AI and IoT-Based Emergency Medical Command Method and System for Emergencies

Abstract

The present invention provides an emergency medical command method and system based on AI and the Internet of Things, which relates to the technical field of emergency medical command. It includes collecting thermal imaging data through an infrared sensor to obtain the body temperature and location information of the wounded, calculating the regional danger level in combination with environmental parameters, and generating a scene situation assessment result. Based on the assessment result, the rescue area is divided and a resource scheduling plan is formulated, which is visually presented on the digital twin platform, and the rescue situation is superimposed through augmented reality technology. The resource configuration is dynamically adjusted according to the real-time location information. The present invention can realize intelligent medical command at the scene of emergencies, improve the rescue efficiency and resource utilization rate.

Description

Technical Field

[0001] The present invention relates to emergency medical command technology, and particularly to an emergency medical command method and system based on AI and the Internet of Things for emergencies. Background Art

[0002] In emergencies, the timeliness and accuracy of emergency medical rescue are crucial. The existing emergency medical command systems have the following deficiencies: First, the early warning and classification of emergencies are not precise enough to quickly judge the danger level of the event; second, the resource scheduling efficiency is low and cross-regional and multi-departmental collaborative linkage cannot be achieved; third, there is a lack of intelligent decision-making support and the rescue plan cannot be dynamically adjusted according to real-time data.

[0003] With the development of artificial intelligence and Internet of Things technology, it has become possible to apply technologies such as intelligent perception and digital twin to emergency medical command. However, there is still a lack of a comprehensive solution that can achieve real-time monitoring of the status of the wounded, intelligent assessment of the on-site situation, and optimal scheduling of medical resources. Therefore, there is an urgent need for an emergency medical command method based on AI and the Internet of Things for emergencies to improve the scientificity and efficiency of emergency medical rescue. Summary of the Invention

[0004] Embodiments of the present invention provide an emergency medical command method and system based on AI and the Internet of Things for emergencies, which can solve the problems in the prior art.

[0005] In the first aspect of the embodiments of the present invention,

[0006] An emergency medical command method based on AI and the Internet of Things for emergencies is provided, including:

[0007] Collecting thermal imaging data of the emergency site through an infrared sensor, extracting the body temperature distribution data and the location data of the wounded from the thermal imaging data, determining the environmental monitoring range according to the location data of the wounded, and collecting environmental parameter data within the range through an environmental sensor;

[0008] Constructing a body temperature change trend graph according to the body temperature distribution data of the wounded, calculating the body temperature abnormality degree value, and at the same time calculating the environmental threat degree value according to the environmental parameter data, calculating the regional danger level based on the body temperature abnormality degree value and the environmental threat degree value, and counting the wounded distribution density according to the location data of the wounded, and generating a scene situation assessment result according to the regional danger level and the wounded distribution density;

[0009] Based on the scene situation assessment result, dividing the rescue area according to the wounded distribution density, and determining the resource allocation weight of each rescue area based on the regional danger level, and generating a medical resource scheduling plan;

[0010] Build a 3D model of the emergency site in the digital twin platform, visually map the scene situation assessment results and the medical resource scheduling plan, overlay the rescue situation annotations in real time through augmented reality technology, and update the location information of rescue personnel and medical equipment in real time based on the target tracking algorithm;

[0011] According to the location information of rescue personnel and medical equipment, dynamically adjust the medical resource scheduling plan through a recursive optimization algorithm to achieve real-time optimal allocation of medical resources.

[0012] In an alternative embodiment,

[0013] Collect the thermal imaging data of the emergency site through an infrared sensor, and extract the casualty body temperature distribution data and casualty location data from the thermal imaging data, including:

[0014] Collect the thermal imaging data of the emergency site through an infrared sensor, perform adaptive histogram equalization enhancement processing on the thermal imaging data, obtain the enhanced thermal imaging data, determine the filter window size based on the local temperature variance of the enhanced thermal imaging data, and use the adaptive median filter algorithm to denoise the enhanced thermal imaging data to obtain the denoised thermal imaging data;

[0015] Establish a mapping function between the gray value and the actual temperature according to the temperature response characteristics of the infrared sensor, convert the gray value in the denoised thermal imaging data into a temperature value, generate a temperature distribution map, set a double-threshold detection interval based on the normal body temperature range of the human body, and perform regional connectivity analysis on the pixel points whose temperature values fall within the double-threshold detection interval in the temperature distribution map to obtain the candidate casualty area;

[0016] Perform morphological processing on the candidate casualty area, extract the region contour features, area features and aspect ratio features, and screen out the regions that do not conform to the features according to the human body size feature template to obtain the confirmed casualty area;

[0017] Establish a temperature field distribution model of the confirmed casualty area, extract the temperature numerical sequence, temperature gradient distribution and temperature uniformity characteristics in the area as the casualty body temperature distribution data; calculate the minimum circumscribed rectangle according to the contour characteristics of the confirmed casualty area, and use the rectangle center point coordinates as the casualty location data.

[0018] In an alternative embodiment,

[0019] Construct a body temperature change trend chart based on the casualty body temperature distribution data, calculate the body temperature abnormality degree value, and at the same time calculate the environmental threat degree value according to the environmental parameter data, calculate the regional danger level based on the body temperature abnormality degree value and the environmental threat degree value, count the casualty distribution density in combination with the casualty location data, and generate the scene situation assessment result according to the regional danger level and the casualty distribution density, including:

[0020] Establish a body temperature monitoring window, divide the body temperature monitoring window into an observation interval and a prediction interval, perform piecewise fitting on the body temperature data in the observation interval to extract the fluctuation law, establish a body temperature baseline based on the prediction interval, calculate the degree of deviation of the fluctuation law from the body temperature baseline to obtain the fluctuation anomaly feature, and judge the body temperature anomaly degree value according to the fluctuation anomaly feature;

[0021] Construct an influence network containing environmental parameters according to the environmental parameter data, use the leading abnormal parameter as the root node and the affected parameter as the child node to establish a directed graph structure of the transfer relationship between parameters, analyze the abnormal diffusion path based on the directed graph structure, and calculate the environmental threat degree value in combination with the scope and speed of abnormal diffusion;

[0022] Dynamically update the inference rules of the neural network according to the change trends of the body temperature anomaly degree value and the environmental threat degree value, and calculate the regional danger level using the inference rules;

[0023] Conduct spatial distribution analysis on the casualty location data, establish a casualty movement trajectory map, extract the trajectory intersection points as key nodes, and calculate the degree of personnel aggregation at each key node to obtain the casualty distribution density;

[0024] Use the regional danger level as the weight coefficient to perform weighted calculation on the casualty distribution density to obtain the scene situation assessment result.

[0025] In an optional embodiment,

[0026] Conduct spatial distribution analysis on the casualty location data, establish a casualty movement trajectory map, extract the trajectory intersection points as key nodes, and calculate the degree of personnel aggregation at each key node to obtain the casualty distribution density, including:

[0027] Perform segmented processing on the casualty location data according to the time sequence to obtain a time window sequence, calculate the change of the position coordinates within the time window sequence to obtain the displacement vector, connect the displacement vectors in sequence to construct a casualty movement trajectory map, perform smoothing processing on the casualty movement trajectory map to eliminate abnormal jumps, and extract the target movement path according to the smoothed trajectory;

[0028] Draw the target movement path in the spatial coordinate system, calculate the intersection angle and the shortest distance between the paths, determine the trajectory intersection points when the intersection angle and the shortest distance meet the preset conditions, count the intersection frequencies at each trajectory intersection point, calculate the number of personnel within the unit area around the trajectory intersection points to obtain the degree of personnel aggregation, and determine the key nodes according to the weighted values of the intersection frequencies and the degree of personnel aggregation;

[0029] A local coordinate system is established with the key node as the center. The cross frequency and the degree of personnel aggregation are respectively mapped to the size of the influence range and the attenuation rate. A node influence field is constructed according to the influence range and the attenuation rate, and multiple node influence fields are superimposed to generate a distribution influence map;

[0030] The real-time change characteristics of the casualty movement trajectory map are converted into dynamic weights, and the distribution influence map is dynamically weighted to obtain a density distribution model;

[0031] The target area is divided into grid cells. The distance between the center of each grid cell and the key node is calculated. According to the distance and the density distribution model, the reference density value in each grid cell is calculated, and the time-varying characteristics of the dynamic weights are superimposed on the reference density value to obtain the casualty distribution density reflecting the dynamic distribution of personnel.

[0032] In an optional embodiment,

[0033] Based on the scenario situation assessment result, the rescue area is divided according to the casualty distribution density, and the resource allocation weight of each rescue area is determined based on the regional danger level. The generated medical resource scheduling plan includes:

[0034] Receive the scenario situation assessment result and the casualty distribution density data. Based on the cellular automaton, a prediction model is constructed to analyze the change trend of the scenario situation, and the scenario threat degree and the passable path of each area at the future moment are obtained;

[0035] Calculate the density difference between adjacent areas according to the casualty distribution density data. The areas with the density difference less than the density threshold and the spatial distance less than the distance threshold are marked as areas to be merged. Combine the scenario threat degree to optimize the boundary of the areas to be merged to generate rescue partitions;

[0036] Based on the corresponding relationship between the medical resource input volume and the casualty treatment effect in the historical rescue data, calculate the resource utilization efficiency of each rescue partition, and calculate the regional danger level according to the casualty distribution density, the scenario threat degree and the resource utilization efficiency in the rescue partition;

[0037] Construct a resource transfer network with the rescue partition as the node and the passable path as the edge. Perform weighted calculation on the regional danger level and the resource utilization efficiency to obtain the resource allocation weight, and generate a medical resource scheduling plan with the minimum transfer loss based on the resource transfer network;

[0038] Monitor the change of the scenario situation assessment result. When the change of the scenario threat degree exceeds the situation threshold, re-divide the rescue partition based on the updated scenario threat degree and the casualty distribution density, adjust the regional danger level and the resource allocation weight, and update the medical resource scheduling plan.

[0039] In an optional embodiment,

[0040] Construct a prediction model based on cellular automata, analyze the changing trend of the scene situation, and obtain the scene threat level and passable paths at future moments for each region, including:

[0041] Divide the rescue scene into regular grid cells, map the road traffic status, building damage degree, hazard source intensity, and environmental threat level in the scene situation data to each grid cell to form the initial state matrix of the grid cells;

[0042] Calculate the evolution value of the road traffic status according to the passing frequency of rescue vehicles, calculate the evolution value of the building damage degree according to the secondary disaster monitoring data, calculate the diffusion value of the hazard source intensity and environmental threat level between adjacent grid cells, and construct a state transition equation with the evolution value of the road traffic status, the evolution value of the building damage degree, the hazard source intensity, and the diffusion value of the environmental threat level;

[0043] Use iterative calculation to solve the state transition equation to obtain the state matrix of each grid cell at the prediction moment, and calculate the scene threat level at the prediction moment according to the state matrix;

[0044] Calculate the passing probability based on the road traffic status and scene threat level of the grid cells, construct a path cost function, and use a search algorithm to solve for the passable paths;

[0045] Receive the scene situation data in real time. When the data is updated, map the updated data to a new initial state matrix, re - execute the prediction calculation, and dynamically update the scene threat level and passable paths.

[0046] In an alternative embodiment,

[0047] Construct a three - dimensional model of the emergency site in the digital twin platform, perform visual mapping on the scene situation assessment results and the medical resource scheduling plan, overlay the rescue situation annotations in real time through augmented reality technology, and update the location information of rescue personnel and medical equipment in real time based on the target tracking algorithm, including:

[0048] Use vehicle - mounted lidar in the digital twin platform to obtain the point cloud data of the emergency site, register the point cloud data to obtain a dense point cloud, segment the scene structure of the dense point cloud based on the region - growing algorithm to generate a three - dimensional model of the emergency site, and establish a unified spatial coordinate system in the three - dimensional model;

[0049] Based on the unified spatial coordinate system, map the scene threat level in the scene situation assessment results to the three - dimensional model to form a situation heat map, convert the equipment layout data and personnel deployment data in the medical resource scheduling plan into position coordinates in the three - dimensional scene and generate situation annotations to obtain a digital twin scene with situation information;

[0050] The feature point matching algorithm is used to realize the real-time registration of the mobile device camera and the digital twin scene with situational information, establish the mapping relationship between the camera view and the unified space coordinate system, and overlay and display the situational heat map and situational annotations in the camera view through augmented reality technology to obtain the augmented reality situational view;

[0051] Based on the target tracking algorithm, the position information of the medical device is obtained in real time, the position information collected by the rescue personnel positioning terminal is received, the position information is coordinate-transformed according to the unified space coordinate system, and the transformed position information is overlaid and displayed in the augmented reality situational view based on the mapping relationship;

[0052] When the scene situation assessment result is updated, a new situational heat map is generated; when the medical resource scheduling plan changes, the situational annotations are updated; when the position information of the rescue personnel and the position information of the medical device are updated, the position display in the augmented reality situational view is updated in real time based on the mapping relationship.

[0053] In the second aspect of the embodiments of the present invention,

[0054] An emergency medical command system based on AI and the Internet of Things is provided, including:

[0055] The first unit is used to collect the thermal imaging data of the emergency scene through an infrared sensor, extract the casualty body temperature distribution data and casualty position data according to the thermal imaging data, determine the environmental monitoring range according to the casualty position data, and collect the environmental parameter data through the environmental sensor within the range;

[0056] The second unit is used to construct a body temperature change trend graph according to the casualty body temperature distribution data, calculate the body temperature abnormality degree value, and at the same time calculate the environmental threat degree value according to the environmental parameter data, calculate the regional danger level based on the body temperature abnormality degree value and the environmental threat degree value, count the casualty distribution density in combination with the casualty position data, and generate a scene situation assessment result according to the regional danger level and the casualty distribution density;

[0057] The third unit is used to divide the rescue area according to the casualty distribution density based on the scene situation assessment result, and determine the resource allocation weight of each rescue area based on the regional danger level to generate a medical resource scheduling plan;

[0058] The fourth unit is used to construct a three-dimensional model of the emergency scene in the digital twin platform, perform visual mapping on the scene situation assessment result and the medical resource scheduling plan, overlay the rescue situation annotations in real time through augmented reality technology, and update the position information of the rescue personnel and the position information of the medical device in real time based on the target tracking algorithm;

[0059] The fifth unit is used to dynamically adjust the medical resource scheduling plan through a recursive optimization algorithm according to the rescuers' location information and medical equipment location information, so as to achieve real-time optimal allocation of medical resources.

[0060] In the third aspect of the embodiments of the present invention,

[0061] a kind of electronic device is provided, including:

[0062] a processor;

[0063] a memory for storing instructions executable by the processor;

[0064] wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0065] In the fourth aspect of the embodiments of the present invention,

[0066] a computer-readable storage medium is provided, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0067] In this embodiment, thermal imaging data and environmental parameter data of the accident scene are collected through infrared sensors and environmental sensors, and combined with the body temperature distribution and location information of the wounded, a comprehensive situation awareness of the accident scene is realized. This method can quickly and accurately evaluate the on-site situation and provide reliable data support for emergency decision-making. Based on the scenario situation assessment results, an intelligent algorithm is used to divide the rescue area and allocate resources, generating a scientific and reasonable medical resource scheduling plan. This method can effectively improve the utilization efficiency of medical resources and ensure the orderly progress of rescue work. By using digital twin and augmented reality technologies, three-dimensional visualization and real-time situation update of the accident scene are realized. At the same time, the medical resource scheduling plan is dynamically adjusted through a recursive optimization algorithm, realizing real-time optimal allocation of medical resources. This method greatly improves the scientificity and efficiency of emergency command and helps to enhance the emergency rescue ability of accidents. Description of the Drawings

[0068] Figure 1 It is a schematic flow chart of the emergency medical command method for accidents based on AI and the Internet of Things in the embodiments of the present invention;

[0069] Figure 2 It is a comparative analysis chart of key indicators in the embodiments of the present invention;

[0070] Figure 3 It is a simulation analysis chart of the path safety factor changing with time in the embodiments of the present invention. Detailed Embodiments

[0071] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part rather than all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0072] The technical solutions of the present invention will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0073] Figure 1 The figure is a schematic flowchart of an emergency medical command method based on AI and the Internet of Things in an embodiment of the present invention. As Figure 1 shown, the method includes:

[0074] Collect thermal imaging data at the emergency site through an infrared sensor, extract casualty body temperature distribution data and casualty location data from the thermal imaging data, determine the environmental monitoring range according to the casualty location data, and collect environmental parameter data within the range through an environmental sensor;

[0075] Construct a body temperature change trend graph based on the casualty body temperature distribution data, calculate the body temperature abnormality degree value, and at the same time calculate the environmental threat degree value according to the environmental parameter data. Calculate the regional risk level based on the body temperature abnormality degree value and the environmental threat degree value, count the casualty distribution density in combination with the casualty location data, and generate a scene situation assessment result according to the regional risk level and the casualty distribution density;

[0076] Based on the scene situation assessment result, divide the rescue area according to the casualty distribution density, and determine the resource allocation weight of each rescue area based on the regional risk level to generate a medical resource scheduling plan;

[0077] Construct a three-dimensional model of the emergency site in the digital twin platform, perform visual mapping on the scene situation assessment result and the medical resource scheduling plan, overlay rescue situation annotations in real time through augmented reality technology, and update the location information of rescue personnel and medical equipment in real time based on the target tracking algorithm;

[0078] According to the location information of rescue personnel and medical equipment, dynamically adjust the medical resource scheduling plan through a recursive optimization algorithm to achieve real-time optimal allocation of medical resources.

[0079] For example, an environmental monitoring range with a radius of 50 meters is set centered on the location of the wounded. An environmental sensor network is deployed within this range. The environmental sensors include temperature sensors, humidity sensors, toxic gas sensors, dust concentration sensors, etc., and the sampling frequency is 1 Hz. The measurement range of the temperature sensor is -40°C to 85°C, and the accuracy is ±0.5°C; the measurement range of the humidity sensor is 0 - 100%RH, and the accuracy is ±3%RH; the toxic gas sensor detects the concentration of gases such as CO and SO2; the dust sensor detects the concentration of PM2.5 and PM10.

[0080] In an alternative embodiment,

[0081] Thermal imaging data of the accident scene is collected by an infrared sensor, and the body temperature distribution data and the location data of the wounded are extracted from the thermal imaging data, including:

[0082] Thermal imaging data of the accident scene is collected by an infrared sensor, and the thermal imaging data is subjected to adaptive histogram equalization enhancement processing to obtain enhanced thermal imaging data. Based on the local temperature variance of the enhanced thermal imaging data, the size of the filtering window is determined, and the adaptive median filtering algorithm is used to denoise the enhanced thermal imaging data to obtain denoised thermal imaging data;

[0083] A mapping function between the gray value and the actual temperature is established according to the temperature response characteristics of the infrared sensor, and the gray values in the denoised thermal imaging data are converted into temperature values to generate a temperature distribution map. Based on the normal body temperature range of the human body, a double-threshold detection interval is set, and the pixel points in the temperature distribution map whose temperature values fall within the double-threshold detection interval are subjected to regional connectivity analysis to obtain candidate wounded areas;

[0084] Morphological processing is performed on the candidate wounded areas, the regional contour features, area features, and aspect ratio features are extracted, and the areas that do not conform to the features are screened out according to the human body size feature template to obtain confirmed wounded areas;

[0085] A temperature field distribution model of the confirmed wounded area is established, and the temperature numerical sequence, temperature gradient distribution, and temperature uniformity features within the area are extracted as the body temperature distribution data of the wounded; the minimum circumscribed rectangle is calculated according to the contour features of the confirmed wounded area, and the coordinates of the center point of the rectangle are used as the location data of the wounded.

[0086] Exemplarily, an infrared sensor array is deployed at the scene of an emergency for thermal imaging acquisition. The infrared sensor can adopt a non-cooled vanadium oxide detector with a wavelength response range of 8 - 14 microns, a temperature resolution of 0.05 °C, and an image resolution of 640 × 480 pixels. The sensor array consists of 9 sensor units arranged in a 3×3 matrix, with an overlapping rate of 20% for the fields of view of adjacent sensors, enabling a large-range scan of 120 degrees × 90 degrees. The acquisition frequency is set to 30 Hz, and each frame of thermal imaging data contains the scene temperature grayscale information.

[0087] First, histogram equalization enhancement processing is performed on the acquired thermal imaging data. The gray level intervals are adaptively divided according to the image gray level distribution characteristics, and key enhancement is performed on the low-contrast regions. In practical applications, the gray level range of the thermal imaging data is divided into 256 levels, the pixel cumulative distribution of each gray level is calculated to generate a cumulative distribution function, and the original gray level values are mapped to a new gray space through this function to achieve the adaptive enhancement of the image contrast.

[0088] Adaptive median filtering denoising is performed after image enhancement. First, the local temperature variance of the thermal imaging data is calculated. When the local area temperature variance is greater than 2 °C, it indicates that the area contains more detailed information, and a 3×3 small window is used for filtering; when the local temperature variance is less than 0.5 °C, it indicates that the area is relatively flat, and a 7×7 large window is used for filtering. During the filtering process, the median pixel within the window is retained, effectively removing the salt-and-pepper noise and random noise in the thermal imaging data.

[0089] Based on the temperature response characteristics of the infrared sensor, a conversion relationship from gray level to actual temperature is established. By acquiring the thermal imaging data of a standard black body at different ambient temperatures, a set of corresponding relationship points between gray level and temperature is obtained. A piecewise linear interpolation method is used to establish a mapping function to convert the gray level values in the denoised thermal imaging data into specific temperature values. For example, the gray level 0 corresponds to the ambient temperature of 20 °C, the gray level 255 corresponds to the highest temperature of 50 °C, and the intermediate temperature values are obtained through linear interpolation.

[0090] A double-threshold detection interval is set for casualty identification. Considering the normal body temperature range of the human body and measurement errors, the detection lower limit is set to 35 °C and the upper limit is set to 40 °C. The temperature distribution map is traversed to mark all the pixel points that fall within this temperature interval. An 8-neighborhood connectivity analysis method is used for region growing of the marked pixel points. When the temperature difference between adjacent pixel points is less than 0.5 °C, they are determined to be in the same region, and finally, multiple candidate casualty regions are obtained.

[0091] Perform morphological processing on the candidate casualty area, including opening operation to remove small-area noise regions and closing operation to fill holes within the region. Extract the contour features of the region, calculate the perimeter and area of the contour; extract the aspect ratio feature, and calculate the aspect ratio of the minimum bounding rectangle of the region. Set the screening conditions according to the human body size feature template: the area of the region should be within the range of 0.3 - 2 square meters, and the aspect ratio should be between 1.5 - 4. Regions that do not meet the conditions are excluded to obtain the confirmed casualty area.

[0092] For each confirmed casualty area, establish a temperature field distribution model. First, extract the temperature numerical sequence within the region, including statistical features such as the highest temperature, lowest temperature, and average temperature. Then calculate the temperature gradient distribution. The Sobel operator can be used to calculate the temperature gradients in the x and y directions, and the gradient magnitude and direction are synthesized. Finally, calculate the temperature uniformity feature, which can be represented by the standard deviation or entropy value of the temperature within the region. These features together constitute the casualty body temperature distribution data.

[0093] Finally, calculate the minimum bounding rectangle according to the contour features of the confirmed casualty area. The calculation of the minimum bounding rectangle can be realized by using the rotational Kalman filtering algorithm. The image coordinates of the center point of the rectangle are used as the casualty position data. If necessary, the image coordinates can also be converted into actual geographical coordinates for subsequent rescue positioning.

[0094] In this embodiment, it can effectively process the thermal imaging data at the emergency scene, accurately extract the body temperature distribution and position information of the casualties, and provide an important basis for emergency rescue. This method adopts an image preprocessing technology that combines adaptive histogram equalization and adaptive median filtering, effectively improving the quality and reliability of the thermal imaging data. By establishing an accurate mapping relationship between the gray value and the actual temperature, and combining the double-threshold detection method of human body temperature characteristics, potential casualty areas can be accurately identified. The method that combines morphological processing and the human body size feature template effectively removes the misdetected areas and improves the accuracy of casualty identification. It can not only obtain the position information of the casualties, but also extract detailed body temperature distribution features, including the temperature numerical sequence, temperature gradient distribution, and temperature uniformity, etc. These information are of great value for evaluating the condition of the casualties and judging the severity of the injuries, and can provide more comprehensive decision-making support for rescue personnel.

[0095] In an alternative embodiment,

[0096] Construct a body temperature change trend graph based on the casualty body temperature distribution data, calculate the body temperature abnormality degree value, and at the same time calculate the environmental threat degree value according to the environmental parameter data. Calculate the regional risk level based on the body temperature abnormality degree value and the environmental threat degree value, and combine the casualty position data to statistically calculate the casualty distribution density. Generate the scene situation assessment result according to the regional risk level and the casualty distribution density, including:

[0097] A body temperature monitoring window is established, and the body temperature monitoring window is divided into an observation interval and a prediction interval. The body temperature data in the observation interval is segmented and fitted to extract the fluctuation law. A body temperature baseline is established based on the prediction interval, and the degree of deviation of the fluctuation law from the body temperature baseline is calculated to obtain the fluctuation anomaly feature. The body temperature anomaly degree value is judged according to the fluctuation anomaly feature;

[0098] An influence network including environmental parameters is constructed according to the environmental parameter data. The leading abnormal parameter is used as the root node, and the affected parameter is used as the child node. A directed graph structure of the transfer relationship between parameters is established. Based on the directed graph structure, the abnormal diffusion path is analyzed, and the environmental threat degree value is calculated by combining the range and speed of abnormal diffusion;

[0099] According to the change trends of the body temperature anomaly degree value and the environmental threat degree value, the inference rules of the neural network are dynamically updated, and the regional danger level is calculated by using the inference rules;

[0100] The spatial distribution of the casualty location data is analyzed, a casualty movement trajectory map is established, the trajectory intersection points are extracted as key nodes, and the degree of personnel aggregation at each key node is calculated to obtain the casualty distribution density;

[0101] Using the regional danger level as the weight coefficient, the casualty distribution density is weighted and calculated to obtain the scene situation assessment result.

[0102] Exemplarily, by monitoring the body temperature of the casualty in real time, a body temperature monitoring window is established. The time range of the body temperature monitoring window is divided into two parts. One part is the observation interval, which is used to record the body temperature data in the past period of time, and the other part is the prediction interval, which is used to predict the body temperature change trend in the future period of time. The body temperature data in the observation interval will be divided into several time periods, and the body temperature data in each time period will be curve-fitted to extract the change law of body temperature fluctuation. This fitting can adopt multi-segment linear fitting or smooth curve fitting methods, and the purpose is to restore the basic trend and fluctuation amplitude of body temperature change over time. Then, in the prediction interval, based on the body temperature change law obtained in the observation interval, a body temperature baseline is established as a reference value for normal body temperature. The actual body temperature data in the prediction interval is compared with the baseline to judge whether the body temperature fluctuation deviates from the normal range. The greater the degree of deviation of the body temperature fluctuation, the higher the possibility of abnormal body temperature of the casualty. By setting different deviation thresholds, a body temperature anomaly degree value is finally obtained to quantify the severity of body temperature anomaly.

[0103] Next, collect environmental parameter data. Environmental parameters include, but are not limited to, temperature, humidity, air pressure, harmful gas concentration, light intensity, etc. Among these environmental parameters, first determine which parameters belong to leading abnormal parameters, such as sudden increase in harmful gas concentration, abnormal change in temperature, etc. These parameters will serve as the root nodes affecting the network. Then, based on the mutual influence relationship between environmental parameters, establish a directed graph structure that includes all relevant parameters. This directed graph reflects the transmission path between parameters. For example, an increase in temperature will cause a decrease in humidity or fluctuations in air pressure. According to the influence intensity and transmission order between parameters, analyze how anomalies spread between environmental parameters, and identify the paths, speeds, and scopes of anomaly diffusion. Considering these factors comprehensively, calculate the environmental threat degree value to measure the threat degree of the current environmental anomaly to regional security.

[0104] Use the body temperature anomaly degree value and the environmental threat degree value as inputs to dynamically update the rule base of an inference system based on a neural network. This neural network has been trained to infer the corresponding regional danger level according to different input values. During the inference process, the neural network will consider the change trends of body temperature anomalies and environmental threats, perform weighted processing on the input data, and output the corresponding regional danger level. The regional danger level is divided into multiple levels, from low risk to high risk, and each level represents the current comprehensive danger level of the region.

[0105] Conduct a spatial distribution analysis of the location data of the wounded. Specifically, collect the current and past location coordinates of the wounded, and use this data to draw a movement trajectory map of the wounded. By analyzing the intersection points between the trajectories of each wounded person, extract the key location nodes where the wounded may gather. Calculate the degree of personnel aggregation for each key node. Specifically, the distribution density of the wounded can be obtained by counting the number of wounded persons who have appeared at this location and their residence time.

[0106] Use the obtained regional danger level as the weight coefficient and perform weighted calculation with the wounded distribution density. The higher the regional danger level, the greater the risk of the wounded gathering in this region. Through this weighted method, generate the situation assessment result of the current scene. The situation assessment result can comprehensively reflect the current body temperature anomalies of the wounded, environmental threats, the distribution of the wounded, and the comprehensive danger level, providing a basis for rescue decision-making and resource allocation.

[0107] In this embodiment, by establishing a body temperature monitoring window and performing piecewise fitting analysis, the changing trend of body temperature can be accurately captured, potential abnormal situations can be detected in advance, and the precise assessment and early warning of the health status of the wounded can be realized. Based on the analysis method of environmental parameter influence network, the correlation and conduction mechanism between environmental factors can be effectively identified, the potential threat of environmental changes to the wounded can be accurately evaluated, and a more comprehensive basis for environmental risk prevention and control can be provided. By adopting a dynamically updated deep neural network combined with the analysis of the distribution density of the wounded, the precise assessment of the overall situation of the area is realized, which can provide a scientific basis for the allocation of medical resources and emergency response, and improve the efficiency of emergency rescue.

[0108] In an alternative embodiment,

[0109] Perform a spatial distribution analysis on the location data of the wounded, establish a moving trajectory map of the wounded, extract the trajectory intersection points as key nodes, and calculate the degree of personnel aggregation at each key node to obtain the distribution density of the wounded, including:

[0110] Perform segmented processing on the location data of the wounded according to the time sequence to obtain a time window sequence, calculate the change of the position coordinates within the time window sequence to obtain a displacement vector, connect the displacement vectors in time sequence to construct a moving trajectory map of the wounded, perform smoothing processing on the moving trajectory map of the wounded to eliminate abnormal jumps, and extract the target moving path according to the smoothed trajectory;

[0111] Draw the target moving path in the spatial coordinate system, calculate the intersection angle and the shortest distance between the paths, determine the trajectory intersection points when the intersection angle and the shortest distance meet the preset conditions, count the intersection frequencies at each trajectory intersection point, calculate the number of personnel within the unit area around the trajectory intersection points to obtain the degree of personnel aggregation, and determine the key nodes according to the weighted values of the intersection frequencies and the degree of personnel aggregation;

[0112] Establish a local coordinate system with the key nodes as the center, map the intersection frequencies and the degree of personnel aggregation to the size of the influence range and the attenuation rate respectively, construct a node influence field according to the influence range and the attenuation rate, and superimpose multiple node influence fields to generate a distribution influence map;

[0113] Convert the real-time change characteristics of the moving trajectory map of the wounded into dynamic weights, and perform dynamic weighting on the distribution influence map to obtain a density distribution model;

[0114] Divide the target area into grid cells, calculate the distance between the center of each grid cell and the key nodes, calculate the reference density value within each grid cell according to the distance and the density distribution model, and superimpose the time-varying characteristics of the dynamic weights on the reference density value to obtain the distribution density of the wounded reflecting the dynamic distribution of personnel.

[0115] Exemplarily, the real-time position data of all the wounded are collected and segmented according to the time sequence, divided into a sequence of continuous time windows. Within each time window, the position information of the wounded in that time period is extracted, and the change in position coordinates between adjacent time points is calculated to obtain the corresponding displacement vector. These displacement vectors are sequentially connected in time order to form a complete movement trajectory graph of each wounded throughout the time period. To avoid abnormal jumps caused by data noise or positioning errors in a short period, trajectory smoothing processing techniques are adopted, such as using methods like moving average or Kalman filtering to correct the mutation points in the trajectory, thereby ensuring the continuity and smoothness of the trajectory curve, and then extracting the target movement path representing the real movement trend.

[0116] The target movement paths of all the wounded are plotted within a unified spatial coordinate system. All the paths are compared pairwise, and the crossing angle between each pair of paths and the shortest distance between them are calculated. When the crossing angle between a certain pair of paths is within a small range and the shortest distance between the paths is lower than the set threshold condition, it is determined that there is a crossing point in space between these two paths, which is marked as a trajectory crossing point. The frequency of occurrence of all the trajectory crossing points is counted, that is, the number of times each crossing point is passed by the paths of different wounded. At the same time, a unit area is demarcated around each trajectory crossing point, and the number of people present in this area is counted to obtain the degree of personnel aggregation around the crossing point. Then, the crossing frequency of each crossing point and the degree of personnel aggregation are weighted and calculated according to the preset weight coefficient, and the positions with a high degree of personnel aggregation and a high degree of path crossing are screened out and determined as key nodes.

[0117] Suppose a crossing point is found at the coordinates (50, 50), and 5 trajectories pass through this point, with a crossing frequency of 5. 20 people are counted within a radius of 10 meters from this point, then the degree of personnel aggregation is 0.064 people per square meter. If the weight of the crossing frequency is set to 0.6 and the weight of the degree of personnel aggregation is set to 0.4, then the weighted value of this point is 5×0.6 + 0.064×0.4 = 3.0256. A threshold can be set, such as 2.5. When the weighted value exceeds this threshold, this crossing point is determined as a key node.

[0118] At the position of each key node, a local coordinate system is established with this node as the center. The previously calculated crossing frequency is mapped to the size of the influence range, that is, the higher the crossing frequency, the larger the influence range; at the same time, the degree of personnel aggregation is mapped to the attenuation rate, that is, the higher the degree of aggregation, the slower the influence decays within the influence range. Based on the influence range and the attenuation rate, an influence field of each key node is constructed, which is manifested as a spatial influence model that spreads outward from the center and gradually weakens with the increase of distance. The influence fields of all the key nodes are superimposed to obtain an overall distribution influence map, which reflects the comprehensive intensity of the influence of personnel aggregation and trajectory crossing at each position in the entire region.

[0119] Extract the characteristics of the change of the movement trajectory map of the wounded over time, calculate parameters such as the speed of trajectory change and the frequency of direction change, and convert these change characteristics into dynamic weight values. Using these dynamic weights, perform dynamic weighting on the previously generated distribution influence map, so that the distribution influence map not only reflects the static aggregation degree, but also reflects the time-varying characteristics of the personnel distribution, forming a density distribution model.

[0120] Divide the target area into several grid cells, and each grid cell has a center point. Calculate the distance from the center point of each grid cell to all key nodes, and calculate the reference density value within each grid cell according to the distance and the density distribution model. The closer the unit is to the key node, the higher its reference density value. Finally, superimpose the time change characteristics of the dynamic weight on the reference density value to obtain the wounded distribution density result that can reflect the dynamic distribution change of personnel in real time.

[0121] Figure 2 This is the comparative analysis chart of the key indicators of the embodiment of the present invention. As Figure 2 shown, in terms of positioning accuracy, this method achieves a high accuracy of 1.25 meters, an improvement of 66.8% compared with the traditional density heat map. This is mainly due to the algorithm optimization of trajectory smoothing to eliminate abnormal jumps. In terms of the accuracy of personnel density prediction, this method achieves an accuracy of 92.7%, far ahead of the 78.3% of the traditional method, verifying the effectiveness of the key node determination mechanism based on cross-frequency and personnel aggregation degree weighting. The dynamic change response time is only 1.8 seconds, more than 3 times faster than the traditional technology, reflecting the advantage of the proposed density distribution model in capturing time-varying characteristics. The key node recognition rate and the abnormal trajectory detection rate reach 94.2% and 89.5% respectively, both far superior to the existing technology, indicating that this method can better identify trajectory intersection points and screen key nodes. The calculation efficiency index is 4.2 ms / sample, more than twice as fast as the GPS-based trajectory analysis, demonstrating the high efficiency after algorithm optimization. These data fully prove the significant advantages of the proposed wounded trajectory analysis and density modeling method in various key indicators.

[0122] The prior art usually analyzes the distribution of personnel only through static location data, ignoring the dynamic changes and interaction characteristics of the personnel movement trajectories, resulting in inaccurate assessment of the personnel aggregation situation. Especially in complex disaster environments, it is impossible to effectively identify high-risk aggregation areas. In this application, by introducing a time window to segment the location data, a continuous movement trajectory map is constructed, and smoothing processing is used to eliminate noise and extract the real movement path. At the same time, by calculating the path intersection points, their intersection frequencies, and the degree of personnel aggregation, key nodes are identified, and combined with the establishment of a local influence field, an accurate description of the dynamic personnel aggregation situation in the space is achieved. In addition, a dynamic weight mechanism is introduced to map the trajectory change characteristics to the density distribution model to further reflect the personnel flow characteristics. The starting point of the improvement is to solve the limitations of static and single analysis in the prior art and achieve dynamic and spatio-temporal combined distribution density assessment. Finally, the recognition accuracy of crowded areas is improved, providing more reliable data support for subsequent danger warning and rescue decision-making.

[0123] In an alternative embodiment,

[0124] Based on the results of the scenario situation assessment, the rescue area is divided according to the distribution density of the wounded, and the resource allocation weight of each rescue area is determined based on the regional danger level. The generated medical resource scheduling plan includes:

[0125] Receive the results of the scenario situation assessment and the distribution density data of the wounded. Based on the cellular automaton, a prediction model is constructed to analyze the change trend of the scenario situation, and the scenario threat degree and passable paths of each area at future moments are obtained;

[0126] Calculate the density difference between adjacent areas according to the distribution density data of the wounded. Mark the areas where the density difference is less than the density threshold and the spatial distance is less than the distance threshold as the areas to be merged. Combine the scenario threat degree to optimize the boundaries of the areas to be merged and generate rescue sub-areas;

[0127] Based on the corresponding relationship between the medical resource input amount and the wounded treatment effect in the historical rescue data, calculate the resource utilization efficiency of each rescue sub-area, and calculate the regional danger level according to the distribution density, scenario threat degree, and resource utilization efficiency in the rescue sub-area;

[0128] Construct a resource transfer network with the rescue sub-areas as nodes and the passable paths as edges. Perform weighted calculation on the regional danger level and resource utilization efficiency to obtain the resource allocation weight, and generate a medical resource scheduling plan with the minimum transfer loss based on the resource transfer network;

[0129] Monitor the change of the scenario situation assessment results. When the change of the scenario threat degree exceeds the situation threshold, re-divide the rescue sub-areas based on the updated scenario threat degree and the distribution density of the wounded, adjust the regional danger level and resource allocation weight, and update the medical resource scheduling plan.

[0130] Exemplarily, the scene situation assessment results and casualty distribution density data are received, and the entire rescue area is divided into several grid cells. Each grid cell represents a small geographical area and contains corresponding casualty density and danger level information.

[0131] The cellular automaton model is used to simulate these grid cells. The cellular automaton is a discrete model that simulates the dynamic evolution of complex systems by defining simple local rules. An initial state is set for each grid cell, including the current danger level and personnel density. According to the influence rules of neighboring cells, the state of each cell is iteratively updated to predict the change trend of the scene threat degree and the passable path in each area at future moments.

[0132] Suppose a rescue area is divided into a 10x10 grid, and the initial threat degree of each grid cell is between 0 and 10. The update rule is set as follows: if the average threat degree of the surrounding 8 adjacent cells is greater than the current cell, the threat degree of the current cell increases by 1; otherwise, it decreases by 1. After 100 iterations, the future threat degree distribution and passable path of this area can be obtained.

[0133] After obtaining the personnel density and predicted scene threat degree of each cell, the casualty density data of adjacent grid cells are compared, and the density difference between them is calculated. If the density difference between two cells is lower than the preset density threshold and their spatial distance is less than the distance threshold, these two cells are marked as areas to be merged. For all areas to be merged that meet the conditions, based on considering the predicted scene threat degree, by adjusting the boundary, the areas with lower local threat and similar density are merged to obtain the final rescue partition.

[0134] Then, based on the corresponding relationship between the amount of medical resources invested and the casualty treatment effect in historical rescue data, a resource efficiency model is established to evaluate the actual treatment efficiency after medical resources are invested in each rescue partition. Considering comprehensively the current casualty distribution density, predicted scene threat degree, and resource utilization efficiency of each rescue partition, the regional danger level of this partition is calculated. The weighted average method can be used, assuming the weights are 0.4, 0.4, and 0.2 respectively. For example, if the casualty density in a partition is 80 people per square kilometer (normalized to 8 points), the threat degree is 7 points, and the resource utilization efficiency is 6 points, then its danger level is 7.2 points.

[0135] Taking the generated rescue zones as nodes and the predicted passable paths as edges, a resource transfer network graph is constructed to reflect the feasible paths and distance relationships of resource scheduling between zones. On this basis, the regional hazard levels and resource utilization efficiencies of each zone are weighted to determine the resource allocation weights for each zone, ensuring that resources are tilted towards high-hazard and high-demand areas. Using the shortest path algorithm, combined with the resource allocation weights and path loss conditions, a medical resource scheduling plan with the minimum resource transfer loss is generated to ensure the efficient transfer of medical resources while meeting the needs.

[0136] Real-time monitor the results of the scenario situation assessment. When it is detected that the scenario threat level in a certain area has changed significantly and exceeds the preset situation change threshold, based on the updated casualty distribution density and threat level data, re-divide the rescue zones, re-adjust the hazard levels and resource allocation weights of each zone, and dynamically correct the medical resource scheduling plan to ensure the timeliness and rationality of resource allocation.

[0137] In this embodiment, by introducing a cellular automaton prediction model, the threat trends and passable paths of each area in the future can be accurately analyzed, improving the foresight of rescue operations. Dynamically optimize the area division based on the casualty distribution density and scenario threat level to ensure reasonable adjustment of the rescue area boundary, reduce duplicate resource investment and scheduling blind spots. Combine historical rescue data to evaluate the resource utilization efficiency and hazard level of each area in real time, rationally allocate limited medical resources, and improve the accuracy and utilization rate of resource allocation. After constructing the resource transfer network, by dynamically adjusting the resource allocation weights, the loss in the resource scheduling process can be effectively reduced, realizing the rapid and efficient transfer of resources. The overall technical means can respond to on-site threat changes in real time, ensure that medical resources always accurately and efficiently cover key areas, and significantly improve the overall rescue efficiency.

[0138] In an alternative embodiment,

[0139] Based on the cellular automaton, a prediction model is constructed to analyze the change trend of the scenario situation, and the scenario threat level and passable path of each area at future moments are obtained, including:

[0140] The rescue scenario is divided into regular grid cells, and the road traffic status, building damage degree, hazard source intensity, and environmental threat level in the scenario situation data are mapped to each grid cell to form an initial state matrix of the grid cells;

[0141] Calculate the evolution value of the road traffic status according to the passing frequency of rescue vehicles, calculate the evolution value of the building damage degree according to the secondary disaster monitoring data, calculate the diffusion value of the hazard source intensity and environmental threat level between adjacent grid cells, and construct a state transition equation with the evolution value of the road traffic status, the evolution value of the building damage degree, the hazard source intensity, and the diffusion value of the environmental threat level;

[0142] The state transition equation is solved by iterative calculation to obtain the state matrix of each grid cell at the prediction time, and the scene threat degree at the prediction time is calculated based on the state matrix;

[0143] The passing probability is calculated based on the road passing state and the scene threat degree of the grid cell, a path cost function is constructed, and a passable path is obtained by using a search algorithm;

[0144] The scene situation data is received in real time. When the data is updated, the updated data is mapped to a new initial state matrix, and the prediction calculation is re-executed to dynamically update the scene threat degree and the passable path.

[0145] Exemplarily, the rescue scene is divided into m×n regular grid cells. The size of each grid cell can be set according to actual needs, for example, 10m×10m. For each grid cell, four attribute values of its road passing state, building damage degree, hazard source intensity, and environmental threat degree are respectively recorded. The road passing state is represented by a value between 0 and 1, where 0 means completely impassable and 1 means completely unobstructed. The building damage degree is represented by 0-100%, where 0% means intact and 100% means completely damaged. Both the hazard source intensity and the environmental threat degree are represented by values between 0 and 10, where 0 means no threat and 10 means the highest threat level.

[0146] Next, the scene situation data is mapped into the grid cells. For example, if the passing state of a certain road is 0.8, the road passing state of the grid cell where it is located is set to 0.8. If the damage degree of a certain building is 60%, the building damage degree of the grid cell where it is located is set to 60%. If the hazard source intensity at a certain place is 6, the hazard source intensity of the grid cell where it is located is set to 6. The environmental threat degree data is mapped in the same way. In this way, the initial state matrix of the grid cells is formed.

[0147] Then, a state transition equation is constructed. First, the evolution value of the road passing state is calculated. The vehicle passing frequency of each grid cell within a certain time is counted. The higher the frequency, the smoother the road. For example, if 100 vehicles pass through a certain grid cell within 1 hour, the evolution value of its road passing state can be set to 0.9. For the building damage degree, its evolution value is calculated according to the monitored secondary disaster data (such as aftershock intensity). For example, if the aftershock intensity is monitored to be 4, the evolution value of the building damage degree can be set to 10%. For the hazard source intensity and the environmental threat degree, their diffusion values to adjacent grid cells are calculated. The diffusion value is inversely proportional to the distance and directly proportional to the initial intensity. For example, if the hazard source intensity of a certain grid cell is 8, the diffusion value of the hazard source intensity of its adjacent grid cell can be set to 4.

[0148] Using the above evolution value and diffusion value, construct a state transition equation. For each grid cell, its state at time t+1 is equal to the state at time t plus the evolution value and the diffusion value. For example, if the road traffic state of a certain grid cell at time t is 0.7 and the evolution value is 0.1, then the road traffic state at time t+1 is 0.8. The calculation methods for the building damage degree, the hazard source intensity, and the environmental threat degree are similar.

[0149] Adopt an iterative calculation method to solve the state transition equation. Set the prediction time length, such as predicting the situation in the next 2 hours. Discretize the time, for example, each 5 minutes is a time step, then 24 iterations are required. Update the state of all grid cells once for each iteration. After the iteration is completed, obtain the state matrix at the prediction time.

[0150] Based on the predicted state matrix, calculate the scenario threat degree. The scenario threat degree is comprehensively determined by three factors: the building damage degree, the hazard source intensity, and the environmental threat degree. The weighted average method can be used, such as the scenario threat degree = 0.3×building damage degree + 0.4×hazard source intensity + 0.3×environmental threat degree. Calculate the threat degree for each grid cell and draw a heat map to visually display the threat levels of each area.

[0151] Next, calculate the passable paths. First, calculate the passing probability based on the road traffic state and the scenario threat degree of the grid cells. The passing probability is proportional to the road traffic state and inversely proportional to the scenario threat degree. For example, if the road traffic state of a certain grid cell is 0.8 and the scenario threat degree is 3, then its passing probability can be set to 0.8×(1 - 3 / 10) = 0.56. Then construct a path cost function, where the cost is inversely proportional to the passing probability and proportional to the path length. Use search algorithms such as A* to solve for the passable paths with the goal of minimizing the total cost.

[0152] Finally, receive the updated scenario situation data in real time. When new data is received, map it to a new initial state matrix and re-execute the prediction calculation process. This can dynamically update the scenario threat degree and the passable paths to ensure the timeliness of decision-making.

[0153] Figure 3 This is the simulation analysis graph of the path safety factor changing with time in the embodiment of the present invention. As Figure 3 shown, this graph shows the comparative analysis of the rescue path planning technology based on cellular automata and traditional methods in terms of the safety factor changing with time. Through Monte Carlo simulation (10,000 iterations), the graph intuitively presents the time series change trends of the safety factors of the three methods: the technical method of the present invention (circular markers, solid line), static path planning (triangle markers, dashed line), and the GIS-based method (square markers, dotted line).

[0154] The results show that during the dynamic change process of the disaster environment, the safety factor of this technical method always remains at a relatively high level (with an average of 0.85). Even at the critical moments when secondary disasters occur at 30 minutes and the environmental threat increases at 60 minutes, its safety factor still remains above 0.8, significantly higher than the safety threshold (0.5). In contrast, the safety factor of the static path planning method drops sharply and drops to 0.3 at 120 minutes, far lower than the safety threshold; although the GIS-based method is better than the static planning, it also shows an obvious downward trend in the later stage of the disaster.

[0155] Statistical analysis shows that this technical method not only has the highest average safety factor, but also has the smallest standard deviation (±0.05), indicating its stable dynamic adaptation ability. It can adjust the path planning strategy in a timely manner according to the changes in the disaster environment, effectively improving the safety and reliability of rescue operations. This result fully verifies the significant advantages of the prediction model based on cellular automata in the application of complex disaster scenarios.

[0156] In this embodiment, by dividing the scene into grid cells and mapping key data, the state of each area can be accurately captured, and then the changing trends of factors such as road traffic capacity, building damage, spread of hazard sources, and environmental threat level can be dynamically evaluated. This prediction model based on cellular automata can automatically adjust the scene threat level and traffic path through iterative calculation after real-time updating of the scene data. Existing technologies usually rely on static models, which are difficult to quickly adapt to the changes in the post-disaster scene, and the path planning often cannot reflect the changes in road conditions and threat levels in real time. In existing technologies, path planning mostly relies on simple cost functions and does not consider the impact of the dynamic environment on the path, resulting in unsatisfactory effects in complex disaster scenarios. In contrast, this solution adjusts the prediction results in real time through a dynamic iterative method and can still provide accurate threat assessment and optimal path selection in the case of continuous evolution of the disaster. The improvement of this application lies in combining the cellular automata model and the dynamic data update mechanism to achieve continuous prediction and timely adjustment of the scene state. Through the state transition equation and the environmental diffusion model, the evolution of various factors after the disaster can be accurately simulated, ensuring that the system can continuously give the optimal path and threat assessment when the data changes. The improvement of this technology makes the path planning more accurate and the decision-making support more timely during the rescue process, improving the rescue efficiency and safety.

[0157] In an alternative embodiment,

[0158] Build a three-dimensional model of the emergency site in the digital twin platform, visually map the scene situation assessment results and the medical resource scheduling plan, overlay the rescue situation annotations in real time through augmented reality technology, and update the position information of rescue personnel and medical equipment in real time based on the target tracking algorithm, including:

[0159] In the digital twin platform, on-vehicle lidar is used to obtain the point cloud data of the emergency site. The point cloud data is registered to obtain a dense point cloud. Based on the region growing algorithm, the dense point cloud is segmented for the scene structure to generate a three-dimensional model of the emergency site, and a unified spatial coordinate system is established in the three-dimensional model;

[0160] Based on the unified spatial coordinate system, the scene threat degree in the scene situation assessment result is mapped to the three-dimensional model to form a situation heat map. The equipment layout data and personnel deployment data in the medical resource scheduling plan are converted into position coordinates in the three-dimensional scene and a situation annotation is generated to obtain a digital twin scene with situation information;

[0161] The feature point matching algorithm is used to realize the real-time registration of the mobile device camera and the digital twin scene with situation information, establish the mapping relationship between the camera view and the unified spatial coordinate system, and the situation heat map and situation annotation are superimposed and displayed in the camera view through the augmented reality technology to obtain an augmented reality situation view;

[0162] Based on the target tracking algorithm, the position information of the medical device is obtained in real time, the position information collected by the rescue personnel positioning terminal is received, the position information is coordinate-transformed according to the unified spatial coordinate system, and the transformed position information is superimposed and displayed in the augmented reality situation view based on the mapping relationship;

[0163] When the scene situation assessment result is updated, a new situation heat map is generated; when the medical resource scheduling plan changes, the situation annotation is updated; when the position information of the rescue personnel and the medical device is updated, the position display in the augmented reality situation view is updated in real time based on the mapping relationship.

[0164] Exemplarily, first, the on-vehicle lidar is used to collect the on-site point cloud data. The Velodyne HDL-64E lidar is used to scan at a frequency of 10Hz to obtain high-precision three-dimensional point cloud data. The collected original point cloud data is preprocessed, including noise filtering and outlier removal. The iterative closest point algorithm is used to register multiple frames of point cloud data to generate a high-density point cloud. Based on the region growing algorithm, the point cloud is segmented for the scene, and the point cloud data is divided into different categories such as ground, buildings, and vegetation. The triangular mesh reconstruction method is used to generate the scene three-dimensional model, and a unified spatial coordinate system with the center of the accident site as the origin is established.

[0165] Map the scene situation assessment information into a 3D model. Based on deep learning methods, evaluate the threat level of the scene, and the evaluation result is represented by a numerical value from 0 to 1 indicating the degree of danger. Convert the threat level numerical value into a heat map through color mapping, where red represents high-risk areas and green represents safe areas. At the same time, convert the equipment layout and personnel deployment information in the medical resource scheduling plan into 3D coordinate points, and generate corresponding annotation symbols in the model. Different types of medical equipment and rescue personnel use different annotation styles.

[0166] The mobile device captures real-time images of the scene through the camera, extracts SIFT feature points, and matches them with the feature points pre-extracted in the 3D model. Estimate the camera pose based on the RANSAC algorithm and establish the mapping relationship between the camera view and the unified coordinate system. Overlay the situation heat map and annotation information into the camera view through the OpenGL ES rendering engine to achieve the augmented reality display effect.

[0167] Medical devices are installed with Bluetooth positioning tags to obtain real-time positions through the indoor positioning base station network. The positioning terminals carried by rescue personnel obtain position information based on the integrated positioning method of Beidou and inertial navigation. Uniformly convert this position data into the scene coordinate system and update the display in real time in the augmented reality view. When the position of the device or personnel changes, the system automatically updates its display position in the view.

[0168] The system re-evaluates the scene situation every 30 seconds and updates the heat map display. When the medical resource scheduling plan changes, immediately update the corresponding annotation information. The position information is updated at a frequency of 10Hz to ensure the real-time nature of the positions of various elements in the augmented reality view.

[0169] The existing technologies mainly rely on manual observation and simple digital map annotation for emergency rescue command, making it difficult to grasp the real-time changes in the scene situation and resource distribution accurately, and lacking intuitive information display means, which affects the timeliness and accuracy of command decisions. This application proposes a new visualization command method based on digital twin and augmented reality technologies. Collect on-site point cloud data through vehicle-mounted lidar and establish an accurate 3D model to achieve precise mapping of situation information under a unified spatial coordinate system. Adopt feature point matching and target tracking algorithms to overlay and display the scene threat level heat map, medical resource deployment annotations, and real-time position information of personnel and equipment in the mobile device view, establishing a dynamic association between the real scene and digital information. When the scene situation and resource deployment change, the system can automatically update the display content. This solution overcomes the defects of incomplete information acquisition, non-intuitive display, and untimely update in traditional methods, realizes the real-time perception, three-dimensional presentation, and dynamic update of the on-site situation of emergencies, provides commanders with a more intuitive and accurate situation awareness ability, and significantly improves the scientificity and efficiency of emergency rescue command.

[0170] Optionally, when conducting medical resource scheduling, first establish a resource demand model based on the results of scenario situation assessment. Determine the number and level of ambulances required in each area according to the casualty distribution density and the degree of abnormal body temperature. Among them, first-aid vehicles with rescue functions are preferentially allocated in critical areas.

[0171] For medical equipment, the system will intelligently match first-aid equipment such as ventilators, defibrillators, and monitors according to the injury characteristics reflected by the thermal imaging data, and reasonably plan the layout position of the equipment based on the equipment transfer radius. In terms of drug dispensing, combine the trend of body temperature change and environmental threat factors to predict the types of possible complications and reserve targeted drugs in advance.

[0172] For blood transfusion guarantee, the system analyzes the number of casualties and the degree of injury, estimates the types and quantities of blood products required, and formulates a deployment plan in combination with the inventory situation and transportation time of blood banks. In the scheduling of medical staff, based on the regional danger level and the distribution of casualties, calculate the allocation ratio of emergency doctors, nurses, and specialist doctors required, and at the same time consider the matching degree between the professional expertise of medical staff and the types of injuries to achieve precise allocation. The system will also plan the casualty transfer routes and distribution plans according to the bed capacity, specialty characteristics, and treatment capabilities of each hospital to ensure the efficient coordination of medical resources.

[0173] In another alternative embodiment, the following medical resources are allocated by intelligent means: medical transportation tools include ambulances of different levels, medical helicopters, and special transfer vehicles; first-aid equipment includes portable ventilators, electrocardiogram monitors, defibrillators, infusion pumps, portable B-ultrasounds, mobile DRs, etc.; medical consumables include tracheal intubation, infusion equipment, hemostatic bandages, disposable medical protective supplies, etc.; drug types include cardiopulmonary resuscitation drugs, hemostatic drugs, anti-shock drugs, anti-infection drugs, antidote drugs, infusion drugs, etc.; blood transfusion products include various types of red blood cells, plasma, platelets, etc.; medical staff include emergency department doctors, intensive care unit doctors, trauma surgeons, nurses, first-aid workers, etc.

[0174] The system analyzes historical case data through deep learning algorithms, establishes a medical resource allocation knowledge base, predicts the resource demand in real time according to the changes in the on-site situation, and generates an optimal allocation plan in combination with the resource reserve situation of surrounding medical institutions.

[0175] During the implementation of the plan, the location and usage status of medical resources are monitored in real time based on the Internet of Things technology. When there is an imbalance between resource supply and demand, the system will automatically trigger an alternative plan to ensure the continuous and orderly development of the rescue work.

[0176] In the second aspect of the embodiments of the present invention,

[0177] Provide an emergency medical command system based on AI and the Internet of Things, the system includes:

[0178] The first unit is used to collect thermal imaging data at the scene of an emergency through an infrared sensor, extract casualty body temperature distribution data and casualty location data from the thermal imaging data, determine the environmental monitoring range according to the casualty location data, and collect environmental parameter data through an environmental sensor within the range;

[0179] The second unit is used to construct a body temperature change trend graph based on the casualty body temperature distribution data, calculate the body temperature abnormality degree value, calculate the environmental threat degree value according to the environmental parameter data at the same time, calculate the regional danger level based on the body temperature abnormality degree value and the environmental threat degree value, count the casualty distribution density in combination with the casualty location data, and generate a scene situation assessment result according to the regional danger level and the casualty distribution density;

[0180] The third unit is used to divide the rescue area according to the casualty distribution density based on the scene situation assessment result, determine the resource allocation weight of each rescue area according to the regional danger level, and generate a medical resource scheduling plan;

[0181] The fourth unit is used to construct a three-dimensional model of the emergency scene in the digital twin platform, perform visual mapping on the scene situation assessment result and the medical resource scheduling plan, overlay rescue situation annotations in real time through augmented reality technology, and update the positions of rescue personnel and medical equipment in real time based on the target tracking algorithm;

[0182] The fifth unit is used to dynamically adjust the medical resource scheduling plan according to the positions of rescue personnel and medical equipment through a recursive optimization algorithm to achieve real-time optimal allocation of medical resources.

[0183] In the third aspect of the embodiments of the present invention,

[0184] A kind of electronic device is provided, including:

[0185] A processor;

[0186] A memory for storing instructions executable by the processor;

[0187] Wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0188] In the fourth aspect of the embodiments of the present invention,

[0189] A computer-readable storage medium is provided, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0190] The present invention may be a method, an apparatus, a system, and / or a computer program product. The computer program product may include a computer-readable storage medium having thereon computer-readable program instructions for performing various aspects of the present invention.

[0191] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An emergency medical command method based on AI and the Internet of Things, characterized in that, Including: Collecting thermal imaging data of the emergency scene through an infrared sensor, extracting casualty body temperature distribution data and casualty location data from the thermal imaging data, determining the environmental monitoring range based on the casualty location data, and collecting environmental parameter data within the range through an environmental sensor; Constructing a body temperature change trend graph according to the casualty body temperature distribution data, calculating the body temperature abnormality degree value, at the same time calculating the environmental threat degree value according to the environmental parameter data, calculating the regional danger level based on the body temperature abnormality degree value and the environmental threat degree value, counting the casualty distribution density in combination with the casualty location data, and generating a scene situation assessment result based on the regional danger level and the casualty distribution density; Based on the scene situation assessment result, dividing the rescue area according to the casualty distribution density, and determining the resource allocation weight of each rescue area based on the regional danger level, and generating a medical resource scheduling plan; Constructing a three-dimensional model of the emergency scene in the digital twin platform, visually mapping the scene situation assessment result and the medical resource scheduling plan, overlaying rescue situation annotations in real time through augmented reality technology, and updating the position information of rescue personnel and the position information of medical equipment in real time based on the target tracking algorithm; According to the position information of rescue personnel and the position information of medical equipment, dynamically adjusting the medical resource scheduling plan through a recursive optimization algorithm to achieve real-time optimal allocation of medical resources; Based on the scene situation assessment result, dividing the rescue area according to the casualty distribution density, and determining the resource allocation weight of each rescue area based on the regional danger level, the generated medical resource scheduling plan includes: Receiving the scene situation assessment result and the casualty distribution density data, constructing a prediction model based on the cellular automaton, analyzing the change trend of the scene situation, and obtaining the scene threat degree and the passable path of each area at the future moment; Calculating the density difference between adjacent areas according to the casualty distribution density data, marking the areas with a density difference less than the density threshold and a spatial distance less than the distance threshold as areas to be merged, and optimizing the boundaries of the areas to be merged in combination with the scene threat degree to generate rescue zones; Calculating the resource utilization efficiency of each rescue zone based on the corresponding relationship between the medical resource input volume and the casualty treatment effect in the historical rescue data, and calculating the regional danger level according to the casualty distribution density, the scene threat degree and the resource utilization efficiency within the rescue zone; Constructing a resource transfer network with the rescue zones as nodes and the passable paths as edges, calculating the weighted resource allocation weight by calculating the regional danger level and the resource utilization efficiency, and generating a medical resource scheduling plan with the minimum transfer loss based on the resource transfer network; Monitoring the change of the scene situation assessment result, when the change of the scene threat degree exceeds the situation threshold, re-dividing the rescue zone based on the updated scene threat degree and the casualty distribution density, adjusting the regional danger level and the resource allocation weight, and updating the medical resource scheduling plan.

2. The method according to claim 1, wherein Collecting thermal imaging data of the emergency scene through an infrared sensor, the extraction of casualty body temperature distribution data and casualty location data from the thermal imaging data includes: Collect the thermal imaging data of the emergency scene through an infrared sensor, perform adaptive histogram equalization enhancement processing on the thermal imaging data to obtain enhanced thermal imaging data, determine the filter window size based on the local temperature variance of the enhanced thermal imaging data, and use the adaptive median filtering algorithm to denoise the enhanced thermal imaging data to obtain denoised thermal imaging data; Establish a mapping function between the gray value and the actual temperature according to the temperature response characteristics of the infrared sensor, convert the gray value in the denoised thermal imaging data into a temperature value, generate a temperature distribution map, set a double-threshold detection interval based on the normal body temperature range of the human body, and perform regional connectivity analysis on the pixel points whose temperature values fall within the double-threshold detection interval in the temperature distribution map to obtain a candidate casualty area; Perform morphological processing on the candidate casualty area, extract the regional contour features, area features, and aspect ratio features, and screen out the areas that do not conform to the features according to the human body size feature template to obtain the confirmed casualty area; Establish a temperature field distribution model for the confirmed casualty area, extract the temperature numerical sequence, temperature gradient distribution, and temperature uniformity features within the area as the casualty body temperature distribution data; calculate the minimum circumscribed rectangle according to the contour features of the confirmed casualty area, and use the coordinates of the rectangle center point as the casualty location data.

3. The method according to claim 1, characterized in that, Construct a body temperature change trend graph according to the casualty body temperature distribution data, calculate the body temperature abnormality degree value, and at the same time calculate the environmental threat degree value according to the environmental parameter data. Calculate the regional risk level based on the body temperature abnormality degree value and the environmental threat degree value, combine the casualty location data to count the casualty distribution density, and generate a scene situation assessment result based on the regional risk level and the casualty distribution density, including: Establish a body temperature monitoring window, divide the body temperature monitoring window into an observation interval and a prediction interval, perform segmented fitting on the body temperature data in the observation interval to extract the fluctuation law, establish a body temperature baseline based on the prediction interval, calculate the degree of deviation of the fluctuation law from the body temperature baseline to obtain the fluctuation abnormality feature, and judge the body temperature abnormality degree value according to the fluctuation abnormality feature; Construct an influence network containing environmental parameters according to the environmental parameter data, use the leading abnormal parameter as the root node and the affected parameter as the sub-node to establish a directed graph structure of the parameter transfer relationship, analyze the abnormal diffusion path based on the directed graph structure, and calculate the environmental threat degree value by combining the scope and speed of the abnormal diffusion; Dynamically update the inference rules of the neural network according to the change trends of the body temperature abnormality degree value and the environmental threat degree value, and use the inference rules to calculate the regional risk level; Perform spatial distribution analysis on the casualty location data, establish a casualty movement trajectory graph, extract the trajectory intersection points as key nodes, and calculate the degree of personnel aggregation at each key node to obtain the casualty distribution density; Use the regional risk level as the weight coefficient to perform weighted calculation on the casualty distribution density to obtain the scene situation assessment result.

4. The method according to claim 3, wherein Perform spatial distribution analysis on the casualty location data, establish a casualty movement trajectory graph, extract the trajectory intersection points as key nodes, and calculate the degree of personnel aggregation at each key node to obtain the casualty distribution density, including: The casualty location data is segmented according to the time sequence to obtain a time window sequence. The displacement vector is calculated by computing the change in the position coordinates within the time window sequence. The casualty movement trajectory graph is constructed by connecting the displacement vectors in sequence. The casualty movement trajectory graph is smoothed to eliminate abnormal jumps, and the target movement path is extracted based on the smoothed trajectory. The target movement path is plotted in the spatial coordinate system. The crossing angle and the shortest distance between the paths are calculated. When the crossing angle and the shortest distance meet the preset conditions, the trajectory crossing points are determined. The crossing frequency at each trajectory crossing point is counted, and the number of people within the unit area around the trajectory crossing point is calculated to obtain the degree of personnel aggregation. The key nodes are determined based on the weighted values of the crossing frequency and the degree of personnel aggregation. A local coordinate system is established with the key nodes as the center. The crossing frequency and the degree of personnel aggregation are respectively mapped to the size of the influence range and the attenuation rate. The node influence field is constructed based on the influence range and the attenuation rate, and multiple node influence fields are superimposed to generate the distribution influence map. The real-time change characteristics of the casualty movement trajectory graph are converted into dynamic weights, and the distribution influence map is dynamically weighted to obtain the density distribution model. The target area is divided into grid cells. The distance between the center of each grid cell and the key node is calculated. The reference density value within each grid cell is calculated based on the distance and the density distribution model. The time-varying characteristics of the dynamic weights are superimposed on the reference density value to obtain the casualty distribution density reflecting the dynamic distribution of personnel.

5. The method according to claim 1, wherein A prediction model is constructed based on the cellular automaton to analyze the change trend of the scene situation, and the scene threat degree and the passable paths at each area in the future time are obtained, including: The rescue scene is divided into regular grid cells. The road traffic status, building damage degree, hazard source intensity, and environmental threat degree in the scene situation data are mapped to each grid cell to form the initial state matrix of the grid cells. The evolution value of the road traffic status is calculated according to the passing frequency of the rescue vehicle. The evolution value of the building damage degree is calculated according to the secondary disaster monitoring data. The diffusion values of the hazard source intensity and the environmental threat degree between adjacent grid cells are calculated. The road traffic status evolution value, the building damage degree evolution value, the diffusion values of the hazard source intensity and the environmental threat degree are constructed into the state transition equation. The state transition equation is solved by iterative calculation to obtain the state matrix of each grid cell at the prediction time, and the scene threat degree at the prediction time is calculated according to the state matrix. The passing probability is calculated based on the road traffic status and the scene threat degree of the grid cells, and the path cost function is constructed and solved by a search algorithm to obtain the passable paths. The scene situation data is received in real time. When the data is updated, the updated data is mapped to a new initial state matrix, and the prediction calculation is re-executed to dynamically update the scene threat degree and the passable paths.

6. The method according to claim 1, wherein A three-dimensional model of the emergency site is constructed in the digital twin platform. The scene situation assessment results and the medical resource scheduling plan are visually mapped. The rescue situation annotations are superimposed in real time through the augmented reality technology, and the position information of the rescue personnel and the medical equipment is updated in real time based on the target tracking algorithm, including: In the digital twin platform, on-vehicle lidar is used to obtain the point cloud data of the emergency site, the point cloud data is registered to obtain a dense point cloud, the dense point cloud is segmented by the region growing algorithm to generate a 3D model of the emergency site, and a unified spatial coordinate system is established in the 3D model; Based on the unified spatial coordinate system, the scene threat degree in the scene situation assessment result is mapped to the 3D model to form a situation heat map, and the equipment layout data and personnel deployment data in the medical resource scheduling plan are converted into position coordinates in the 3D scene and a situation annotation is generated, obtaining a digital twin scene with situation information; The feature point matching algorithm is used to realize the real-time registration of the mobile device camera and the digital twin scene with situation information, establish the mapping relationship between the camera view and the unified spatial coordinate system, and superimpose and display the situation heat map and the situation annotation in the camera view through the augmented reality technology to obtain an augmented reality situation view; Based on the target tracking algorithm, the position information of the medical device is obtained in real time, the position information collected by the rescue personnel positioning terminal is received, the position information is coordinate-transformed according to the unified spatial coordinate system, and the transformed position information is superimposed and displayed in the augmented reality situation view based on the mapping relationship; When the scene situation assessment result is updated, a new situation heat map is generated; when the medical resource scheduling plan changes, the situation annotation is updated; when the position information of the rescue personnel and the medical device is updated, the position display in the augmented reality situation view is updated in real time based on the mapping relationship.

7. An emergency medical command system based on AI and the Internet of Things for implementing the method according to any one of the preceding claims 1-6, characterized in that, It includes: The first unit is used to collect the thermal imaging data of the emergency site through an infrared sensor, extract the casualty body temperature distribution data and casualty position data according to the thermal imaging data, determine the environmental monitoring range according to the casualty position data, and collect the environmental parameter data through the environmental sensor within the range; The second unit is used to construct a body temperature change trend graph according to the casualty body temperature distribution data, calculate the body temperature abnormality degree value, calculate the environmental threat degree value according to the environmental parameter data at the same time, calculate the regional danger level based on the body temperature abnormality degree value and the environmental threat degree value, count the casualty distribution density in combination with the casualty position data, and generate a scene situation assessment result according to the regional danger level and the casualty distribution density; The third unit is used to divide the rescue area according to the casualty distribution density based on the scene situation assessment result, and determine the resource allocation weight of each rescue area based on the regional danger level to generate a medical resource scheduling plan; The fourth unit is used to construct a 3D model of the emergency site in the digital twin platform, perform visual mapping on the scene situation assessment result and the medical resource scheduling plan, superimpose the rescue situation annotation in real time through the augmented reality technology, and update the position information of the rescue personnel and the medical device in real time based on the target tracking algorithm; The fifth unit is used to dynamically adjust the medical resource scheduling plan through a recursive optimization algorithm according to the position information of the rescue personnel and the medical device to realize the real-time optimal allocation of medical resources.

8. An electronic device, characterized in that, It includes: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, the method according to any one of claims 1 to 6 is implemented.

Images