An AI-based auxiliary diagnosis method and system for viral pneumonia

Abstract

An AI-based auxiliary diagnosis method and system for viral pneumonia, which relates to the field of pneumonia diagnosis, includes: monitoring the abnormal state of the target population in the target area, and when it is detected that the target human body is in the abnormal state within a preset time period and the number of abnormal behaviors exceeds the preset threshold, the location, movement trajectory and real-time crowd density are mapped to the three-dimensional building model to obtain spatial environment association data; the spatial environment association data is input into the propagation prediction model to obtain the propagation risk probability of each target sub-area in the target area; the target sub-area with a propagation risk probability greater than the preset risk threshold is marked as a target risk area, and the first warning information containing the target risk area and the target human body is generated. The implementation of this method can improve the accuracy of viral pneumonia warning information.

Description

Technical Field

[0001] The present application relates to the field of pneumonia diagnosis, and in particular to an AI-based auxiliary diagnosis method and system for viral pneumonia. Background Art

[0002] With the acceleration of urbanization and the increase in population mobility, the prevention and control of respiratory infectious diseases such as viral pneumonia faces severe challenges. In crowded places, rapid identification of potential infected persons and timely adoption of prevention and control measures are crucial to effectively block the transmission chain of infectious diseases.

[0003] In the relevant technologies, infrared temperature measurement equipment can be used to screen body temperature in crowded places, and video monitoring systems can be used to manually observe suspicious symptoms such as coughing. At the same time, environmental monitoring equipment is deployed in key areas to collect environmental parameters such as temperature and humidity, and a simple transmission risk assessment model is established in combination with historical data.

[0004] However, relying solely on data obtained from discrete monitoring points is difficult to accurately reflect the actual spread of infectious diseases in the spatial environment, especially in scenarios with complex building structures and dynamically changing crowd density. The existing risk assessment models cannot effectively combine spatial environmental characteristics with the patterns of human activities, resulting in insufficient accuracy of early warning information. Summary of the invention

[0005] The present application provides an AI-based auxiliary diagnosis method and system for viral pneumonia, which is used to improve the accuracy of viral pneumonia warning information.

[0006] In a first aspect, the present application provides an AI-based auxiliary diagnosis method for viral pneumonia, which is applied to an AI-based auxiliary diagnosis system for viral pneumonia, the method comprising: monitoring the target population in the target area for abnormal conditions, wherein the abnormal conditions include abnormal body temperature and preset viral pneumonia symptoms; when it is detected that the target human body is in the abnormal state within a preset time period and the number of abnormal behaviors exceeds a preset threshold, obtaining a three-dimensional building model of the target area, the occurrence location of the abnormal state, the movement trajectory of the target human body, and the real-time crowd density in the target area; mapping the occurrence location, the movement trajectory, and the real-time crowd density to the three-dimensional building model to obtain spatial environment associated data; inputting the spatial environment associated data into a propagation prediction model to obtain the propagation risk probability of each target sub-area in the target area; marking the target sub-area whose propagation risk probability is greater than the preset risk threshold as a target risk area, and generating a first warning information containing the target risk area and the target human body.

[0007] By adopting the above technical solution, the abnormal status of the people in the target area can be monitored, and the target people with abnormal body temperature and preset viral pneumonia symptoms can be found in time. When the abnormal behavior of the target person exceeds the threshold within the preset time period, multiple data are obtained and mapped to the three-dimensional building model to obtain spatial environment related data, and then the transmission risk probability of each target sub-area is calculated through the transmission prediction model, and the high-risk area is marked and early warning information is generated. Taking into account multiple factors such as personnel status, regional environment and personnel activity patterns, the accuracy of early warning information is improved, and potential transmission risk areas can be more accurately identified.

[0008] In combination with some embodiments of the first aspect, in some embodiments, the step of monitoring the target population in the target area for abnormal conditions specifically includes: obtaining body temperature data of the target population in the target area based on an infrared temperature measuring device; identifying whether the target population in the video data within the target area has the preset viral pneumonia symptoms, and obtaining symptom characteristic data; performing spatiotemporal correlation analysis on the body temperature data and the symptom characteristic data to determine the target human body in the target population that is in the abnormal state, and recording the number of abnormal behaviors of the target human body within a preset time period.

[0009] By adopting the above technical solution, the infrared temperature measuring device can obtain body temperature data, and the video data analysis can obtain symptom characteristic data. The temporal and spatial correlation analysis of the two can determine the target human body in abnormal state and record the number of abnormal behaviors. The infrared temperature measuring device can quickly obtain the body temperature of the person, the video data analysis can accurately capture the preset symptoms, and the temporal and spatial correlation analysis can organically combine the two, effectively avoiding the limitations of single data monitoring, making the judgment of the abnormal state of the target population more comprehensive and accurate, so as to detect potential infected persons earlier and more accurately.

[0010] In combination with some embodiments of the first aspect, in some embodiments, the step of mapping the occurrence location, the movement trajectory and the real-time crowd density to the three-dimensional building model to obtain spatial environment associated data specifically includes: dividing the target area into multiple target sub-areas based on the three-dimensional building model; calculating the spatial distance and ventilation coverage between each of the target sub-areas; obtaining the temperature and humidity data and airflow data in each of the target sub-areas; performing spatiotemporal mapping of the occurrence location, the movement trajectory and the spatial distance of each of the target sub-areas to the activity trajectory density of the target human body in each of the target sub-areas; performing weighted calculation of the activity trajectory density and the spatial distance of each of the target sub-areas to obtain a spatial position association degree; based on the ventilation coverage, temperature and humidity data and airflow data of each of the target sub-areas, performing environmental factor correction on the spatial position association degree; and weightedly combining the real-time crowd density of each of the target sub-areas with the spatial position association degree based on preset weights to obtain spatial environment associated data.

[0011] By adopting the above technical solution, the target sub-areas are divided based on the three-dimensional building model, the analysis units are refined, and the risks are accurately located. The spatial distance and ventilation coverage between sub-areas are calculated, the convenience of personnel flow and air circulation conditions can be intuitively presented, and temperature, humidity and airflow data can be obtained. The multi-source data is obtained through time-space mapping, weighted calculation and environmental correction to obtain spatial environment correlation data, providing richer and more accurate input for the propagation prediction model.

[0012] In combination with some embodiments of the first aspect, in some embodiments, before the step of monitoring the target population in the target area for abnormal conditions, wherein the abnormal conditions include abnormal body temperature and preset viral pneumonia symptoms, the method also includes: obtaining historical monitoring data containing confirmed cases, wherein the historical monitoring data includes a three-dimensional building model of the historical target area, the onset location of the confirmed case, the movement trajectory of the confirmed case, the real-time crowd density data of the historical target area, and the transmission situation record in the historical target area; and dividing the historical target area into multiple historical target sub-areas based on the three-dimensional building model, and calculating the spatial distance and ventilation coverage range between each of the historical target sub-areas; performing spatiotemporal mapping on the onset location, the movement trajectory and the spatial distance of each of the historical target sub-areas to obtain the activity trajectory density of the confirmed case in each of the historical target sub-areas; performing correlation analysis on the activity trajectory density and the real-time crowd density data of the historical target area to obtain the transmission risk characteristics of each of the historical target sub-areas; and inputting the transmission risk characteristics and the transmission situation records into the transmission risk initial model for training to obtain a transmission prediction model.

[0013] By adopting the above technical solution, we first obtain historical monitoring data, which contains a lot of key information of confirmed cases, and use the three-dimensional building model to divide the historical target sub-areas, calculate the spatial distance and ventilation coverage, make the regional analysis more detailed, and map the location of the disease with the spatial distance of the sub-area in time and space. Combined with the real-time crowd density data correlation analysis, we can derive the transmission risk characteristics, and input the transmission risk characteristics and transmission situation records into the initial model training. The model continuously learns and optimizes, and finally the transmission prediction model can more accurately assess the probability of transmission risk based on real-time data, thereby enhancing the adaptability of the diagnostic system to virus transmission risk assessment in different scenarios.

[0014] In combination with some embodiments of the first aspect, in some embodiments, after the step of marking the target sub-area where the probability of transmission risk is greater than a preset risk threshold as a target risk area and generating a first warning information containing the target risk area and the target person, the method also includes: when the target area is a transportation station, tracking the activity trajectory of the target person to obtain the waiting area and entry time of the target person; obtaining the train schedule corresponding to the waiting area and the entry time; generating a second warning information containing the train schedule and the target person; and sending the second warning information to the target train client corresponding to the train schedule.

[0015] By adopting the above technical solution, when the target area is a transportation station, the target person's activity trajectory is tracked to determine the waiting area and entry time point, and the target person's activities in the station are located. The train schedule is obtained based on this, and the generated second warning information is highly targeted. After being sent to the train client, the train staff can know the potential risks in time, realizing the effective connection between transportation station and train information, so that prevention and control measures can be deployed on the train in advance.

[0016] In combination with some embodiments of the first aspect, in some embodiments, after the step of sending the second warning information to the target train client corresponding to the train schedule, the method also includes: performing a risk level assessment on the carriage corresponding to the train schedule based on the second warning information and the activity trajectory of the target human body to obtain the carriage risk level; generating ventilation parameters corresponding to each carriage according to the carriage risk level, and generating ventilation control prompt information containing the ventilation parameters; and sending the ventilation control prompt information to the target train client.

[0017] By adopting the above technical solution, the risk level of the carriage is evaluated based on the second warning information and the target human activity trajectory. The activity trajectory provides an important basis for the evaluation, making the risk level division more realistic. Generating ventilation parameters according to different risk levels effectively reduces the risk of virus transmission in the carriage.

[0018] In combination with some embodiments of the first aspect, in some embodiments, after the step of marking the target sub-area where the transmission risk probability is greater than a preset risk threshold as a target risk area and generating a first warning information containing the target risk area and the target human body, the method further includes: generating a risk area distribution identifier based on the transmission risk probability of each target sub-area; and sending the first warning information and the risk area distribution identifier to the target client.

[0019] By adopting the above technical solution, the first warning information is integrated with the risk area distribution logo, making the information transmission more comprehensive and intuitive, and facilitating relevant personnel to quickly and clearly grasp the overall risk distribution of the target area.

[0020] In a second aspect, an embodiment of the present application provides an AI-based viral pneumonia auxiliary diagnosis system, the AI-based viral pneumonia auxiliary diagnosis system comprising: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the AI-based viral pneumonia auxiliary diagnosis system to perform the method described in the first aspect and any possible implementation method of the first aspect.

[0021] In a third aspect, an embodiment of the present application provides a computer program product comprising instructions. When the above-mentioned computer program product is run on an AI-based viral pneumonia auxiliary diagnosis system, the above-mentioned AI-based viral pneumonia auxiliary diagnosis system performs the method described in the first aspect and any possible implementation method of the first aspect.

[0022] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions. When the instructions are run on an AI-based viral pneumonia auxiliary diagnosis system, the AI-based viral pneumonia auxiliary diagnosis system performs the method described in the first aspect and any possible implementation method of the first aspect.

[0023] It can be understood that the AI-based viral pneumonia auxiliary diagnosis system provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the method provided in the embodiment of the present application. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method, which will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0025] 1. This application can timely detect target people with abnormal body temperature and preset viral pneumonia symptoms by monitoring the abnormal status of people in the target area. When the target person's abnormal behavior exceeds the threshold within the preset time period, multiple data are obtained and mapped to the three-dimensional building model to obtain spatial environment related data, and then the transmission risk probability of each target sub-area is calculated through the transmission prediction model, high-risk areas are marked and early warning information is generated, comprehensively considering multiple factors such as personnel status, regional environment and personnel activity patterns, improving the accuracy of early warning information, and more accurately identifying potential transmission risk areas.

[0026] 2. This application obtains body temperature data through infrared temperature measurement equipment, obtains symptom characteristic data through video data analysis, and uses spatiotemporal correlation analysis to determine the target human body in abnormal state and record the number of abnormal behaviors. Infrared temperature measurement equipment can quickly obtain the body temperature of people, video data analysis can accurately capture preset symptoms, and spatiotemporal correlation analysis organically combines the two, effectively avoiding the limitations of single data monitoring, making the judgment of abnormal state of target population more comprehensive and accurate, so as to detect potential infected persons earlier and more accurately.

[0027] 3. This application divides the target sub-areas based on the three-dimensional building model to refine the analysis units, facilitate accurate positioning of risks, calculate the spatial distance and ventilation coverage between sub-areas, intuitively present the convenience of personnel flow and air circulation conditions, obtain temperature, humidity and airflow data, and obtain spatial environment correlation data after time-space mapping, weighted calculation and environmental correction of multi-source data, providing richer and more accurate input for the propagation prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flow chart of an AI-based auxiliary diagnosis method for viral pneumonia in an embodiment of the present application;

[0029] Figure 2 is another flowchart of the AI-based auxiliary diagnosis method for viral pneumonia in an embodiment of the present application;

[0030] Figure 3 It is a schematic diagram of the physical device structure of the AI-based viral pneumonia auxiliary diagnosis system in the embodiment of the present application. DETAILED DESCRIPTION

[0031] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification of the present application, the singular expressions "one", "a kind of", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more of the listed items.

[0032] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0033] For ease of understanding, the following is a description of the process of the method provided by this implementation. Figure 1, is a flow chart of the AI-based auxiliary diagnosis method for viral pneumonia in an embodiment of the present application.

[0034] S101. Monitor the target population in the target area for abnormal conditions, where the abnormal conditions include abnormal body temperature and preset viral pneumonia symptoms.

[0035] Among them, the target area refers to a specific area where auxiliary diagnosis of viral pneumonia is required, such as shopping malls, schools, stations and other crowded places, which are used to limit the monitoring scope. The target population refers to the people who are active in the target area and are the main subjects of monitoring. Abnormal conditions include abnormal body temperature, that is, the body temperature of the human body exceeds the normal range, which may be fever, etc., as well as preset viral pneumonia symptoms, such as coughing, shortness of breath and other symptoms commonly seen in viral pneumonia patients, which are used to determine whether a person may be infected with viral pneumonia.

[0036] Specifically, this step is continuously performed during the daily operation of the target area, and the target population is monitored in real time through monitoring equipment arranged in the target area. The temperature data of the target population is obtained non-contactly through infrared temperature measurement equipment, and the video data is analyzed by using a video monitoring system combined with image recognition technology to determine whether there are preset viral pneumonia symptoms in the target population, thereby obtaining symptom characteristic data. Then, the temperature data and the symptom characteristic data are analyzed in time and space to comprehensively judge the health status of the personnel, determine the target human body in an abnormal state, and record the number of abnormal behaviors within the preset time period, such as multiple coughing, continuous abnormal body temperature, etc.

[0037] In some embodiments, abnormal state monitoring can be achieved in a variety of ways: Optionally, high-precision infrared temperature measurement equipment is installed at key locations such as the entrance and passage of the target area, and the body temperature of people passing by is measured at regular intervals (such as 5 minutes), and the body temperature data is transmitted to the monitoring system in real time. At the same time, high-definition cameras are installed in public areas (such as halls, corridors, etc.) within the target area, and video data is analyzed in real time using deep learning-based image recognition algorithms to identify whether people have coughing, shortness of breath, and other actions, and extract relevant feature data; Optionally, wearable body temperature monitoring equipment is equipped for some people who often move in the target area (such as staff, long-term tenants, etc.), and body temperature data is uploaded in real time, and then combined with video surveillance data in the area for comprehensive analysis. It is understandable that other methods can also be used to achieve abnormal state monitoring, such as adding environmental monitoring equipment to assist in judgment, etc., which are not limited here.

[0038] S102: When it is detected that the target human body is in the abnormal state within a preset time period and the number of abnormal behaviors exceeds a preset threshold, a three-dimensional building model of the target area, a location where the abnormal state occurs, a movement trajectory of the target human body, and a real-time crowd density in the target area are obtained.

[0039] Among them, the preset time period is a pre-set length of time for counting the duration of the abnormal state of the target human body, such as 1 hour, 2 hours, etc., which is used to measure the stability of the abnormal state. The three-dimensional building model is a digital presentation of the building structure of the target area, including spatial information such as building layout, room distribution, and channel location, providing a spatial basis for subsequent analysis. The location of occurrence refers to the specific location in the target area where the target human body is located when the abnormal state occurs, which is used to locate the source of the abnormality. The movement trajectory is a record of the target human body's activity route in the target area, which helps to analyze its activity range and possible transmission paths. The real-time crowd density reflects the density of people in different locations in the target area, which is of great significance to the risk assessment of virus transmission.

[0040] Specifically, when the monitoring system detects that a target person is in an abnormal state for a preset period of time through continuous monitoring and data analysis, and the number of abnormal behaviors exceeds the preset threshold, the subsequent data acquisition process is started. The three-dimensional building model of the target area is retrieved from the pre-stored building information library, and the model can be generated by professional building surveying and mapping software. At the same time, the location of the abnormal state is determined based on the location information recorded by the monitoring equipment, and the movement trajectory of the target person is tracked using the positioning system in the target area (such as indoor positioning technology based on Wi-Fi or Bluetooth), and real-time crowd density data is obtained through the crowd counting equipment installed in the target area (such as infrared sensors, cameras, etc.).

[0041] In some embodiments, relevant data can be obtained in a variety of ways: Optionally, an interface is established with the property management system of the target area to directly obtain the three-dimensional building model data from its database. The video surveillance system in the target area is used in combination with image recognition technology to track and locate the target human body, determine the location where the abnormal state occurs and the movement trajectory, and obtain the real-time crowd density through statistical analysis of the number of people in the video screen; Optionally, the target area is scanned using a laser scanning sensor installed on the ceiling to obtain real-time crowd density data. It is understandable that other methods can also be used to obtain these data, such as using drones to survey and map building models, etc., which are not limited here.

[0042] S103: Map the occurrence location, the movement trajectory and the real-time crowd density to the three-dimensional building model to obtain spatial environment related data.

[0043] Among them, the occurrence location refers to the specific location coordinates where the target person appears abnormal, such as the entrance of a store in a shopping mall, a waiting area at a station, etc. It clarifies the location of the starting point of the abnormal situation in space. The movement trajectory is a record of the route taken by the target person in the target area. It consists of a series of continuous location points and can reflect the activity range and path direction of the target person. Real-time crowd density refers to the distribution of the density of people at each location in the target area at the current moment, usually expressed as the number of people per unit area, such as the number of people per square meter. The three-dimensional building model is a digital three-dimensional presentation of the building structure of the target area.

[0044] Specifically, after obtaining data such as the occurrence location, movement trajectory, and real-time crowd density, a data mapping operation is performed based on the three-dimensional building model of the target area. First, the target area is divided into multiple target sub-areas based on the three-dimensional building model, such as by floor, functional area, etc. Then the spatial distance between each target sub-area is calculated, which helps to analyze the difficulty of personnel flow and virus transmission, and at the same time obtain the ventilation coverage of each target sub-area. The ventilation situation has an important impact on virus transmission. Then the temperature and humidity data and airflow data in each target sub-area are obtained. These environmental factors will affect the survival and transmission ability of the virus. The occurrence location and movement trajectory are mapped with the spatial distance of each target sub-area in time and space to obtain the activity trajectory density of the target human body in each target sub-area, and then the activity trajectory density and the spatial distance are weighted to obtain the spatial position correlation degree, and the environmental factors are corrected based on the ventilation coverage, temperature and humidity and airflow data. Finally, the real-time crowd density and the corrected spatial position correlation degree are weighted and combined according to the preset weights to obtain the spatial environment correlation data.

[0045] In some embodiments, data mapping can be achieved in a variety of ways: Optionally, using geographic information system (GIS) technology, the three-dimensional building model is imported into the GIS platform, and the occurrence location, movement trajectory and real-time crowd density data are used as attribute data to be associated and mapped with each target sub-area in the building model. The distance and spatial relationship between the target sub-areas are calculated by the spatial analysis algorithm, and the temperature, humidity and airflow data obtained by the environmental monitoring equipment are combined to perform weighted calculation and environmental correction in the GIS platform to finally generate spatial environment related data; Optionally, a special data mapping software is developed, and a three-dimensional building model of the target area is first constructed in the software, divided into target sub-areas and the spatial relationship is defined. Then the occurrence location, movement trajectory and real-time crowd density data are imported, and parameters such as spatial distance and ventilation coverage are calculated according to the preset calculation rules, and the temperature, humidity and airflow data are obtained using the sensor data interface. Weighted calculation and environmental factor correction are completed in the software to obtain spatial environment related data. It can be understood that other methods can also be used to achieve this, such as using building information modeling (BIM) software for data integration and analysis, etc., which are not limited here.

[0046] S104: Input the spatial environment association data into a propagation prediction model to obtain the propagation risk probability of each target sub-area in the target area.

[0047] Among them, spatial environment-related data is a complex data set that integrates the location of occurrence, movement trajectory, real-time crowd density, three-dimensional building model spatial information, and environmental factors (temperature, humidity, airflow, etc.), which fully reflects the relationship between human activities and the spatial environment in the target area. The transmission prediction model is a mathematical model built based on a large amount of historical data and related algorithms. It is used to predict the risk of virus transmission in a specific environment. Its input is spatial environment-related data, and its output is the transmission risk probability of each target sub-area in the target area. The target sub-area is a smaller area unit divided according to certain rules in the target area, such as dividing a shopping mall into store areas, public passage areas, etc.

[0048] Specifically, after obtaining the spatial environment-related data, when it is necessary to evaluate the virus transmission risk in the target area, the spatial environment-related data is input into the pre-trained transmission prediction model. The transmission prediction model performs complex calculations and analyses on each target sub-area based on the input data, combined with its internal algorithms and parameters. The model considers the comprehensive impact of multiple factors such as the density of personnel activity trajectories, the density of human traffic, spatial distance, ventilation conditions, and environmental factors on the spread of the virus, thereby deriving the transmission risk probability for each target sub-area. For example, if a target sub-area has frequent personnel activities, high human traffic density, poor ventilation, and abnormal personnel activity trajectories passing through, its transmission risk probability may be high.

[0049] In some embodiments, the propagation prediction model can be used for risk prediction in a variety of ways: Optionally, a propagation prediction model based on a machine learning algorithm (such as a neural network) is used. First, a large amount of historical monitoring data, including data related to confirmed cases and environmental data, is collected to train the model. After the training is completed, the spatial environment-related data acquired in real time is preprocessed to meet the model input requirements, and then input into the model. The model outputs the propagation risk probability of each target sub-area through the calculation and layer-by-layer transmission of neurons; Optionally, a propagation prediction model based on statistical analysis is used. The probability of virus transmission under different combinations of factors is statistically calculated based on historical data, and a probability distribution model is established. The various parameters in the spatial environment-related data are substituted into the probability distribution model, and the probability values ​​corresponding to different parameter combinations are calculated to obtain the propagation risk probability of each target sub-area. It is understandable that other methods can also be used, such as a hybrid model combining deep learning and physical models, which is not limited here.

[0050] S105: Mark the target sub-region where the transmission risk probability is greater than a preset risk threshold as a target risk region, and generate first warning information including the target risk region and the target human body.

[0051] Among them, the transmission risk probability is a value calculated by the transmission prediction model to quantify the possibility of virus transmission in the target sub-area, ranging from 0 to 1. The larger the value, the higher the risk. For example, 0.8 has a greater transmission risk than 0.5. The preset risk threshold is a judgment standard set in advance, such as 0.7. Areas exceeding this value are considered high-risk. The target sub-area is a subdivision unit of the target area. The target risk area is a high-risk target sub-area. The first warning information is a prompt containing information about the target risk area and the target human body, notifying relevant personnel to take measures.

[0052] Specifically, after obtaining the transmission risk probability, the system immediately compares the threshold, and the target sub-area above the threshold is marked as the target risk area. At the same time, the system integrates the location range of the target risk area and the target human information (if identifiable), such as the time and location of the abnormal state, to generate the first warning information. For example, in a school scenario, if a classroom area is marked, the warning information will have the location of the classroom and the activity trajectory of related personnel, which is convenient for staff to handle.

[0053] In some embodiments, the implementation is as follows: Optionally, the monitoring system is equipped with a risk assessment module, which receives probability data and compares the threshold value. If the conditions are met, a visual mark is made in the three-dimensional model (such as color change and flashing), and the target human body information is extracted to generate an early warning according to the template, which is sent to the management personnel through SMS, email or pop-up window; Optionally, the intelligent algorithm analyzes the data, and when the threshold is exceeded, the early warning program is started, and the GIS map data is called to locate and mark the target risk area, and the early warning information containing the map screenshot and text description is generated in combination with the target human body trajectory, and sent to the designated mobile terminal or monitoring screen. Other methods such as voice alarm can also be combined, which are not limited here.

[0054] The following is a more detailed description of the process of the method provided by this implementation. Figure 2 , is another flow chart of the AI-based auxiliary diagnosis method for viral pneumonia in an embodiment of the present application.

[0055] S201. Acquire historical monitoring data containing confirmed cases, where the historical monitoring data includes a three-dimensional building model of the historical target area, the onset location of the confirmed case, the movement trajectory of the confirmed case, real-time crowd density data of the historical target area, and a record of the spread of the case in the historical target area.

[0056] Among them, historical monitoring data is a record of relevant information about the target area in the past period of time, which is used to analyze and build a transmission prediction model. The historical target area refers to a specific area where cases have occurred or monitoring has been carried out in the past, such as hospitals and communities where epidemics have occurred. The three-dimensional building model is a digital three-dimensional presentation of the building structure of the historical target area, including information such as building layout and spatial structure, such as the floor distribution and department layout of the hospital. The onset location is the specific location where the confirmed case first showed symptoms or was detected to be infected, such as the hospital ward number, the specific address in the community, etc. The movement trajectory is the route that the confirmed case took in the historical target area during the illness, such as the patient's walking path between different departments in the hospital, the daily activity route in the community, etc. The real-time crowd density data reflects the density of people at various locations in the historical target area at different times, such as the flow of people in the hospital lobby at different time periods. The transmission record is a record of the spread range and speed of the virus in the area, such as the number of people infected in the community and the specific area to which it spreads. It should be noted that the method in this embodiment obtains the written consent of the parties after fully respecting their privacy, explaining the purpose in detail to the parties and ensuring the security and confidentiality of the information, and then obtains the historical monitoring data of confirmed cases, covering three-dimensional building models, onset and movement trajectories, real-time crowd density, and transmission records, which are used to build a transmission prediction model.

[0057] Specifically, in order to build an effective transmission prediction model, historical monitoring data needs to be collected. Data can be obtained from relevant epidemic prevention and control databases, medical institutions' medical record systems, community management records and other channels. For the three-dimensional building model of the historical target area, it can be obtained by digitizing the architectural design drawings at that time or by professional surveying and mapping methods. The location of the disease is extracted from the medical record information of the confirmed case, and the movement trajectory can be determined based on the monitoring records of the hospital and the community's personnel activity survey. The real-time crowd density data is obtained through the records of the crowd flow monitoring equipment (such as infrared sensors, cameras, etc.) deployed in the area at that time, and the transmission records are compiled by the statistics and records of the health department during the epidemic prevention and control process. For example, taking the community where the epidemic once occurred as an example, the community's architectural planning drawings are obtained from the community management center to build a three-dimensional building model, the home address of the confirmed case at the time of the disease is found from the case file as the location of the disease, the patient's activity route in the community is tracked through the surveillance video in the community as the movement trajectory, the crowd density data of different time periods is obtained from the crowd flow counting equipment at the entrance of the community, and the number of infected people and the scope of transmission in the community are counted by the health department as the transmission record.

[0058] S202: Divide the historical target area into a plurality of historical target sub-areas based on the three-dimensional building model, and calculate the spatial distance and ventilation coverage between each of the historical target sub-areas.

[0059] Specifically, based on the acquired three-dimensional building model of the historical target area, professional geographic information system (GIS) software or building information model (BIM) tools are used to divide the area. According to the building structure and functional characteristics, the historical target area is divided into multiple reasonable historical target sub-areas. For example, in the hospital scene, the inpatient department can be divided into different sub-areas according to floors and departments. Then the spatial distance between each historical target sub-area is calculated through the measurement function of the software. For the ventilation coverage range, if there is a ventilation system design drawing, it can be determined based on the drawing information; if not, the ventilation coverage range can be determined by analyzing factors such as the vent location, wind direction, and wind speed through on-site testing or simulation software.

[0060] For example, in the processing of historical monitoring data of the hospital, the three-dimensional building model is imported into the GIS software, and the hospital is divided into different historical target sub-areas according to functional areas such as wards, outpatient clinics, corridors, etc. The distance measurement tool of the software is used to calculate the shortest travel distance between each sub-area as the spatial distance. For the ventilation coverage, the effective area range of fresh air supply in each sub-area is determined based on the hospital ventilation system design data.

[0061] S203, performing spatiotemporal mapping of the spatial distance between the onset location, the movement trajectory and each of the historical target sub-regions to obtain the activity trajectory density of the confirmed case in each of the historical target sub-regions.

[0062] Specifically, the location of the onset of the confirmed case is taken as the starting point, and its movement trajectory is decomposed in chronological order. For the location at each time point, the spatial distance to each historical target sub-area is calculated. Then, according to the relationship between time and space distance, the area and length of stay of the patient at different times are mapped to the corresponding historical target sub-area. The activity trajectory density is calculated by counting the number of activity trajectory points or the length of stay of the patient in each historical target sub-area, combined with the area or spatial size of the sub-area.

[0063] S204: performing correlation analysis on the activity trajectory density and the real-time crowd density data of the historical target area to obtain the transmission risk characteristics of each sub-area of ​​the historical target.

[0064] Specifically, for each historical target sub-area, the corresponding activity trajectory density data is matched and compared with the real-time crowd density data. If the activity trajectory density of a historical target sub-area is high, it means that the cases are more active in this area, and the real-time crowd density in this area is also high, then the possibility of virus transmission in this area is greater, and its transmission risk characteristics may be manifested as high transmission risk, rapid spread, etc. The transmission risk characteristics of each historical target sub-area are determined by statistics and calculations of a large number of data points, such as calculating the correlation coefficient between the two. Taking a shopping mall as an example, if the activity trajectory density of a certain area on a certain floor shows that the case has stayed there many times, and the real-time crowd density data in the area shows that the flow of people is frequent, then the transmission risk characteristics of the area are high risk, the virus is easy to spread and may spread quickly to surrounding areas.

[0065] Using data analysis software (such as the data analysis libraries Pandas and NumPy in Python), the activity trajectory density data and real-time crowd density data are integrated according to the historical target sub-areas. By writing algorithms, the degree of correlation between the two in time and space is calculated, such as calculating the consistency of their changing trends in different time periods, and then the transmission risk characteristics are determined.

[0066] S205: Input the propagation risk characteristics and the propagation situation records into a propagation risk initial model for training to obtain a propagation prediction model.

[0067] Specifically, the transmission risk characteristic data and corresponding transmission record data of each historical target sub-area are used as input samples and input into the initial transmission risk model. The model processes and learns the input data according to the built-in algorithm and structure. For example, the model analyzes the relationship between factors such as the density of human flow and the trajectory of case activity in the transmission risk characteristics and the number of infected people and the range of transmission in the transmission records, and continuously adjusts the parameters within the model to improve the accuracy of the prediction. Taking the historical monitoring data of the hospital as an example, the transmission risk characteristics of each ward area (such as personnel flow, case activity, etc.) and the actual virus transmission situation (such as the number of infected people in the ward, whether it has spread to adjacent wards, etc.) are input into the initial model. The model optimizes the parameters through multiple iterative calculations, and finally obtains a transmission prediction model that can accurately predict the probability of transmission risk in similar scenarios.

[0068] Use a machine learning framework (such as TensorFlow or PyTorch) to build an initial model of transmission risk. First, preprocess the transmission risk characteristics and transmission situation record data to make them meet the model input requirements, and then divide the data into training sets, validation sets, and test sets in a certain proportion. During the training process, the model parameters are continuously adjusted through optimization algorithms (such as gradient descent method), and the training strategy is adjusted according to the evaluation results of the validation set until the model achieves good performance indicators on the test set, and a transmission prediction model is obtained.

[0069] S206. Monitor the target population in the target area for abnormal conditions, where the abnormal conditions include abnormal body temperature and preset viral pneumonia symptoms.

[0070] This step specifically includes:

[0071] The body temperature data of the target population in the target area is obtained based on the infrared temperature measuring device.

[0072] Identify whether the target population in the video data in the target area has the preset viral pneumonia symptoms, and obtain symptom feature data;

[0073] The temperature data and the symptom characteristic data are subjected to a spatiotemporal correlation analysis to determine the target human body in the abnormal state in the target population, and the number of abnormal behaviors of the target human body within a preset time period is recorded.

[0074] Specifically, infrared temperature measurement equipment is deployed at key locations (such as entrances and passages) in the target area, and the equipment continuously monitors the body temperature of the target people passing by. When a person enters the monitoring range of the equipment, the equipment emits infrared rays and receives infrared rays reflected by the human body, and calculates the human body temperature based on the relationship between infrared energy and temperature, thereby obtaining body temperature data. For example, an infrared thermal imager is installed at the school gate. When students and faculty enter the school, the thermal imager automatically captures human thermal radiation information, quickly measures body temperature, and records body temperature data in real time. Video data is collected by multiple surveillance cameras arranged in the target area, and the video data is analyzed using deep learning-based image recognition technology. This technology uses a pre-trained neural network model to identify the behavior and expressions of people in the video and determine whether there are preset viral pneumonia symptoms. For example, the facial expressions and body postures of people in the video are analyzed to detect whether there are frequent coughing, body fluctuations during shortness of breath, etc., so as to obtain symptom feature data, and the body temperature data and symptom feature data of the same target person are matched and correlated according to the time sequence and spatial location (such as the specific activity area in the target area). If a target person has an abnormal body temperature at a certain moment, and the video data shows that the person has preset symptoms such as multiple coughs during the time period, and moves within a certain spatial range (such as a certain floor area in a shopping mall), then the target person is determined to be in an abnormal state. The system continuously records and counts the number of abnormal behaviors of the target person within the preset time period, such as coughing 5 times in 1 hour and abnormal body temperature 2 times, etc.

[0075] S207. When it is detected that the target human body is in the abnormal state within a preset time period and the number of abnormal behaviors exceeds a preset threshold, a three-dimensional building model of the target area, the location where the abnormal state occurs, the movement trajectory of the target human body, and the real-time crowd density in the target area are obtained.

[0076] It can be understood that this step is similar to step S102 and will not be described in detail here.

[0077] S208: Divide the target area into a plurality of target sub-areas based on the three-dimensional building model.

[0078] Among them, the 3D building model is a comprehensive and accurate digital 3D presentation of the building structure in the target area, covering the building's outline, internal layout (such as room distribution, channel direction) and spatial position relationship of each structure, which can be constructed through professional mapping software combined with field measurement data. The target area is a specific place where monitoring and analysis are currently being implemented, such as large shopping malls, office buildings, etc. The target sub-area is a smaller area unit after the target area is subdivided according to certain rules and needs. Its division helps to analyze the situation in different areas more accurately. For example, in a shopping mall, each floor and different functional areas can be divided into target sub-areas.

[0079] Specifically, taking a shopping mall as an example, its 3D building model is imported into a professional geographic information system (GIS) software or building information model (BIM) tool. According to the functional zoning, personnel flow characteristics and management needs of the shopping mall, the segmentation function in the software is used to divide the shopping mall into multiple target sub-areas, such as dividing a floor into multiple store areas, public passage areas, etc. The key segmentation points or lines can be marked in the 3D building model, such as the dividing lines between floors, the boundary lines of different functional areas, etc., and then the area division function of the software is used to divide the model into multiple target sub-areas along these marked lines, and each sub-area is given a unique identifier to facilitate data management and call.

[0080] S209: Calculate the spatial distance and ventilation coverage between each target sub-area.

[0081] Among them, spatial distance refers to the actual geometric distance between each target sub-area in three-dimensional space, which can be divided into straight-line distance (the shortest connecting line length between two points) and actual travel distance (the walking distance considering the building structure and channel restrictions), such as the straight-line distance between the centers of two store areas or the shortest path length along the channel. Ventilation coverage refers to the area within the target area where the ventilation system (such as fresh air system, air conditioning ventilation ducts, etc.) can effectively regulate air circulation and exchange. Its size and shape are affected by the layout, power and building structure of the ventilation equipment.

[0082] Specifically, for spatial distance calculation, the software's measurement tools are used in the three-dimensional building model to select the characteristic points of the target sub-area (such as the center of the area, the entrance point, etc.), measure the straight-line distance between these points, or calculate the actual travel distance based on the building channel information. Taking an office building as an example, the distance between the stairs or elevator passages on different floors of the office area is calculated. For ventilation coverage, if there are ventilation system design drawings, it can be determined based on the ventilation equipment location, air outlet distribution, and wind direction simulation information in the drawings; if there are no drawings, field tests can be conducted, such as releasing smoke at the vents or using air velocity meters to detect air flow at different locations, so as to determine the ventilation coverage. Spatial distance calculation can be automated scripted to batch select characteristic points of the target sub-area and calculate the distance, and store the results in the database. The ventilation coverage can be determined using computational fluid dynamics (CFD) simulation software, which can be simulated and analyzed based on the building structure and ventilation equipment parameters to obtain a visual ventilation coverage map, and extract relevant data for subsequent analysis.

[0083] S210: Acquire temperature, humidity and airflow data in each target sub-area.

[0084] Among them, temperature and humidity data refers to the numerical information of the temperature and humidity of the air in the target sub-area. Temperature affects the survival and transmission speed of the virus, and humidity affects the stability of virus transmission, such as 25°C and 60% relative humidity. Airflow data refers to information such as the speed and direction of air flow in the target sub-area. Airflow will affect the diffusion path of the virus. For example, the airflow speed in a certain area is 0.5 m / s and the direction is from east to west. Specifically, temperature and humidity sensors and airflow sensors are arranged in each target sub-area. The temperature and humidity sensor converts the physical signal into an electrical signal by sensing the changes in temperature and humidity in the air, and then obtains the temperature and humidity data in digital form through analog-to-digital conversion. The airflow sensor uses hot wire, impeller and other principles to measure the airflow speed and direction, which are also converted into digital data.

[0085] S211, performing spatiotemporal mapping of the occurrence position, the movement trajectory and the spatial distance of each target sub-region to the activity trajectory density of the target human body in each target sub-region.

[0086] Among them, the occurrence location refers to the specific location coordinates where the target human body is located when an abnormal state first occurs (such as abnormal body temperature or preset viral pneumonia symptoms), such as the location coordinates of a store entrance in a shopping mall. The movement trajectory is a route record formed by connecting a series of location points that the target human body passes through in the target area, such as the walking route from the entrance to each store and then to the exit in a shopping mall. The spatial distance is the geometric distance between the target sub-area and the occurrence location and each point on the movement trajectory. Spatiotemporal mapping combines time and space information to associate the changes in the occurrence location and movement trajectory in the time dimension with the spatial position of the target sub-area. The activity trajectory density is used to measure the frequency of the target human body's activities in each target sub-area, expressed as the number of activity trajectory points per unit area or unit space, such as the number of activity trajectory points per square meter.

[0087] Specifically, taking the shopping mall scene as an example, when the target human body shows an abnormal state in the shopping mall, the location where it occurs (such as the elevator entrance on a certain floor) is determined. Then, according to its subsequent movement trajectory (the position change sequence obtained by positioning technology), the spatial distance between the location and each point on the movement trajectory and each target sub-area (such as each store area, public rest area, etc.) is calculated. In chronological order, the relationship between the target human body at different positions at different times and the target sub-area is mapped. For example, the target human body passes through 3 target sub-areas within 10 minutes, and stays in one of the areas for a long time and has more activity trajectory points, then the activity trajectory density of this area is relatively high. By counting the number of activity trajectory points in each target sub-area and combining the area of ​​the area, the activity trajectory density is calculated, and the movement trajectory data of the target human body is obtained using a positioning system (such as indoor positioning based on Wi-Fi or Bluetooth), and it is imported into the geographic information system (GIS) software together with the three-dimensional building model of the target area and the information of the divided target sub-areas. In GIS, spatial analysis functions are used to calculate the location of occurrence, the distance between the moving trajectory points and the target sub-area, mark the trajectory points according to time sequence, count the number of trajectory points in each target sub-area, and then divide it by the area to obtain the activity trajectory density.

[0088] S212: Perform weighted calculation on the activity trajectory density and the spatial distance of each target sub-region to obtain a spatial position correlation.

[0089] The distance between the relevant location (location of occurrence, moving trajectory point). Weighted calculation is a method of calculating after assigning corresponding weights according to the importance of different factors. Here, the influence of activity trajectory density and spatial distance on the correlation degree of virus transmission risk is considered. Spatial position correlation is a quantitative indicator of the closeness of the spatial relationship between the target human body and the target sub-area obtained by combining the activity trajectory density and spatial distance. The larger the value, the closer the correlation is and the higher the transmission risk may be. Specifically, for each target sub-area, the corresponding activity trajectory density and spatial distance data are weighted. For example, if it is believed that the activity trajectory density has a greater impact on the risk of virus transmission, a higher weight (such as 0.7) can be assigned and a lower weight (such as 0.3) can be assigned to the spatial distance. The calculation method is spatial position correlation = activity trajectory density × weight 1 + spatial distance × weight 2. Assuming that the activity trajectory density of a target sub-area is high and the spatial distance is close, the spatial position correlation obtained after weighted calculation will be high, indicating that the area is closely related to the target human body and the risk of virus transmission is relatively high. In terms of implementation, in data processing software (such as the data analysis library Pandas in Python), the activity trajectory density and spatial distance data are read, and weighted summation calculation is performed according to the pre-set weight value to obtain the spatial position correlation of each target sub-area, and the result is stored in a new data column for subsequent analysis and processing.

[0090] S213: Based on the ventilation coverage, temperature and humidity data, and airflow data of each target sub-area, the spatial position correlation is corrected for environmental factors.

[0091] Among them, the ventilation coverage is the area within the target sub-area where the ventilation system is effective, which affects the spread and dilution of the virus. Temperature and humidity data include temperature and humidity values, which affect the survival and transmission characteristics of the virus. Airflow data is the speed and direction information of air flow, which determines the transmission path of the virus in the air. Environmental factor correction is to adjust the spatial position correlation according to environmental factors such as ventilation, temperature and humidity, and airflow, so that it can more accurately reflect the actual risk of virus transmission. The spatial position correlation is the correlation degree indicator calculated in the previous step without considering environmental factors.

[0092] Specifically, for each target sub-area, obtain its ventilation coverage, temperature and humidity data, and airflow data. If a target sub-area is well ventilated, has a wide ventilation coverage, has a fast airflow speed and a direction that is conducive to air circulation, and has a suitable temperature and humidity for virus survival and transmission, then these environmental factors will reduce the transmission risk of the area, and the spatial position correlation will be corrected accordingly. For example, in a well-ventilated shopping mall lobby area, the spatial position correlation may be reduced due to good ventilation conditions. Conversely, in areas with poor ventilation, suitable temperature and humidity for viruses, and poor airflow, the spatial position correlation will increase. By establishing a correction model or rule for the correlation between environmental factors and spatial positions, adjusting the correlation according to environmental data, a relationship table or mathematical model between environmental factors and transmission risk correction coefficients can be established based on a large amount of experimental data or historical case analysis. In the data processing process, the relationship table is queried or substituted into the model to calculate the correction coefficient based on the environmental data of the target sub-area, and then the spatial position correlation is multiplied by the correction coefficient to obtain the corrected spatial position correlation. For example, a machine learning algorithm is used to train a model to predict the transmission risk correction coefficient based on environmental factors, with ventilation, temperature and humidity, and airflow data as input and the correction coefficient as output to correct the spatial location correlation.

[0093] S214: Perform a weighted combination of the real-time crowd density of each target sub-area and the spatial position correlation based on a preset weight to obtain spatial environment correlation data.

[0094] The real-time crowd density refers to the number of people per unit area or unit space in the target sub-area at the current moment, such as 5 people per square meter. It reflects the degree of crowd gathering. The denser the crowd, the higher the possibility of virus transmission. The spatial position correlation is a quantitative indicator that combines factors such as the density of the target human activity trajectory and the spatial distance of the target sub-area. It indicates the degree of closeness between the target human body and the target sub-area in space and the degree of correlation of transmission risk. For each target sub-area, its real-time crowd density data and the calculated spatial position correlation are obtained. According to the preset weight, the real-time crowd density is multiplied by its weight value, and the spatial position correlation is multiplied by the corresponding weight value, and then the two products are added to obtain the spatial environment correlation data of the target sub-area. For example, in a target sub-area in the waiting hall of the station, the real-time crowd density is high, assuming 8 people per square meter, and the spatial position correlation is calculated to be 0.6 (assumption). If the preset real-time crowd density weight is 0.4 and the spatial position correlation weight is 0.6, then the spatial environment correlation data = 8×0.4+0.6×0.6=3.2+0.36=3.56.

[0095] S215: Input the spatial environment association data into a propagation prediction model to obtain the propagation risk probability of each target sub-area in the target area.

[0096] It can be understood that this step is similar to step S104 and will not be described in detail here.

[0097] S216: Mark the target sub-region where the transmission risk probability is greater than the preset risk threshold as a target risk region, and generate first warning information including the target risk region and the target human body.

[0098] It can be understood that this step is similar to step S105 and will not be described in detail here.

[0099] S217: When the target area is a transportation station, the activity trajectory of the target person is tracked to obtain the waiting area and entry time of the target person.

[0100] Among them, transportation stations are public places where people gather and disperse, such as railway stations, bus stations, subway stations, etc., which have the characteristics of large and complex personnel flow. The target person is a person who is monitored in an abnormal state (abnormal body temperature and / or has preset viral pneumonia symptoms) in the transportation station. Activity trajectory tracking is the use of a variety of technical means (such as surveillance cameras, positioning systems, etc.) to record the target person's route of action in the transportation station. The waiting area is the specific location where the target person waits for the bus in the transportation station, such as a specific waiting area in a waiting hall of a railway station. The entry time point is the exact moment when the target person enters the transportation station to prepare for the bus, such as 10:30 on October 1, 2023.

[0101] Specifically, surveillance cameras widely distributed in transportation stations form a surveillance network. When the target person enters the transportation station, the camera continues to capture his or her figure. The target person is identified and tracked through image recognition technology, and the waiting area is determined in combination with the area division information in the station (such as the signs of different waiting halls, platforms, etc.). At the same time, the monitoring system records the time when the target person passes through security checks or enters a specific area (such as the entrance to the waiting hall) as the entry time point. For example, in a subway station, the target person enters from the subway station entrance, and the camera records the time when passing through the gate, and then waits for the train on the platform. The monitoring system determines that the platform where the target person is located is the waiting area based on his or her route of action, and records the time information of the entire process.

[0102] In terms of implementation, a high-performance image recognition server is deployed in the monitoring system of the traffic station, and the target human recognition model trained by the deep learning algorithm is used to analyze the monitoring video in real time. The map information and regional division data of the station are imported into the system. When the target human body is recognized, the waiting area and the time of entry are determined according to its position changes in different camera images and combined with the map information, and these data are stored in the database for subsequent query and use.

[0103] S218. Obtain the train number corresponding to the waiting area and the entry time point.

[0104] The waiting area is the specific location where the target person is waiting for the train. Different waiting areas usually correspond to different train running directions or routes. The entry time point is the time when the target person enters the transportation station to prepare for the train, which is associated with the train's running schedule. The train schedule is a train with a specific number, running route and departure time, such as the G123 train.

[0105] Specifically, the management system of the transportation station stores detailed train schedules and station layout information. According to the waiting area of ​​the target person, the direction or line of the train that the person may take can be determined. Combined with the time of entering the station, the train schedule is searched for trains near that time point and in line with the direction or line. For example, at a railway station, the target person is in the waiting area for the train to Beijing and enters the station at 9 am. By querying the schedule, the train schedule departing for Beijing between 9 am and 10 am is found, such as the G56 train.

[0106] In terms of implementation, the train schedule data of transportation stations is stored in the database, and a data interface with the monitoring system and early warning system is established. After obtaining the waiting area and entry time of the target person, the train schedule information that meets the conditions is screened out from the database by writing a query program. At the same time, the data can be updated in real time to ensure that the obtained train schedule information is accurate, so as to meet the needs of accurate management of train transmission risks in epidemic prevention and control.

[0107] S219: Generate second warning information including the train schedule and the target person.

[0108] Specifically, the obtained train schedule information and relevant information of the target person (such as identity information, if available; abnormal status, etc.) are integrated. The format of the warning information can be a structured data format or a clear and concise text format. For example, the content of the warning information may be: A person in an abnormal state [name (if known)] is found at [entry time] in [waiting area] preparing to take [train schedule]. Please take prevention and control measures in time. Such information allows the recipient to quickly understand key information and make corresponding decisions.

[0109] S220. Send the second warning information to the target train client corresponding to the train schedule.

[0110] Specifically, after generating the second warning information, the information management system of the transportation station first establishes a connection through the interface of the railway communication network, which can cover the entire railway transportation system to ensure the reliability and timeliness of information transmission. Then, according to the train schedule information, the network address or identification information of the target train client corresponding to the train schedule is queried in the railway operation management database. After determining the target train client, the warning information is encoded and encapsulated and sent out in a format suitable for railway communication network transmission (such as a specific message protocol). For example, at a high-speed railway station, when it is detected that an abnormal person may take the G123 train, the station system finds the client on the G123 train through the railway dedicated communication network and sends the warning information. After receiving it, the train staff can make prevention and control preparations in advance, such as disinfecting the relevant carriages and strengthening the health monitoring of passengers.

[0111] In some embodiments, the following steps may be further included after step S220:

[0112] Based on the second warning information and the activity trajectory of the target person, a risk level assessment is performed on the carriage corresponding to the train service to obtain a risk level of the carriage;

[0113] Generate ventilation parameters corresponding to each compartment according to the compartment risk level, and generate ventilation control prompt information including the ventilation parameters;

[0114] The ventilation control prompt information is sent to the target train client.

[0115] Specifically, the train schedule and related conditions that the target person may take are determined based on the second warning information, and then the area and stay time that the target person passes through in the transportation station are analyzed in combination with the activity trajectory. If the target person has a wide range of activities in the station, stays for a long time, and is close to the boarding gates of multiple carriages, then these carriages are more likely to be affected. By setting evaluation rules, such as assigning different weights to factors such as the proximity of the activity trajectory coverage area to the carriage and the length of stay, the risk level of the carriage is obtained. For example, if the target person passes near the boarding gates of carriages 3-5 in the station and stays for more than 10 minutes, these carriages may be rated as high risk levels. Using data analysis software (such as the data processing library in Python), the warning information and activity trajectory data are integrated and analyzed. According to the pre-set risk assessment algorithm (such as a weighted calculation model based on distance and time), the risk score of each carriage is calculated, and the risk level is divided according to the score range, such as 80-100 points for high risk, 50-79 points for medium risk, and 0-49 points for low risk.

[0116] Set corresponding ventilation parameter standards for different carriage risk levels. For high-risk carriages, in order to quickly dilute the possible viruses, set the ventilation volume to a larger value, such as 15 air changes per hour, and adopt a fresh air mode to ensure that the introduced air is fresh and pollution-free. For medium-risk carriages, the ventilation volume can be set to 10 air changes per hour, using a mixed wind mode to partially introduce fresh air. For low-risk carriages, the ventilation volume is maintained at a normal level, such as 6 air changes per hour. Generate ventilation control prompt information based on these set ventilation parameters. For example, carriage No. 3 of the G56 train is a high-risk carriage, and the ventilation parameters are: ventilation volume 15 times per hour, fresh air mode. Please ask the staff to adjust in time and establish a mapping relationship table between ventilation parameters and risk levels in the train ventilation control system. Query the table according to the carriage risk level to obtain the corresponding ventilation parameters, and then use the information generation template to generate ventilation control prompt information.

[0117] The system establishes a connection with the target train client through the railway communication network. The ventilation control prompt information is encoded and encapsulated, and sent to the target train client in a format suitable for communication network transmission. For example, using a railway-specific data transmission protocol (such as the GSM-R protocol), the ventilation control prompt information is packaged into a data packet in a specific format and sent to the designated receiving port of the train client. After receiving the information, the train client parses the data packet, extracts the ventilation parameters and operating instructions, and then controls the carriage ventilation system to adjust as required to ensure that the air circulation in the carriage achieves the expected effect and reduces the risk of virus transmission.

[0118] In some embodiments, the following steps may be further included after step S216:

[0119] Generate a risk area distribution identifier according to the transmission risk probability of each target sub-area;

[0120] The first warning information and the risk area distribution identifier are sent to a target client.

[0121] Specifically, according to the transmission risk probability of each target sub-area, a corresponding identification rule is set. If the transmission risk probability is greater than 0.7, the risk area distribution mark of the target sub-area is set to red (or coded as 1), indicating high risk; if the probability is between 0.3-0.7, it is set to yellow (or coded as 0.5), indicating medium risk; if it is less than 0.3, it is set to green (or coded as 0), indicating low risk. For example, in a shopping mall scenario, the transmission risk probability of a clothing sales area is 0.8, and its risk area distribution mark is red, while the transmission risk probability of a bathroom area is 0.2, and the mark is green. In this way, the risk distribution in the target area can be quickly and intuitively presented. First, the communication address or identification information of the target client is determined, which can be obtained by pre-registration or in the system configuration. Then the first warning information and the risk area distribution mark are integrated and packaged in a format suitable for the target client to receive and process (such as JSON format). The packaged information is sent to the target client through a network communication protocol (such as TCP / IP protocol).

[0122] The following describes the AI-based viral pneumonia auxiliary diagnosis system in the embodiment of the present invention from the perspective of hardware processing. Figure 3 , which is a schematic diagram of a physical device structure of an AI-based viral pneumonia auxiliary diagnosis system in an embodiment of the present application.

[0123] It should be noted that Figure 3 The structure of the AI-based viral pneumonia auxiliary diagnosis system shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0124] like Figure 3As shown, the AI-based viral pneumonia auxiliary diagnosis system includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage part 308 to the random access memory (RAM) 303, such as executing the method described in the above embodiment. In RAM 303, various programs and data required for system operation are also stored. CPU 301, ROM 302 and RAM 303 are connected to each other through bus 304. Input / output (I / O) interface 305 is also connected to bus 304.

[0125] The following components are connected to the I / O interface 305: an input section 306 including an audio input device, a button switch, etc.; an output section 307 including a liquid crystal display (LCD) and an audio output device, an indicator light, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed so that a computer program read therefrom is installed into the storage section 308 as needed.

[0126] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 309, and / or installed from a removable medium 311. When the computer program is executed by the central processing unit (CPU) 301, various functions defined in the present invention are performed.

[0127] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.

[0128] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from that marked in the accompanying drawings.

[0129] Specifically, the AI-based viral pneumonia auxiliary diagnosis system of this embodiment includes a processor and a memory, and a computer program is stored in the memory. When the computer program is executed by the processor, the AI-based viral pneumonia auxiliary diagnosis method provided in the above embodiment is implemented.

[0130] As another aspect, the present invention also provides a computer-readable storage medium, which may be included in the AI-based viral pneumonia auxiliary diagnosis system described in the above embodiment; or it may exist independently without being assembled into the AI-based viral pneumonia auxiliary diagnosis system. The above storage medium carries one or more computer programs, and when the above one or more computer programs are executed by a processor of the AI-based viral pneumonia auxiliary diagnosis system, the AI-based viral pneumonia auxiliary diagnosis system implements the AI-based viral pneumonia auxiliary diagnosis method provided in the above embodiment.

[0131] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, 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 application.

[0132] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0133] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. An AI-based auxiliary diagnosis method for viral pneumonia, characterized in that: Applied to an AI-based viral pneumonia auxiliary diagnosis system, the method comprises: Obtain body temperature data of target people in the target area based on infrared temperature measurement equipment; Identify whether the target population in the video data in the target area has preset viral pneumonia symptoms, and obtain symptom feature data; Performing spatiotemporal correlation analysis on the body temperature data and the symptom characteristic data, determining the target human body in an abnormal state among the target population, and recording the number of abnormal behaviors of the target human body within a preset time period, wherein the abnormal state includes abnormal body temperature and preset viral pneumonia symptoms; When it is detected that the target person is in the abnormal state within a preset time period and the number of abnormal behaviors exceeds a preset threshold, a three-dimensional building model of the target area, the location where the abnormal state occurs, the movement trajectory of the target person, and the real-time crowd density in the target area are obtained; Dividing the target area into a plurality of target sub-areas based on the three-dimensional building model; Calculating the spatial distance and ventilation coverage between each of the target sub-areas; Acquiring temperature, humidity and airflow data in each of the target sub-areas; Performing spatiotemporal mapping of the occurrence position, the movement trajectory and the spatial distance of each target sub-region to the activity trajectory density of the target human body in each target sub-region, wherein the activity trajectory density is calculated by dividing the number of trajectory points in each target sub-region by the area of ​​the target sub-region; Performing weighted calculation on the activity trajectory density and the spatial distance of each target sub-region to obtain a spatial position correlation degree; Based on the ventilation coverage, temperature and humidity data, and airflow data of each target sub-area, the spatial position correlation is corrected for environmental factors; Based on a preset weight, the real-time crowd density of each target sub-area is weighted and combined with the spatial position correlation degree to obtain spatial environment correlation data; Inputting the spatial environment association data into a propagation prediction model to obtain a propagation risk probability for each target sub-area in the target area; The target sub-region with a transmission risk probability greater than a preset risk threshold is marked as a target risk region, and first warning information including the target risk region and the target human body is generated.

2. The method according to claim 1, characterized in that Before the step of obtaining the body temperature data of the target population in the target area based on the infrared temperature measuring device, the method further includes: Acquire historical monitoring data containing confirmed cases, wherein the historical monitoring data includes a three-dimensional building model of a historical target area, the onset location of the confirmed case, the movement trajectory of the confirmed case, real-time crowd density data of the historical target area, and a record of the spread of the case in the historical target area; and dividing the historical target area into a plurality of historical target sub-areas based on the three-dimensional building model, and calculating the spatial distance and ventilation coverage between each of the historical target sub-areas; Performing spatiotemporal mapping on the spatial distance between the onset location, the movement trajectory and each of the historical target sub-regions to obtain the activity trajectory density of the confirmed cases in each of the historical target sub-regions; Correlation analysis is performed on the activity trajectory density and the real-time crowd density data of the historical target area to obtain the transmission risk characteristics of each historical target sub-area; The transmission risk characteristics and the transmission situation records are input into the transmission risk initial model for training to obtain a transmission prediction model.

3. The method according to claim 1, characterized in that After the step of marking the target sub-region with a transmission risk probability greater than a preset risk threshold as a target risk region and generating first warning information including the target risk region and the target human body, the method further includes: When the target area is a traffic station, the activity trajectory of the target person is tracked to obtain the waiting area and entry time of the target person; Obtaining the train schedule corresponding to the waiting area and the entry time point; generating second warning information including the train schedule and the target person; The second warning information is sent to the target train client corresponding to the train schedule.

4. The method according to claim 3, characterized in that: After the step of sending the second warning information to the target train client corresponding to the train schedule, the method further includes: Based on the second warning information and the activity trajectory of the target person, a risk level of the carriage corresponding to the train schedule is evaluated to obtain a risk level of the carriage; Generating ventilation parameters corresponding to each compartment according to the compartment risk level, and generating ventilation control prompt information including the ventilation parameters; The ventilation control prompt information is sent to the target train client.

5. The method according to claim 1, characterized in that After the step of marking the target sub-region with a transmission risk probability greater than a preset risk threshold as a target risk region and generating first warning information including the target risk region and the target human body, the method further includes: Generating a risk area distribution identifier according to the transmission risk probability of each target sub-area; The first warning information and the risk area distribution identifier are sent to a target client.

6. An AI-based viral pneumonia auxiliary diagnosis system, characterized in that: The AI-based viral pneumonia auxiliary diagnosis system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the AI-based viral pneumonia auxiliary diagnosis system to perform the method described in any one of claims 1-5.

7. A computer-readable storage medium comprising instructions, characterized in that: When the instruction is executed on the AI-based viral pneumonia auxiliary diagnosis system, the AI-based viral pneumonia auxiliary diagnosis system executes the method as described in any one of claims 1-5.

8. A computer program product, characterized in that When the computer program product runs on an AI-based viral pneumonia auxiliary diagnosis system, the AI-based viral pneumonia auxiliary diagnosis system executes the method as described in any one of claims 1-5.

Images