Disaster site injury condition identification device and distributed disaster response processing system

Abstract

The invention relates to a disaster site injury condition identification device and a distributed disaster response processing system. The method comprises the following steps: acquiring disaster data, and constructing an entity recognition model for multi-disaster emergency processing according to the data; meanwhile, vital sign data of the wounded person on a disaster site are obtained and processed through an injury condition classification model obtained based on machine learning training to generate an injury condition result, an injury condition score is obtained according to the injury condition result and the evaluation index corresponding to each model, and then an injury condition identifier corresponding to the wounded person is generated. And then, based on the constructed entity identification model, identifying and analyzing the disaster data and the injury identification, and generating a disaster response plan. And finally, distributing the pre-arranged plan to each disaster site, and obtaining a scheduling strategy of the corresponding disaster site. The pertinence and the practicability of the disaster response plan are remarkably improved, reasonable allocation of rescue resources and scientific arrangement of the wounded are ensured, and the high efficiency and the intelligence of the disaster response process are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of disaster response and rescue, and in particular relates to a disaster site injury identification device and a distributed disaster response processing system. Background Art

[0002] With the development of computer technology and its deep cross-application in the field of disaster response and rescue, disaster site injury identification devices and distributed disaster response processing systems suitable for disaster response and rescue have emerged. Existing on-site injury identification technology usually relies on on-site medical experts to collect the patient's vital signs according to the standard medical injury identification process to make a judgment on their injury level. The injury identification strategy is usually inefficient and relies on human subjective judgment, which has errors. At the same time, traditional disaster response processing systems often have difficulty in coping with complex and changeable disaster situations. When the amount of data is huge, the processing capacity is limited and the disaster response process cannot be made efficient and intelligent. Summary of the invention

[0003] Based on this, it is necessary to provide a disaster site injury identification device and a distributed disaster response system that can efficiently and intelligently handle disaster response at the disaster site in response to the above technical problems.

[0004] In a first aspect, the present application provides a disaster site injury identification device, comprising:

[0005] The module for acquiring the vital signs of the injured is used to obtain the vital signs data of the injured at the disaster site.

[0006] The injury classification prediction module includes an injury classification unit and an injury prediction unit; the injury classification unit is used to obtain a corresponding injury classification model based on machine learning training; each injury classification model receives vital sign data to generate an injury result; the injury prediction unit is used to obtain an injury score corresponding to each injury result based on each injury result and the evaluation index corresponding to each injury classification model.

[0007] The injury decision identification module is used to obtain the corresponding injury result based on the maximum injury score; it is also used to visualize the injury result of the injured person and generate the corresponding injury identification of the injured person.

[0008] In one embodiment, a method for calculating an evaluation index corresponding to the injury classification model includes:

[0009] The evaluation index corresponding to any injury classification model is its own corresponding M i value, and the M corresponding to the injury classification model i The value is calculated as follows:

[0010]

[0011] Among them, Mi represents the evaluation index corresponding to the i-th injury classification model, n represents the total number of injury types, DX i It indicates the number of injured persons with the i-th injury type that are correctly predicted by the current injury classification model as injured persons with the i-th injury type when training the current injury classification model; It represents the average value of the number of injured persons correctly predicted as corresponding injury types during the training of the injury classification model.

[0012] In one embodiment, the method for calculating the injury score corresponding to each injury result includes:

[0013]

[0014] Among them, ACEL represents the injury score vector corresponding to each injury result, M i represents the evaluation index corresponding to the i-th injury classification model, n represents the total number of injury types, A i,j represents the prediction result of the i-th injury classification model for the j-th injury type category, where A i,j =1, indicating that the i-th injury classification model predicts the injured person as the j-th injury type, A i,j =0, indicating that the i-th injury classification model did not predict the injured person as the j-th injury type category.

[0015] In one embodiment, a training method for any injury classification model includes:

[0016] A vital sign data set is constructed based on the historical injury electronic medical records, where each vital sign data in the vital sign data set includes blood pressure, blood oxygen concentration, respiratory rate and heart rate.

[0017] According to the fuzzy comprehensive evaluation method, each vital sign data is labeled to obtain the injury type category corresponding to each vital sign data, and based on the corresponding injury type category, a sample set of on-site injury classification is obtained.

[0018] Each vital sign data is normalized, and the injury type categories with on-site injury classification samples less than the set value are screened out as minor injury types.

[0019] Each minor injury type category is processed to generate a new on-site injury typing sample, until the number of on-site injury typing samples corresponding to the minor injury type category is not less than the set value, and a balanced on-site injury typing sample set is obtained.

[0020] The initial values ​​of the pending network parameters of the injury classification model to be trained are determined according to the brainstorming optimization algorithm.

[0021] The vital sign data in the balanced on-site injury classification sample set is used as the input of the current injury classification model to be trained with a determined initial value, and the injury type category corresponding to each vital sign data is used as the expected output of the current injury classification model to be trained with a determined initial value.

[0022] A loss function is constructed for the actual output and expected output of the current injury classification model to be trained, back propagation is performed based on the loss function, and the pending network parameters of the current injury classification model to be trained are modified until the loss function meets the set requirements, thereby completing the training of the current injury classification model to be trained.

[0023] In a second aspect, the present application provides a distributed disaster response processing system, the system comprising:

[0024] The disaster information acquisition device is used to collect disaster data; the disaster data includes disaster news data, historical disaster data, and real-time disaster monitoring data.

[0025] The recognition model building device is used to build an entity recognition model for multi-disaster emergency response based on the collected disaster data.

[0026] The disaster site injury identification device is used to collect the injury identification corresponding to the injured; the injury identification includes injury type, injury results and vital signs data.

[0027] The disaster response device is used to identify and analyze disaster data and injury identification based on the constructed entity recognition model, and generate a corresponding disaster response plan; the disaster response plan includes the allocation of rescue resources and the placement of the injured.

[0028] The disaster response processing device is used to distribute the disaster response plan to each disaster site and obtain the dispatching strategy corresponding to the disaster site.

[0029] In one embodiment, disaster data and injury identification are identified and analyzed based on the constructed entity recognition model to generate a corresponding disaster response plan, including:

[0030] Based on the collected disaster data and injury identification, an entity recognition model for the disaster-response area is constructed; the entity recognition model is used to identify disaster entities in the disaster data and injury identification.

[0031] Based on the disaster entities, the disaster entity relationships between disaster data and injury identification are extracted, and the association between each disaster entity relationship is identified to obtain a new disaster map.

[0032] Determine whether the new disaster is in the old graph. If so, extract the relationship path of the new disaster graph in the old graph; the old graph is an entity recognition model that has been constructed based on the current disaster types.

[0033] Analyze whether there are any contradictions between the relationship paths in the new disaster map and the old map. If so, mark the new relationships that do not appear in the new disaster map and obtain analysis results based on the new relationships.

[0034] The new disaster situation map is updated based on the analysis results and the relationship in the old map to obtain an updated new disaster situation map; and the disaster response plan is obtained by analyzing the updated new disaster situation map.

[0035] In one embodiment, disaster entity relationships are extracted from disaster data based on disaster entities, and association identification is performed on each disaster entity relationship to obtain a new disaster situation map, including:

[0036] Identify the multi-disaster entities associated with the old disaster entities and obtain the degree of association with the old disaster entities; obtain the relationship between the multi-disaster entities based on the degree of association and construct a new disaster map.

[0037] Identify the multi-disaster entities associated with the old disaster entities and obtain the degree of association with the old disaster entities, including:

[0038] The trained entity recognition model for the disaster-response area is used to identify one or more disaster entities in the disaster data to obtain identified old disaster entities.

[0039] A search engine is used to search each old disaster entity in the disaster data to obtain multiple web pages related to the old disaster entity.

[0040] The entity recognition model for the disaster-response area is used to perform disaster entity recognition on the disaster text information in each web page, and multiple newly recognized disaster entities associated with the old disaster entities are obtained.

[0041] The number of occurrences of each newly identified disaster entity is counted to obtain the degree of association between the newly identified disaster entity and the old disaster entity.

[0042] In one embodiment, the disaster response processing device further includes:

[0043] The division unit is used to divide the disaster site into several disaster response areas based on the disaster response plan.

[0044] The collection unit is used to collect on-site data of each disaster response area in real time; the on-site data includes data on the injured and medical resources and the relative distribution of medical resources and the injured in the corresponding disaster response area.

[0045] The judgment unit is used to judge whether the relative distribution of medical resources and wounded in each disaster response area meets the dispatch conditions; if so, the disaster response area that meets the dispatch conditions is determined as the target disaster response area.

[0046] The scheduling unit is used to generate a scheduling strategy according to the relative distribution of medical resources and wounded in the target disaster response area; the scheduling strategy is used to change the relative distribution of medical resources and wounded in the target disaster response area.

[0047] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0048] Collect disaster data; disaster data includes disaster news data, historical disaster data, and real-time disaster monitoring data.

[0049] An entity recognition model for multi-disaster emergency response is constructed based on the collected disaster data.

[0050] Collect the corresponding injury identification of the injured; the injury identification includes injury type, injury results and vital signs data.

[0051] Based on the constructed entity recognition model, disaster data and injury identification are identified and analyzed to generate corresponding disaster response plans; the disaster response plans include the allocation of rescue resources and the placement of the wounded.

[0052] Distribute the disaster response plan to each disaster site and obtain the dispatch strategy for the corresponding disaster site.

[0053] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the following steps are implemented:

[0054] Collect disaster data; disaster data includes disaster news data, historical disaster data, and real-time disaster monitoring data.

[0055] An entity recognition model for multi-disaster emergency response is constructed based on the collected disaster data.

[0056] Collect the corresponding injury identification of the injured; the injury identification includes injury type, injury results and vital signs data.

[0057] Based on the constructed entity recognition model, disaster data and injury identification are identified and analyzed to generate corresponding disaster response plans; the disaster response plans include the allocation of rescue resources and the placement of the wounded.

[0058] Distribute the disaster response plan to each disaster site and obtain the dispatch strategy for the corresponding disaster site.

[0059] The above-mentioned disaster site injury identification device and distributed disaster response processing system, computer equipment and storage medium collect disaster data including disaster news data, historical disaster data, and real-time disaster monitoring data, and build an entity recognition model for multi-disaster emergency response based on these data. At the same time, the vital signs data of the injured at the disaster site are obtained, and the injury classification model obtained based on machine learning training is used to process it. The injury results are generated based on the vital signs data received by each model, and then the injury scores are obtained based on these results and the evaluation indexes corresponding to each model to determine the final injury results, and then the results are visualized to generate the corresponding injury identification of the injured. After that, the disaster data and injury identification are identified and analyzed based on the constructed entity recognition model, and a disaster response plan including the allocation of rescue resources and the placement of the injured is generated. Finally, the plan is distributed to each disaster site, and the dispatch strategy for the corresponding disaster site is obtained. This not only significantly improves the pertinence and practicality of the disaster response plan, but also ensures the reasonable allocation of rescue resources and the scientific placement of the injured, realizes the rapid response and effective treatment of disasters, and improves the efficiency and intelligence of the disaster response process. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0061] Figure 1 A structural block diagram of a disaster scene injury identification device provided by an embodiment of the present invention;

[0062] Figure 2 A three-dimensional view of a disaster scene injury identification device provided by an embodiment of the present invention;

[0063] Figure 3 A flowchart of a training method for any injury classification model provided in an embodiment of the present invention;

[0064] Figure 4 A structural block diagram of a distributed disaster response system provided by an embodiment of the present invention;

[0065] Figure 5 The entity recognition model provided in the embodiment of the present invention is based on the construction to identify and analyze disaster data and injury identification, and generate a line flow chart of the corresponding disaster response plan;

[0066] Figure 6A flowchart of identifying a multi-disaster entity associated with an old disaster entity and obtaining a degree of association with the old disaster entity provided by an embodiment of the present invention;

[0067] Figure 7 A structural block diagram of a disaster response processing device of a distributed disaster response system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0069] The disaster site injury identification device and distributed disaster response processing system provided in the embodiments of the present application can enable rescue personnel to quickly identify the injuries of the wounded at the disaster site, improve rescue efficiency, accurately guide rescue measures, and avoid secondary injuries. The distributed disaster response processing system enhances emergency response capabilities, can quickly initiate rescue and flexibly allocate resources. Its distributed architecture avoids single point failures, ensures system reliability and data security, and can also achieve multi-department collaboration and resource sharing, integrate rescue forces from different regions and departments, and improve the ability to respond to large and complex disasters. Exemplarily, the disaster site injury identification device and distributed disaster response processing system provided in the embodiments of the present application can be applied to at least one scenario including but not limited to the following scenarios.

[0070] First, the disaster site injury identification device and the distributed disaster response system are applied to natural disaster sites. For example, at natural disaster sites such as earthquakes, floods, and typhoons, the disaster site injury identification device can quickly mark the injuries of the injured, such as serious injuries, minor injuries, or those requiring special medical care. These identification devices usually have eye-catching colors and graphics, so that rescuers can understand the condition of the injured at a glance, so as to give priority to treating the seriously injured and reasonably allocate medical resources. At the same time, the distributed disaster response system can collect and analyze data at the disaster site in real time, including the number of injured, the distribution of injuries, the distribution of rescue resources, etc., to provide a scientific basis for rescue decision-making and improve rescue efficiency.

[0071] Second, the disaster site injury identification device and the distributed disaster response system are applied to industrial production accident sites. For example, at the scene of industrial production accidents such as toxic and harmful, dusty, flammable and explosive, the disaster site injury identification device is also crucial. Because there may be complex dangerous factors at the accident site, rescuers need to quickly understand the location and injury of the injured in order to take effective rescue measures. Through the disaster site injury identification device, rescuers can quickly locate the injured and provide initial treatment based on the injury identification. In addition, the distributed disaster response system can monitor environmental changes at the accident site in real time, such as gas concentration, temperature, etc., to provide safety guidance for rescuers and reduce the risk of secondary injuries.

[0072] Third, the disaster site injury identification device and the distributed disaster response processing system are applied to the scene of urban public safety incidents. For example, at the scene of urban public safety incidents, such as fires, explosions, traffic accidents, etc., the disaster site injury identification device and the distributed disaster response processing system also play an important role. Through the disaster site injury identification device, rescue personnel can quickly understand the injuries and locations of the injured, and quickly evacuate and treat the injured. At the same time, the distributed disaster response processing system can summarize and analyze the information at the disaster site in real time, including the number of injured, the distribution of rescue resources, road conditions, etc., to provide timely and accurate information support for rescue decisions. In addition, the system can also coordinate the rescue forces of all parties to achieve optimal allocation and efficient use of resources, and improve the overall rescue effect.

[0073] In one embodiment, if Figure 1 As shown, a disaster scene injury identification device is provided, comprising:

[0074] The wounded person's vital sign acquisition module 101 is used to acquire the vital sign data of the wounded at the disaster site.

[0075] Vital sign data include blood pressure, blood oxygen concentration, respiratory rate and heart rate.

[0076] By obtaining the vital signs data of the injured on the scene, we can quickly understand the physiological status of the injured, provide key basis for subsequent injury assessment, emergency treatment and medical resource allocation, and effectively improve the timeliness and accuracy of disaster relief.

[0077] The injury classification prediction module 102 includes an injury classification unit 1021 and an injury prediction unit 1022; the injury classification unit 1021 is used to obtain a corresponding injury classification model based on machine learning training; each injury classification model receives vital sign data to generate an injury result; the injury prediction unit 1022 is used to obtain an injury score corresponding to each injury result according to each injury result and the evaluation index corresponding to each injury classification model;

[0078] The training method of the injury classification model involves the following steps: First, a data set containing vital signs such as blood pressure, blood oxygen concentration, respiratory rate and heart rate is constructed based on the historical injury electronic medical records, and the fuzzy comprehensive evaluation method is used to label these data to determine the injury type, thereby generating a sample set of on-site injury classification. Next, the vital sign data is normalized, and the small injury type categories with fewer samples than the set value are identified. New samples are generated through processing to ensure that the number of samples of each injury type category is balanced, forming a balanced on-site injury classification sample set. Then, the brainstorming optimization algorithm is used to determine the initial values ​​of the pending network parameters of the injury classification model to be trained. Subsequently, the vital sign data in the balanced sample set is used as the input of the model, and the corresponding injury type category is used as the expected output. The loss function is constructed, and the model parameters are adjusted through the back propagation algorithm until the loss function meets the set requirements, completing the training of the injury classification model.

[0079] This embodiment uses the injury classification model obtained through machine learning training to accurately receive and analyze vital sign data to generate accurate injury results. At the same time, combined with the evaluation index of each injury classification model, it can further quantify the injury, provide a scientific basis for rescue decision-making, and significantly improve the efficiency and accuracy of injury assessment at the disaster site. It achieves rapid response and effective handling of disasters, and improves the efficiency and intelligence of the disaster response process.

[0080] The injury decision identification module 103 is used to obtain the corresponding injury result based on the maximum injury score; it is also used to visualize the injury result of the injured person and generate an injury identification corresponding to the injured person.

[0081] Injury identification includes injury type, injury outcome and vital signs data.

[0082] like Figure 2 As shown, a three-dimensional view of a disaster site injury identification device is provided. The device obtains the vital signs data of the injured at the disaster site through the acquisition unit 01, and transmits the data to the processing unit 02 through the communication interface. The processing unit 02 uses the processor to process the vital signs data to determine the final injury result, and then sends the result data to the display unit 03 to display the injury result information. After the confirmation unit 04 confirms the injury result information, the processing unit 02 controls the printing unit 05 to print the result information to generate an injury identification.

[0083] The injury identification is a barcode or QR code with a unique identification printed on a waterproof surface. It has a color display that identifies the injury type, including the injury type, injury results and vital signs data of the injured person. By scanning the barcode or QR code, the injured person's information can be tracked and updated. It can also be worn on the injured person's wrist or ankle as a unique identity identification of the injured person.

[0084] This embodiment obtains the vital signs data of the injured at the disaster site, processes it using an injury classification model trained based on machine learning, generates injury results based on the vital signs data received by each model, and then obtains an injury score based on these results and the evaluation index corresponding to each model to determine the final injury result, and then visualizes the result to generate an injury identification corresponding to the injured.

[0085] In one embodiment, illustratively, in this embodiment, the calculation method of the evaluation index corresponding to the injury classification model may include:

[0086] The evaluation index corresponding to any injury classification model is its own corresponding M i value, and the M corresponding to the injury classification model i The value is calculated as follows:

[0087]

[0088] Among them, M i represents the evaluation index corresponding to the i-th injury classification model, n represents the total number of injury types, DX i It indicates the number of injured persons with the i-th injury type that are correctly predicted by the current injury classification model as injured persons with the i-th injury type when training the current injury classification model; It represents the average value of the number of injured persons correctly predicted as corresponding injury types during the training of the injury classification model.

[0089] In one embodiment, illustratively, in this embodiment, the method for calculating the injury score corresponding to each injury result may include:

[0090]

[0091] Among them, ACEL represents the injury score vector corresponding to each injury result, M i represents the evaluation index corresponding to the i-th injury classification model, n represents the total number of injury types, A i,j represents the prediction result of the i-th injury classification model for the j-th injury type category, where A i,j =1, indicating that the i-th injury classification model predicts the injured person as the j-th injury type, A i,j =0, indicating that the i-th injury classification model did not predict the injured person as the j-th injury type category.

[0092] The injury classification model is artificial neural network, support vector machine, decision tree or random forest, etc.

[0093] By integrating the evaluation indexes and prediction results of multiple injury classification models and calculating the injury scores corresponding to each injury result, the injuries of the injured can be comprehensively and accurately assessed, providing strong support for medical treatment. This evaluation and score calculation method not only improves the accuracy of injury classification, but also improves the efficiency and intelligence of the disaster response process.

[0094] In one embodiment, if Figure 3 As shown, illustratively, in this embodiment, the training method of any injury classification model may include the following steps:

[0095] Step 201, constructing a vital sign data set based on historical injury electronic medical records, wherein each vital sign data in the vital sign data set includes blood pressure, blood oxygen concentration, respiratory rate and heart rate.

[0096] Step 202 , labeling each piece of vital sign data according to the fuzzy comprehensive evaluation method, obtaining the injury type category corresponding to each piece of vital sign data, and obtaining a sample set of on-site injury classification based on the corresponding injury type category.

[0097] Step 203, normalize each piece of vital sign data, and select the injury type category with a number of on-site injury classification samples less than a set value as a minor injury type category.

[0098] Step 204, each minor injury type category is processed to generate a new on-site injury typing sample, until the number of on-site injury typing samples corresponding to the minor injury type category is not less than a set value, thereby obtaining a balanced on-site injury typing sample set.

[0099] Step 205, determining the initial values ​​of the pending network parameters of the injury classification model to be trained according to the brainstorming optimization algorithm.

[0100] Step 206, the vital sign data in the equalized on-site injury classification sample set is used as the input of the current injury classification model to be trained with a determined initial value, and the injury type category corresponding to each vital sign data is used as the expected output of the current injury classification model to be trained with a determined initial value.

[0101] Step 207, construct a loss function for the actual output and expected output of the current injury classification model to be trained, perform back propagation based on the loss function, and modify the pending network parameters of the current injury classification model to be trained until the loss function meets the set requirements, thereby completing the training of the current injury classification model to be trained.

[0102] In step 203, each piece of vital sign data is normalized to eliminate the dimensional influence between different vital sign data indicators.

[0103] In step 204, the ADASYN oversampling algorithm is used to process each minor injury type to generate a new on-site injury classification sample. Sample balancing is to deal with the extreme imbalance in the number of different types of injured persons to reduce the risk of model overfitting.

[0104] In step 205, the pending network parameters are determined by the type of injury classification model. For example, the pending network parameters of the random forest model include the number of decision trees, the depth of the decision tree, the maximum number of leaf nodes, the minimum number of sample splits, and the feature sampling ratio.

[0105] In the specific training process, the balanced on-site injury classification sample set is divided into training set, validation set and test set in proportion; the scale ratio of training set, validation set and test set is usually 6:2:2, which can be appropriately adjusted according to the sample size.

[0106] By integrating historical injury electronic medical records to construct a vital sign data set, and using the fuzzy comprehensive evaluation method to label the injury type to form an on-site injury classification sample set, normalization is then performed and the number of samples in the minor injury category is balanced to ensure the comprehensiveness and accuracy of model training. Next, the brainstorming optimization algorithm is used to initialize the network parameters of the injury classification model, and the balanced sample set is used as input for training. The loss function is constructed and the model parameters are optimized based on back propagation until the set requirements are met. This series of steps not only improves the training efficiency and generalization ability of the injury classification model, but also effectively solves the model deviation problem caused by the scarcity of samples in the minor injury category, providing strong support for rapid injury assessment and rescue resource allocation at disaster sites.

[0107] In a second aspect, the present application provides a distributed disaster response processing system, such as Figure 4 As shown, the system includes:

[0108] The disaster information acquisition device 301 is used to collect disaster data; the disaster data includes disaster news data, historical disaster data, and real-time disaster monitoring data.

[0109] Disaster news data, obtain the latest disaster information by real-time capture and analysis of various news reports.

[0110] Historical disaster data, review and analyze previous disaster cases, and extract valuable experiences and lessons.

[0111] Real-time disaster monitoring data uses modern sensors and monitoring technology to obtain real-time data from disaster sites, such as seismic waves, rainfall, wind speed, etc.

[0112] By integrating diversified disaster information resources, we provide comprehensive and timely data support for disaster warning, risk assessment and emergency response, thereby enhancing the scientific nature and effectiveness of disaster management.

[0113] The recognition model building device 302 is used to build an entity recognition model for multi-disaster emergency response based on the collected disaster data.

[0114] First, search engines are used to search for disaster news from various places, and the retrieved disaster news is obtained through web crawlers; the obtained disaster news contains information that is not related to the disaster, so it is necessary to manually process the obtained disaster news to remove the information that is not related to the disaster in the news, and split each news into multiple sentences; each sentence is taken as a piece of data, and the sequence labeling method is used to annotate each word in each data, and the sequence labeling method is used as the label of each data to construct a multi-disaster corpus, and the data in the multi-disaster corpus is divided into a training set and a test set according to a preset ratio.

[0115] The data in the disaster-response area corpus training set is input into the BERT pre-trained model to obtain the embedded representation of the input data, and the bidirectional long short-term memory network is used to encode the embedded representation; then the encoded representation is input into the CRF, i.e., conditional random field, to obtain the label of each word in each data; the back propagation technology is used to continuously optimize the parameters in the model, and the model is tested with the data in the test set. By continuously iterating the model until the accuracy of the model in the test set reaches the preset threshold, a multi-disaster entity recognition model is obtained, and the corresponding model parameters of the model are saved.

[0116] This embodiment can automatically identify and extract key information from complex disaster data efficiently, provide accurate data support for emergency decision-making, significantly improve the speed and efficiency of disaster response, and thus more effectively deal with various disaster situations.

[0117] The disaster site injury identification device 303 is used to collect the injury identification corresponding to the injured; the injury identification includes the injury type, injury result and vital sign data.

[0118] Vital signs data can reflect the severity and type of the casualties' injuries, ensuring that the casualties receive timely and appropriate treatment.

[0119] The disaster response device 304 is used to identify and analyze disaster data and injury identification based on the constructed entity recognition model, and generate a corresponding disaster response plan; the disaster response plan includes the allocation of rescue resources and the placement of the injured.

[0120] The entity recognition model will conduct in-depth analysis of a large amount of disaster data and extract key information elements, such as disaster type, location, impact range, disaster intensity, etc. For injury identification, the model will accurately identify the specific degree of injury represented by different identification symbols, including minor injuries, serious injuries, critical injuries, etc., as well as detailed information such as the injured part and bleeding. By identifying these data and identifications, the model will use algorithms and data analysis techniques to comprehensively evaluate the severity of the disaster, possible development trends, and the overall condition of the injured. On this basis, combined with past experience in responding to similar disasters and professional emergency response knowledge, a corresponding disaster response plan is generated. This plan not only effectively improves the timeliness and pertinence of disaster response, but also ensures the rational allocation and efficient use of rescue resources, providing a solid guarantee for minimizing disaster losses.

[0121] The disaster response processing device 305 is used to distribute the disaster response plan to each disaster site and obtain the dispatching strategy corresponding to the disaster site.

[0122] Based on the disaster response plan, the disaster site is first scientifically and reasonably divided into several disaster response areas through the division unit, and then the acquisition unit is used to obtain the data of the injured and medical resources in each area in real time, as well as the relative distribution between them. Next, the judgment unit will evaluate whether the medical resources and the distribution of the injured in each area meet the dispatching conditions one by one. Once it is determined that a certain area meets the conditions, it will be regarded as the target disaster response area. Finally, the dispatch unit will quickly generate and implement effective dispatching strategies according to the specific conditions of the target disaster response area to optimize the allocation of medical resources and the treatment process of the injured. This not only significantly improves the pertinence and practicality of the disaster response plan, but also ensures the rational allocation of rescue resources and the scientific placement of the injured, realizes rapid response and effective handling of disasters, and improves the efficiency and intelligence of the disaster response process.

[0123] In one embodiment, if Figure 5 As shown, illustratively, identifying and analyzing disaster data and injury identification based on the constructed entity recognition model to generate a corresponding disaster response plan may include the following steps:

[0124] Step 401, constructing an entity recognition model for the disaster-stricken area based on the collected disaster data and injury identification; the entity recognition model is used to identify disaster entities in the disaster data and injury identification.

[0125] Step 402, based on the disaster entity, the disaster data and the injury identification are extracted from the disaster entity relationship, and the association of each disaster entity relationship is identified to obtain a new disaster map.

[0126] Step 403, determine whether the new disaster is in the old graph, if so, extract the relationship path of the new disaster graph in the old graph; the old graph is an entity recognition model that has been constructed based on the current disaster types.

[0127] Step 404, analyzing whether there are any contradictions between the relationship paths in the new disaster map and the old map. If so, marking new relationships that do not appear in the new disaster map, and obtaining analysis results based on the new relationships.

[0128] Step 405, updating the new disaster map based on the analysis results and the relationship in the old map to obtain an updated new disaster map; analyzing the updated new disaster map to obtain a disaster response plan.

[0129] Among them, in step 402, based on the disaster entity, the disaster entity relationship is extracted from the disaster data and the injury identification, and the association identification of each disaster entity relationship is performed to obtain a new disaster map, which specifically includes:

[0130] Step 4021, obtaining multiple disaster entities related to the old disaster entity; based on the multiple disaster entities related to the old disaster entity, using a search engine to search for disaster data of the multiple disaster entities related to the old disaster entity in each web page.

[0131] Step 4022, based on the disaster data and injury identification, relationships are set for multiple disaster entities to obtain set relationship types; the set relationship types include disaster risk assessment, solutions, causes, and precursors.

[0132] Step 4023, according to the set relationship type, use Python regular expressions to match sentences containing both old disaster entities and the set relationship type from the disaster data to obtain successfully matched sentences; based on the successfully matched sentences, extract the corresponding values ​​of each set relationship type.

[0133] Step 4024, based on the new disaster name, relationship type, and the value corresponding to the relationship type, a triplet of the new disaster is constructed and saved to obtain a new disaster map.

[0134] In step 403, it is determined whether the new disaster is in the old map. If so, the relationship path between the new disaster map and the old map is extracted, including:

[0135] Step 4031, construct an old graph for disasters, and use the cosine similarity matching algorithm to obtain the correlation between the new disaster and the existing disasters in the old graph;

[0136] The cosine similarity matching formula is as follows:

[0137]

[0138] Among them, A i and B i They represent the new disaster names and the disaster names in the old atlas respectively.

[0139] When the matching value is greater than a preset threshold, it indicates that the identified new disaster or a disaster similar to the new disaster type has been saved in the old map, that is, the old map stores relevant entities of the corresponding disaster.

[0140] When the matching value is less than the preset threshold, it indicates that the identified new disaster has never appeared in the old map or the disaster type in the old map has a low correlation with the new disaster type.

[0141] Step 4032, determine whether the new disaster is in the old map based on the correlation. If the new disaster is in the old map, obtain the disaster related to the new disaster type in the old map to obtain the disaster that successfully matches the new disaster.

[0142] Step 4033, extracting the relationship path of the new disaster map in the old map based on the disasters that successfully match the new disaster.

[0143] This series of steps not only improves the accuracy and efficiency of disaster response, but also ensures the pertinence and effectiveness of disaster response plans, and provides a scientific basis for disaster rescue and treatment of the wounded.

[0144] In one embodiment, if Figure 6 As shown, exemplarily, extracting disaster entity relationships from disaster data based on disaster entities, and identifying associations between disaster entity relationships to obtain a new disaster situation map may include the following steps:

[0145] Identify the multi-disaster entities associated with the old disaster entities and obtain the degree of association with the old disaster entities; obtain the relationship between the multi-disaster entities based on the degree of association and construct a new disaster map.

[0146] Identify the multi-disaster entities associated with the old disaster entities and obtain the degree of association with the old disaster entities, including:

[0147] Step 501: Use the trained entity recognition model for the disaster-stricken area to identify one or more disaster entities in the disaster data to obtain identified old disaster entities.

[0148] Step 502: Use a search engine to search for each old disaster entity in the disaster data to obtain a plurality of web pages related to the old disaster entity.

[0149] Step 503, using the entity recognition model for the disaster-response area to perform disaster entity recognition on the disaster text information in each web page, and obtain multiple newly recognized disaster entities associated with the old disaster entities.

[0150] Step 504 , counting the number of occurrences of each newly identified disaster entity, and obtaining the correlation degree between the newly identified disaster entity and the old disaster entity.

[0151] Wherein, step 504 also includes judging the number of occurrences of the disaster entity just identified according to a set threshold value to obtain relevant judgment results:

[0152] When the number of occurrences of the newly identified disaster entity is greater than or equal to the set threshold, the disaster entity is determined to be a multi-disaster entity associated with the old disaster entity, and each strongly associated disaster entity is saved in the database document.

[0153] When the number of occurrences is less than a set threshold, the disaster entity is determined to be a multi-disaster entity that is not associated with the old disaster entity, and each unrelated multi-disaster entity is discarded.

[0154] The entity recognition model for disaster response areas obtained through training is used to identify old disaster entities in disaster data, and a search engine is used to search for multiple web pages related to each old disaster entity. Then, the entity recognition model is used again to conduct in-depth analysis of the disaster text information in these web pages to identify new disaster entities associated with the old disaster entities. Subsequently, the frequency of occurrence of each newly identified disaster entity is counted, and the degree of association between them and the old disaster entities is calculated based on this. Based on these association degrees, a detailed new disaster map is constructed, which clearly reveals the relationship between multiple disaster entities. This process not only improves the accuracy and efficiency of disaster entity recognition, but also provides more scientific and accurate decision-making support for disaster prediction, response and rescue by constructing a comprehensive disaster map, greatly improving the overall effectiveness of disaster management.

[0155] In one embodiment, if Figure 7 As shown, illustratively, the disaster response processing device may also include:

[0156] The division unit 3051 is used to divide the disaster site into a number of disaster response areas based on the disaster response plan.

[0157] The collection unit 3052 is used to collect the on-site data of each disaster response area in real time; the on-site data includes the data of the injured and the medical resources and the relative distribution of the medical resources and the injured in the corresponding disaster response area.

[0158] The judgment unit 3053 is used to judge whether the relative distribution of medical resources and wounded in each disaster response area meets the dispatching conditions; if so, the disaster response area that meets the dispatching conditions is determined as the target disaster response area.

[0159] The scheduling unit 3054 is used to generate a scheduling strategy according to the relative distribution of medical resources and wounded persons in the target disaster response area; the scheduling strategy is used to change the relative distribution of medical resources and wounded persons in the target disaster response area.

[0160] In this embodiment, the division unit can customize the division of the disaster site into several disaster response areas according to actual applications, wherein the area number of each disaster response area can also be customized to identify the disaster response area according to actual applications.

[0161] Among them, the collection unit is also specifically used to collect the real-time location of medical staff and the amount of medical resources supplied to each disaster response area in real time. Determine the disaster response area where the corresponding medical staff is located according to the real-time location of the medical staff. Count the total number of wounded, the total number of medical staff and the remaining amount of medical resources in each disaster response area, wherein the remaining amount of medical resources is determined by the amount of medical resources used and the amount of medical resources supplied in the corresponding disaster response area. According to the total number of medical staff and the total number of wounded in each disaster response area, the ratio of medical staff to wounded in the corresponding disaster response area is obtained respectively. According to the remaining amount of Chinese medicine resources and the total number of wounded in each disaster response area, the ratio of the remaining amount of Chinese medicine resources to the wounded in the corresponding disaster response area is obtained respectively.

[0162] In the judgment unit, it is judged whether the relative distribution of medical resources and wounded in each disaster response area meets the relative distribution threshold; if so, it is determined that the disaster response area whose relative distribution of medical resources and wounded is less than the relative distribution threshold meets the scheduling conditions.

[0163] In the scheduling unit, a scheduling strategy adapted to the target disaster response area is generated and issued according to the relative distribution of medical resources and wounded in the target disaster response area that meets the scheduling conditions, so as to achieve the distribution adjustment of medical resources and wounded in the disaster response area. Therefore, the present invention realizes the efficient and coordinated development of disaster site rescue work on the basis of information sharing among several disaster response areas, thereby reducing secondary casualties.

[0164] This embodiment aims to optimize the allocation of medical resources and effectively adjust the relative distribution of the injured and medical resources, thereby improving the efficiency and effectiveness of rescue. This systematic process greatly enhances the emergency response capabilities at the disaster site, ensures the rational allocation and utilization of rescue resources, and provides strong technical support for saving lives and reducing injuries.

[0165] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0166] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements steps related to a disaster site injury identification device and a distributed disaster response processing system as described above.

[0167] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0168] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can refer to the partial description of the method embodiments. The device embodiments described above are only schematic, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0169] The above-mentioned embodiments only express several implementation methods of the embodiments of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the embodiments of the present application. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the embodiments of the present application, and these all belong to the protection scope of the embodiments of the present application.

Claims

1. A disaster site injury identification device, characterized in that: The device comprises: The module for acquiring the vital signs of the injured is used to obtain the vital signs data of the injured at the disaster site; An injury classification prediction module comprises an injury classification unit and an injury prediction unit; the injury classification unit is used to obtain a corresponding injury classification model based on machine learning training; each of the injury classification models receives the vital sign data to generate an injury result; the injury prediction unit is used to obtain an injury score corresponding to each injury result according to each injury result and an evaluation index corresponding to each injury classification model; The injury decision identification module is used to obtain the corresponding injury result based on the maximum injury score; it is also used to visualize the injury result of the injured person and generate an injury identification corresponding to the injured person.

2. A disaster scene injury identification device according to claim 1, characterized in that: The calculation method of the evaluation index corresponding to the injury classification model includes: The evaluation index corresponding to any of the injury classification models is its own corresponding M i value, and the injury classification model corresponds to M i The value is calculated as follows: Among them, M i represents the evaluation index corresponding to the i-th injury classification model, n represents the total number of injury types, DX i It indicates the number of injured persons with the i-th injury type that are correctly predicted by the current injury classification model as injured persons with the i-th injury type when training the current injury classification model; It represents the average value of the number of injured persons correctly predicted as corresponding injury types during the training of the injury classification model.

3. A disaster scene injury identification device according to claim 1, characterized in that: The calculation methods of the injury scores corresponding to each injury result include: Among them, ACEL represents the injury score vector corresponding to each injury result, M i represents the evaluation index corresponding to the i-th injury classification model, n represents the total number of injury types, A i,j represents the prediction result of the i-th injury classification model for the j-th injury type category, where A i,j =1, indicating that the i-th injury classification model predicts the injured person as the j-th injury type, A i,j =0, indicating that the i-th injury classification model did not predict the injured person as the j-th injury type category.

4. A disaster scene injury identification device according to claim 2, characterized in that: The training method of any of the injury classification models comprises: Constructing a vital sign data set based on the historical injury electronic medical records, wherein each vital sign data in the vital sign data set includes blood pressure, blood oxygen concentration, respiratory rate and heart rate; Labeling each piece of the vital sign data according to the fuzzy comprehensive evaluation method to obtain the injury type category corresponding to each piece of the vital sign data, and obtaining a sample set of on-site injury classification based on the corresponding injury type category; Normalizing each piece of the vital sign data, and screening out the injury type category whose number of on-site injury classification samples is less than a set value as a minor injury type category; Processing each of the minor injury types to generate new on-site injury typing samples, until the number of on-site injury typing samples corresponding to the minor injury type is not less than the set value, thereby obtaining a balanced on-site injury typing sample set; Determine the initial values ​​of the pending network parameters of the injury classification model to be trained according to the brainstorming optimization algorithm; The vital sign data in the equalized on-site injury classification sample set is used as the input of the current injury classification model to be trained with the determined initial value, and the injury type category corresponding to each vital sign data is used as the expected output of the current injury classification model to be trained with the determined initial value; A loss function is constructed for the actual output and expected output of the current injury classification model to be trained, back propagation is performed based on the loss function, and the pending network parameters of the current injury classification model to be trained are modified until the loss function meets the set requirements, thereby completing the training of the current injury classification model to be trained.

5. A distributed disaster response system, characterized in that: The system comprises: A disaster information acquisition device is used to collect disaster data; the disaster data includes disaster news data, historical disaster data, and real-time disaster monitoring data; A recognition model building device, used to build an entity recognition model for multi-disaster emergency response based on the collected disaster data; The disaster site injury identification device is used to collect the injury identification corresponding to the injured; the injury identification includes the injury type category, the injury result and the vital sign data; A disaster response device, used to identify and analyze the disaster data and the injury identification based on the constructed entity recognition model, and generate a corresponding disaster response plan; the disaster response plan includes the deployment of rescue resources and the placement of the injured; The disaster response processing device is used to distribute the disaster response plan to each disaster site to obtain a dispatching strategy corresponding to the disaster site.

6. A distributed disaster response system according to claim 5, characterized in that: The entity recognition model constructed based on the above structure identifies and analyzes the disaster data and the injury identification, and generates a corresponding disaster response plan, including: Based on the collected disaster data and the injury identification, an entity recognition model for the disaster response area is constructed; the entity recognition model is used to identify disaster entities in the disaster data and the injury identification; Extracting disaster entity relationships between the disaster data and the injury identification based on the disaster entity, and identifying associations between the disaster entity relationships to obtain a new disaster map; Determine whether the new disaster is in the old graph, and if so, extract the relationship path of the new disaster graph in the old graph; the old graph is an entity recognition model that has been constructed based on the current disaster types; Analyze whether there is a path contradiction between the relationship paths in the new disaster map and the old map. If so, mark a new relationship that does not appear in the new disaster map, and obtain an analysis result based on the new relationship; The new disaster situation map is updated based on the analysis result and the relationship in the old map to obtain an updated new disaster situation map; and the disaster response plan is obtained by analyzing the updated new disaster situation map.

7. A distributed disaster response system according to claim 6, characterized in that: The extracting of disaster entity relationships from the disaster data based on the disaster entities and identifying associations between the disaster entity relationships to obtain a new disaster situation map includes: Identify the multi-disaster entity associated with the old disaster entity, and obtain the degree of association with the old disaster entity; obtain the relationship between the multi-disaster entities based on the degree of association, and construct the new disaster map; The identifying of the multi-disaster entity associated with the old disaster entity and obtaining the degree of association with the old disaster entity specifically includes: Using the trained entity recognition model for the disaster-stricken area to identify one or more disaster entities in the disaster data, and obtain the identified old disaster entities; Using a search engine to search each of the old disaster entities in the disaster data to obtain a plurality of web pages related to the old disaster entities; Using the entity recognition model for the disaster-stricken area to perform disaster entity recognition on the disaster text information in each web page, and obtaining a plurality of newly recognized disaster entities associated with the old disaster entities; The number of occurrences of each newly identified disaster entity is counted to obtain the correlation degree between the newly identified disaster entity and the old disaster entity.

8. A distributed disaster response system according to claim 5, characterized in that: The disaster response processing device further comprises: A division unit, used for dividing the disaster site into a plurality of disaster response areas based on the disaster response plan; A collection unit, used for collecting on-site data of each disaster response area in real time; the on-site data includes data of injured persons and medical resources and the relative distribution of medical resources and injured persons in the corresponding disaster response area; A judgment unit, used to judge whether the relative distribution of medical resources and wounded persons in each disaster response area meets the dispatching conditions; if so, determine the disaster response area meeting the dispatching conditions as the target disaster response area; A scheduling unit, used to generate a scheduling strategy according to the relative distribution of medical resources and injured persons in the target disaster response area; The dispatching strategy is used to change the relative distribution of medical resources and wounded persons in the target disaster response area.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the device or system according to any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the device or system according to any one of claims 1 to 8 is implemented.

Images