population distribution data in disaster areas; S14: Determine the size of the grid based on the scope and accuracy requirements of the disaster area; use GIS software or programming tools to divide the disaster area into regular grids; for each grid, extract its terrain characteristics and surface attachment information based on DEM digital elevation data and surface attachment distribution information; S15: Calculate the comprehensive weight of each grid using a linear weighting model; S16: Determine the grading standard based on the comprehensive weight value; S17: Use GIS software or programming tools to color each grid according to the classification standard; display the color of each grid on the map to form an intuitive visualization effect; S18: Adjust the grid weight according to the development of the disaster and real-time monitoring. Increase the weight when the grid is in the main flow line of the debris flow, and reduce the facility weight when the slope is greater than 25°. Set the weight to the highest level when vital signs are detected, and reduce the weight if there is no signal for two consecutive hours.

3. The method for dynamic decision-making of emergency rescue priority according to claim 2, characterized in that: In step S15, the comprehensive weight W is calculated by formula (1): W = α × W population + β × W 设施 (1) Among them, α is the population weight correction coefficient, β is the facility weight correction coefficient, α+β=1, α=0.7, β=0.3; W 人口 is the personnel density weight; W 设施 is the weight of surface attachments; Step S16 includes: dividing the grid into four levels according to the comprehensive weight value and assigning corresponding color identification; The comprehensive weight interval is ≥0.8, and the quantitative value of population density is >100 people / km 2 (or single building ≥50 people), lifeline projects are classified as Level I, indicated by red, and the emergency response requirements are: immediate response, with priority given to deploying heavy rescue equipment; The comprehensive weight range is 0.6 to 0.79, and the population density is quantified to 50 to 100 people / km 2 , hazardous source facilities are classified as Level II, indicated by yellow, and the emergency response requirements are: deploy a professional rescue team within 2 hours; The comprehensive weight range is 0.4 to 0.59, and the population density is quantified to 10 to 50 people / km 2 , general residential areas are classified as Level I, represented by blue, and the emergency response requirements are: initiate basic rescue within 4 hours; The comprehensive weight interval is <0.4, and the quantitative value of personnel density is <10 people / km 2 The ecological protection zone is classified as Level I, represented by green. The emergency response requirements are: monitoring and early warning, and dynamic intervention based on the disaster situation.

4. A method for dynamic decision-making on emergency rescue priority according to any one of claims 1 to 3, characterized in that: In step S2, based on the movement zoning characteristics of landslides or debris flows, the material movement trajectory after the geological disaster is predicted in combination with the fluid dynamics model, the potential impact range of each grid is dynamically adjusted, the weight value is updated, and secondary coloring is performed; including: S21: The calculation formula for the movement distance of two types of disasters, landslide and debris flow; S22: Generate disaster impact probability distribution map through spatial discretization, motion simulation and probability coverage; S23: Dynamically modify the grid weight and color according to the disaster impact probability distribution map; S24: Modify the parameters in the movement distance formula based on real-time terrain feedback, trigger local grid topology reconstruction, and generate a new disaster impact probability distribution map; S25: Develop technical implementation standards based on new disaster impact probability distribution maps; Step S22 includes the following steps: S221: Spatial discretization, dividing the DEM terrain data of the study area into a regular grid system; extracting the necessary terrain features for each grid; S222: Movement Simulation: Select an appropriate movement model based on the disaster type; set model parameters based on existing disaster characteristics and terrain conditions; run the model to simulate the movement path and range of the disaster material on the terrain to determine the potential impact of each grid; S223: Probability Coverage: Using the Monte Carlo method, a large number of random simulations are performed to estimate the probability of each grid being affected by a disaster. Based on the simulation results, a disaster impact probability distribution map is generated, where the color or value of each grid represents the probability of its being affected by the disaster. Step S23 specifically includes: when the grid is in the main flow line of the debris flow with a probability of ≥70%, the weight is increased by 0.2, with an upper limit of 1.0, and the grid color is upgraded from red to a dark red flashing warning; when the grid is at the edge of the movement with a probability of 30%≤<70%, the weight is increased by 0.1, and the color changes from yellow to orange, and from blue to light blue; when the grid is outside the prediction range with a probability of <30%, the weight is reduced to 50% of the original value, with a lower limit of 0.2, and the green grid returns to the background color; The technical implementation standards in step S25 include: When the landslide moves a distance L slide and debris flow movement distance L debris satisfy: L slide ≥300m or L debris When the distance is ≥500m, it is a red alert zone; When the landslide movement distance and debris flow movement distance meet the following conditions: 150m≤L slide <300m or 200m≤L debris When the distance is less than 500m, it is a yellow warning zone.

5. A method for dynamic decision-making on emergency rescue priority according to any one of claims 1 to 3, characterized in that: In step S3, multi-source life signal data of the trapped person is obtained through a multi-source life detection device, a multi-source signal fusion model is used to calculate the comprehensive signal characteristic value of the multi-source life signal data, valid signals are determined according to signal strength, weight gain processing is performed on the response grid, and weights of the unresponsive grids are reset to zero and colors are reset to form a third grid coloring; including: S31: Collect vital signs, thermal imaging, and motion signals of trapped persons through life detectors, thermal imaging equipment, and radar detectors; S32: fusing the trapped person's vital sign signals, thermal imaging signals, and motion signals using a multi-source signal fusion model to obtain a comprehensive signal feature value; S33: Developing a valid signal determination standard based on the comprehensive signal characteristic value; a comprehensive signal characteristic value S ≥ 0.7 is defined as a strong response signal, marked with a red flashing mark; a comprehensive signal characteristic value 0.4 ≤ S < 0.7 is defined as a weak response signal, marked with a yellow pulse; a comprehensive signal characteristic value S < 0.4 is defined as an invalid signal, and no weight adjustment is triggered; S34: Modify the grid weight and color according to the effective signal judgment standard; specifically, perform weight gain processing on the responsive grid, and perform weight zeroing, color resetting and data tracing processing on the unresponsive grid.

6. The method for dynamic decision-making of emergency rescue priority according to claim 5, characterized in that: The multi-source signal fusion model in step S32 is expressed by formula (4): S=λ1·S 生命 +λ2·S 热力 +λ3·S 运动 (4) Among them, S is the comprehensive signal characteristic value; S 生命 is the characteristic value of vital sign signal; S 热力 is the characteristic value of thermal imaging signal; S 运动 is the motion signal characteristic value; λ1 = 0.5, λ2 = 0.3, λ3 = 0.2; The weight gain processing in step S34 includes: Weight after gain W new Calculation is expressed by formula (8): W new =min(1.0,W 原 +ΔW·log2(1+S)) (8) Where ΔW is the gain coefficient, which is 0.3 for strong response and 0.15 for weak response; W 原 is the original weight value; According to the time decay mechanism, if the signal is not updated for every 15 minutes, the gain value will be reduced by 50%; if there is no update for 3 consecutive times, the gain effect will be reset to zero; According to the color synchronization rule, the weight W after gain new The grid area with a value of ≥0.8 is displayed with a dark red background and a golden pulse frame; the system triggers an audible and visual alarm and sends an emergency notification to relevant personnel via SMS push notification. Set the weight after gain to 0.6≤W new The grid area with a value less than 0.8 is displayed with an orange gradient animation; a pop-up window is displayed to alert the user to this area. The weight W after the gain new Grid areas with a value less than 0.6 retain their original color, indicating that the area has a lower priority. The system only records information about this area in the system log and does not issue additional reminders or alarms. The triggering conditions for resetting the weight of the unresponsive grid to zero in step S34 are: first-level judgment, no valid signal for two consecutive detection cycles; second-level judgment, no signs of life confirmed by manual review; The color reset rules are as follows: for red / yellow grids, the weight is reset to zero, the pre-disaster background color is restored, and a semi-transparent "X" mark is superimposed; for blue / green grids, the weight is reset to zero, the original color is retained but the transparency is increased to 70%, and it is displayed as a dotted border; The data tracing mechanism is as follows: the weighted zero grid automatically generates a signal disappearance report, and historical data is retained for 72 hours for manual review.

7. A method for dynamic decision-making on emergency rescue priority according to any one of claims 1 to 3 and 6, characterized in that: In step S4, a multi-dimensional decision tree model is constructed to generate a rescue priority sequence based on the weight values, coloring levels, and traffic accessibility parameters; including: S41: Define the feature dimensions of the decision tree; this includes three categories of six-dimensional decision features; the three categories include disaster urgency, resource accessibility, and rescue effectiveness; the six dimensions include grid color level and comprehensive weight value for assessing disaster urgency, road accessibility index and airdrop feasibility for assessing resource accessibility, and historical success rate and equipment matching for assessing rescue effectiveness; S50: Based on the feature dimension of the decision tree, the CART algorithm is used to split the decision tree nodes, construct a binary tree, select the features and split points that can minimize the Gini impurity, and divide the rescue tasks into different feature priorities; S43: Dynamically generate an initial rescue sequence based on feature priority sorting and decision tree model calculation results.

8. The method for dynamic decision-making of emergency rescue priority according to claim 7, characterized in that: The rescue sequence in step S43 includes: Priority 1, trigger condition combination is: red grid + weight ≥ 0.8 + airdrop available; resource delivery strategy is: immediate dispatch of helicopter + drone cluster delivery Priority 2, trigger conditions are: red grid + weight ≥ 0.8 + no airdrop; resource deployment strategy: heavy machinery to open the way + satellite communication support commando; Priority 3: The trigger conditions are: red grid + 0.6 ≤ weight < 0.8 + sufficient equipment; the resource allocation strategy is: standard rescue team double formation and coordinated advancement; Priority 4: The trigger condition combination is: red grid + 0.6 ≤ weight < 0.8 + insufficient equipment; the resource allocation strategy is: emergency allocation of supplies followed by a second-stage rescue. Priority 5: The trigger conditions are: non-red grid + clear roads + historically high success rate; the resource allocation strategy is: light rescue team rapid response; Priority 6: The trigger conditions are: non-red grid + clear roads + historically low success rate; the resource allocation strategy is: expert command team to reinforce the team; Priority 7: Trigger condition combination: Yellow grid + Impassable road; Resource deployment strategy: Send an advance reconnaissance team to establish a temporary supply point; Priority 8: Trigger condition combination: Blue / Green Grid; Resource allocation strategy: Include in monitoring scope and take no action for now; Step S41 further includes: According to the urgency of the disaster: The dynamic grid processing engine divides the grid area into four levels according to the severity of the disaster and assigns different values: Red-Level I: The value is 4, indicating the most serious disaster and requiring the highest priority rescue; Yellow-Level II: The value is 3, indicating a relatively serious disaster; Blue-Level III: The value is 2, indicating a moderate disaster; Green-Level IV: The value is 1, indicating a relatively minor disaster. Based on the four-dimensional weight calculation unit, the comprehensive weight value of each grid is calculated and normalized to the range of 0 to 1; Regarding resource accessibility: Obtain road traffic index based on real-time traffic big data platform; The road traffic index is expressed by formula (9): R road =e -0.1t (9), Where t represents the time required to reach the grid area; The feasibility of airdrops is determined through the drone reconnaissance system. Airdrop feasibility is represented by a 0-1 Boolean value, indicating whether airdrop rescue can be carried out in the grid area. If the terrain flatness is greater than 5° and there are no high-voltage wires blocking the area, the airdrop is feasible, with a Boolean value of 1; otherwise, the airdrop is not feasible, with a Boolean value of 0. Regarding rescue effectiveness: Based on the adaptive learning database, the historical rescue success rate for similar scenarios is obtained and multiplied by 0.01 as the value of the indicator; Based on the material management information system, the equipment matching degree is calculated according to the matching degree between rescue needs and existing rescue equipment.

9. A method for dynamic decision-making on emergency rescue priority according to any one of claims 1 to 3, 6, and 8, characterized in that: Step S5 includes: S51: Data collection: Collect data from multi-source life detection equipment; collect road traffic index and terrain change data through traffic monitoring systems, satellite navigation, and drone reconnaissance systems; collect material consumption and equipment matching data from the material management information system; S52: Data analysis: Calculates comprehensive signal characteristic values ​​based on detection equipment data to assess the strength of life signs; updates road traffic index based on traffic data every 5 minutes; and recalculates equipment compatibility when material consumption reaches 50%; S53: Perform priority evaluation according to the priority transition rule to obtain new priority data; S54: Update the feature dimensions of the decision tree according to the new priority data, use the CART algorithm, and update the node splitting rules of the decision tree based on the principle of minimizing Gini impurity to obtain the updated priority sequence; S55: Optimize the scheduling and allocation of rescue resources based on the updated priority sequence, and dynamically adjust the resource allocation plan based on real-time feedback and priority transition rules; The priority transition rules in step S53 are as follows: if the grid color level is upgraded, the rescue priority of the relevant area will be increased by 3 levels; when the road is suddenly interrupted and the road traffic index drops by more than 0.3, the priority of the affected area will be reduced by 2 levels; when new signs of life are detected, the priority of the relevant area will be increased to the current highest level + 1.

10. A method for dynamic decision-making on emergency rescue priority according to any one of claims 1-3, 6, and 8, characterized in that: Step S6 includes: S61: Collect and record the rescue time efficiency, resource consumption and success metrics of each grid; S62: Define an optimization objective function based on the rescue time efficiency, resource consumption, and success metric of each grid; the optimization goal is to minimize the rescue time and resource consumption while maximizing the rescue success rate and the key facility preservation rate; S63: Determine the decision variable space; S64: Building a Bayesian optimization model; S65: Clean and process the collected rescue effectiveness indicator data to construct features that aid model training; calculate the posterior distribution of key rescue strategy parameters based on the Bayesian optimization model; use the acquisition function to guide the optimization search for key rescue strategy parameters; perform Monte Carlo simulation based on the digital twin platform to evaluate the effectiveness of the optimized strategy; S66: Implement the optimized rescue strategy in actual rescue operations; collect rescue feedback data after implementation and evaluate the effectiveness of the optimized strategy in actual rescue operations; S67: Feed new rescue data and evaluation results back into the Bayesian optimization model to continuously update and iterate the model; adjust model parameters and optimization strategies based on new rescue data and evaluation results; The collection parameters of the time efficiency index in step S61 include response delay time and operation duration; response delay time is recorded when the event is triggered; operation duration is updated when the task ends; The collection parameters of the resource consumption index include equipment usage intensity and manpower input equivalent; equipment usage intensity is calculated on an hourly basis; manpower input equivalent is summarized at the end of the mission; The acquisition parameters of the success metric include the life rescue success rate P life and key facility preservation rate P facility ; Life rescue success rate P life Evaluate at the end of each phase of the mission; Key facility preservation rate P facility Assessment within 48 hours of disaster; The optimization objective function in step S62 is expressed by formula (10): θ=max(ω1P life +ω2P facility -ω3T total -ω4Q) (10) Among them, according to the "Multi-objective Optimization Weight Allocation Specification": ω1=0.4, ω2=0.3, ω3=0.2, ω4=0.1; PP life is the success rate of life rescue; P facility is the key facility preservation rate; T total is the total rescue time; Q is the resource consumption; In step S63, the decision variable space includes adjustable parameters and their domains and physical meanings; specifically, they are as follows: The domain of the weight coefficient α is [0.5, 0.9], which represents the proportion of personnel density in the comprehensive weight; Road traffic threshold R crit The domain of is [0.4, 0.8], which is the minimum traffic index that triggers road repair work; Airdrop safety angle θ safe The domain of is [5°, 25°], which represents the maximum terrain inclination angle that allows drone airdrops; The gain decay period τ is defined in the range [10, 60] minutes, which represents the half-life of the gain effect of the detection signal.

11. A method for dynamic decision-making on emergency rescue priority according to any one of claims 1-3, 6, and 8, characterized in that: The dynamic parameter adjustment rules in step S6 are as follows: Level I response, online real-time optimization, iterating every 15 minutes, using cloud computing clusters for calculation; Level II response, offline batch optimization, iterating every 2 hours, using edge servers for calculation; Level III response, historical pattern matching, updating every 6 hours, using local workstations for calculation; The parameter safety constraints are: the weight coefficient α shall not be lower than the lower limit α≥0.55 stipulated in the Regulations on the Prevention and Control of Geological Disasters; the adjustment range of the airdrop safety angle shall not exceed 5° at a time to prevent violent fluctuations.

12. An emergency rescue priority dynamic decision-making system, characterized in that: A method for implementing a dynamic decision-making method for emergency rescue priority according to any one of claims 1 to 11, comprising: Multi-source data fusion module for integrating data from satellite remote sensing data, drone mapping data, and IoT sensors; A dynamic mesh processing engine that identifies real-time changes in the terrain and reconstructs the mesh topology based on these changes; A four-dimensional weight calculation unit is used to calculate the weight of each grid by combining four parameters: terrain slope, population density, infrastructure density, and ecological sensitivity. This allows the importance and priority of different areas to be assessed, guiding resource allocation and rescue operations. A visual decision-making interface provides three-dimensional situation simulation and dynamic planning of rescue routes, enabling decision-makers to intuitively understand and analyze disaster situations and optimize rescue routes and strategies. An adaptive learning database is used to store historical disaster patterns and rescue effectiveness comparison data, and improve future rescue decisions by learning from historical data.

Images