A method and system for automatically distributing early warning events for smart cities

Abstract

The present invention discloses a method and system for automatically distributing early warning events for smart cities, and belongs to the field of smart city technology. The present invention obtains smart IoT sensor monitoring data to be analyzed for early warning, and performs attention processing on it to effectively extract key information, thereby forming sensor monitoring attention data. Then, the debugged AI decision analysis algorithm is used to deeply analyze the target abnormal event label information in the monitoring data, which not only reveals the multi-modal state trend of the event, but also evaluates the city interface energy level, so as to fully grasp the development trend and impact range of the abnormal event. Through characterization integration, an abnormal event linkage attribute vector is innovatively constructed, and this comprehensive indicator provides a scientific basis for early warning distribution. Finally, based on rich data support and precise vector analysis, the early warning distribution keywords are accurately marked, and efficient early warning event distribution processing is performed accordingly, which greatly improves the speed and accuracy of emergency response.

Description

Technical Field

[0001] The present invention belongs to the technical field of smart cities, and in particular relates to a method and system for automatically distributing early warning events for smart cities. Background Art

[0002] In the current process of smart city construction, the widespread application of IoT sensor technology has greatly enriched the data foundation of urban management. However, faced with massive, multi-dimensional sensor monitoring data, how to quickly and accurately identify potential abnormal events and issue timely and effective warnings has become a technical problem that needs to be solved urgently. Traditional early warning systems often rely on manually set thresholds or simple pattern recognition, which is difficult to adapt to the complex and changing urban operating environment and the growing data scale. Therefore, there is an urgent need for a technical solution that can automatically process smart IoT sensor monitoring data, intelligently analyze abnormal event trends, and accurately issue warnings. Summary of the invention

[0003] The present invention provides a method and system for automatically distributing warning events for smart cities, which can solve or partially solve the technical problems involved in the above-mentioned background technology.

[0004] The embodiment of the present invention provides an automatic dispatching method for warning events for smart cities, which is applied to a warning event dispatching system. The method includes: acquiring smart IoT sensor monitoring data to be subjected to warning analysis; performing attention processing on the smart IoT sensor monitoring data to obtain sensor monitoring attention data; calling a debugged AI decision analysis algorithm to respectively analyze the multimodal state trend and the city interface energy level of target abnormal event label information in the smart IoT sensor monitoring data to obtain a target multimodal state trend representation and a target city interface energy level representation; integrating the target multimodal state trend representation and the target city interface energy level representation to obtain an abnormal event linkage attribute vector; annotating the warning dispatching keywords of the target abnormal event label information based on the smart IoT sensor monitoring data, the sensor monitoring attention data and the abnormal event linkage attribute vector to obtain an annotation result, and performing warning event dispatching processing based on the annotation result.

[0005] An embodiment of the present invention provides a warning event dispatching system, comprising at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the above method.

[0006] An embodiment of the present invention provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the above method are implemented.

[0007] The present invention obtains the smart IoT sensor monitoring data to be analyzed for early warning, processes it with attention, effectively extracts key information, and forms sensor monitoring attention data. This step significantly enhances the system's sensitivity to abnormal signals, making subsequent analysis more accurate. Next, the debugged AI decision analysis algorithm is used to deeply analyze the target abnormal event label information in the monitoring data, which not only reveals the multimodal state trend of the event, but also evaluates the urban interface energy level, thereby fully grasping the development trend and impact range of the abnormal event. By integrating these two aspects of characterization, the present invention innovatively constructs an abnormal event linkage attribute vector, and this comprehensive indicator provides a scientific basis for early warning distribution. Finally, based on rich data support and accurate vector analysis, the present invention accurately marks the early warning distribution keywords, and performs efficient early warning event distribution processing accordingly, greatly improving the speed and accuracy of emergency response. In summary, the rapid identification, accurate early warning and effective distribution of abnormal events are achieved, providing strong support for urban safety management and disaster prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A flowchart of a method for automatically distributing warning events for smart cities provided by an embodiment of the present invention.

[0009] Figure 2 A schematic diagram of the structure of a warning event distribution system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0010] Figure 1 A method for automatically distributing warning events for smart cities is shown, which is applied to a warning event distribution system. The method includes the following steps 801 to 803.

[0011] Step 801: The warning event dispatching system obtains the smart IoT sensor monitoring data to be subjected to warning analysis; performs attention processing on the smart IoT sensor monitoring data to obtain sensor monitoring attention data.

[0012] Step 802: The warning event dispatching system calls the debugged AI decision analysis algorithm to analyze the multimodal state trend and the city interface energy level of the target abnormal event label information in the smart IoT sensor monitoring data, respectively, to obtain the target multimodal state trend representation and the target city interface energy level representation; the target multimodal state trend representation and the target city interface energy level representation are integrated to obtain the abnormal event linkage attribute vector.

[0013] Step 803: The warning event distribution system labels the warning distribution keywords of the target abnormal event label information based on the smart IoT sensor monitoring data, the sensor monitoring attention data and the abnormal event linkage attribute vector, obtains the labeling results, and performs warning event distribution processing based on the labeling results.

[0014] The early warning event distribution system in the embodiment of the present invention is an important part of modern urban management. Its core lies in the efficient integration and analysis of monitoring data from the smart IoT sensor network to achieve rapid and accurate abnormal event warning and response. The following will elaborate on how the system implements the technical solution through three specific application scenarios of traffic management, environmental protection and disaster warning to ensure the safety and efficiency of urban operation.

[0015] 1. Traffic Management Scenario

[0016] In the traffic management scenario, the early warning event dispatching system first obtains real-time monitoring data from smart IoT devices such as road monitoring cameras, traffic flow sensors, and vehicle GPS. These data cover key information such as vehicle flow, vehicle speed, road occupancy rate, and traffic accident occurrence. The system then processes these data and uses deep learning models to identify key features in the data, such as sudden increases in vehicle flow and abnormally slow vehicle speeds, to form sensor monitoring attention data.

[0017] Next, the system calls the pre-debugged AI decision analysis algorithm to deeply analyze the data after attention processing. On the one hand, the algorithm analyzes the multimodal state trend of traffic flow, such as the changing rules of morning and evening peaks, special flow patterns during holidays, etc. On the other hand, it evaluates the energy level of the urban road interface, that is, the ability of different road sections to carry traffic flow and its saturation. By integrating these two aspects of representation, the system generates an abnormal event linkage attribute vector, such as a road section is about to reach the congestion threshold or a traffic accident has occurred.

[0018] Finally, based on the smart IoT sensor monitoring data, sensor monitoring attention data and abnormal event linkage attribute vectors, the system accurately labels the warning distribution keywords of target abnormal events (such as traffic congestion and traffic accidents). For example, "XX section of road is expected to reach a serious congestion state in half an hour, please dispatch immediately", "A traffic accident occurred at XX intersection, and emergency rescue is required". These labeling results directly guide the emergency response of the traffic management department, realize the rapid distribution of processing instructions, effectively alleviate traffic pressure, and ensure smooth roads.

[0019] 2. Environmental Protection Scenarios

[0020] In the field of environmental protection, the early warning event distribution system also plays a key role. The system first collects real-time data from sensors such as air quality monitoring stations, water quality monitoring stations, and noise monitoring equipment. These data reflect environmental indicators such as air quality index, water pollutant concentration, and noise decibel level. Through the attention processing mechanism, the system screens out monitoring values ​​that exceed the normal range, such as a sudden increase in PM2.5 concentration and excessive levels of specific chemical substances, to form sensor monitoring attention data.

[0021] Subsequently, AI decision analysis algorithms were used to analyze these data, identify multi-modal state trends of environmental indicators, such as seasonal air quality changes, water quality improvement trends after rain, etc., and analyze the environmental carrying capacity of different areas of the city, that is, the relationship between the environment's self-recovery capacity and the current pollution pressure. Combining these two aspects of information, the system constructs an abnormal event linkage attribute vector, pointing to possible environmental crises, such as increased air pollution and threats to water sources.

[0022] Based on the above analysis, the system marks the early warning events in the field of environmental protection with keywords, such as "the air quality in the XX area has reached a dangerous level, and it is recommended to activate the emergency plan", "the chemical substances detected in the XX river exceed the standard, and the pollution source needs to be investigated immediately". These marking results are quickly conveyed to the environmental protection department to guide them to take effective measures, such as restricting the driving of high-emission vehicles and urgently shutting down pollution sources, so as to timely control environmental deterioration and protect public health.

[0023] 3. Disaster Warning Scenarios

[0024] In terms of disaster warning, the application of the warning event distribution system is particularly critical. The system integrates real-time data from a variety of smart IoT sensor devices such as earthquake monitoring stations, meteorological satellites, and hydrological monitoring stations. These data cover dynamic information on natural phenomena such as earthquake fluctuations, meteorological changes, and water level fluctuations. Through attention processing, the system can quickly capture abnormal signals that may indicate the occurrence of disasters, such as earthquake precursors, the formation of extreme weather patterns, and abnormal rises in river water levels.

[0025] The AI ​​decision analysis algorithm further analyzes these data to reveal the multi-modal state trends of natural disasters, such as the periodicity of earthquake activity and the frequent occurrence of extreme climate events. On the other hand, it evaluates the resilience of different geographical regions to disasters, that is, the vulnerability and resilience under factors such as terrain, building density, and population distribution. Through integrated analysis, the system generates an abnormal event linkage attribute vector, which clearly points out the potential disaster type, impact range, and severity.

[0026] Finally, the system annotates the warning events with keywords according to the urgency and impact of the disaster warning, for example, "A strong earthquake is expected to occur in the XX region within the next 24 hours, please activate the emergency response mechanism immediately", "The XX River Basin is about to usher in a severe flood, and residents along the coast need to be evacuated urgently". These annotation results are directly transmitted to the emergency management department and relevant local departments to guide them to take prompt actions, such as evacuating the masses, deploying rescue resources, and activating shelters, so as to effectively reduce the losses caused by the disaster and protect people's lives and property.

[0027] It can be seen that the early warning event distribution system, in the three major scenarios of traffic management, environmental protection and disaster early warning, can achieve rapid identification, accurate early warning and effective distribution of abnormal events by efficiently integrating and analyzing smart IoT sensor monitoring data and using the AI ​​decision analysis algorithm of the present invention. In this way, it not only improves the level of intelligent urban management, but also plays an irreplaceable role in ensuring public safety and promoting sustainable development.

[0028] It is worth mentioning that when discussing the application of the early warning event distribution system, the following technical terms can be deeply understood through the three specific scenarios of traffic management, environmental protection and disaster warning: smart IoT sensor monitoring data to be analyzed for early warning, sensor monitoring attention data, target abnormal event label information, target multimodal state trend representation, target city interface energy level representation, and abnormal event linkage attribute vector.

[0029] First, in the traffic management scenario, the smart IoT sensor monitoring data to be analyzed for early warning mainly includes real-time information from road monitoring cameras, traffic flow sensors, vehicle GPS and other equipment. These data record key traffic indicators such as traffic volume, vehicle speed, and road occupancy rate in detail. The sensor monitoring attention data is obtained by processing these raw data through a deep learning model. It highlights the key features in the data, such as a sudden increase in traffic volume or an abnormally slow speed. These data points are often precursors to traffic congestion or accidents.

[0030] The target abnormal event label information here refers to the specific traffic events that the system needs to identify and warn, such as traffic congestion, traffic accidents, etc. The target multimodal state trend representation is the result of analyzing the multimodal state of traffic flow, which reveals the changing rules of traffic flow, such as the periodicity of morning and evening peaks and special flow patterns during holidays. The target urban interface energy level representation evaluates the carrying capacity and saturation of the urban road interface, that is, the performance of different sections in dealing with traffic flow. In this way, the system generates an abnormal event linkage attribute vector, which is a comprehensive indicator that integrates multimodal state trends and urban interface energy levels. In the traffic management scenario, it may point to a section of road that is about to reach the congestion threshold, or a specific location where a traffic accident has occurred, to guide the traffic management department to respond quickly.

[0031] Secondly, in the environmental protection scenario, the smart IoT sensor monitoring data to be analyzed for early warning covers real-time data from sensors such as air quality monitoring stations, water quality monitoring stations, and noise monitoring equipment. These data reflect environmental indicators such as air quality index, water pollutant concentration, and noise decibel level. Sensor monitoring attention data screens out monitoring values ​​that exceed the normal range, such as a sudden increase in PM2.5 concentration or the exceeding of the standard for specific chemical substances. These data points require special attention from the environmental protection department. The target abnormal event label information here refers to environmental events that require early warning, such as environmental pollution and ecological damage. The target multimodal state trend representation analyzes the changing laws of environmental indicators, such as seasonal changes in air quality and the improvement trend of water quality after rain. The target city interface energy level representation evaluates the environmental carrying energy level of different areas of the city, that is, the relationship between the environmental self-recovery capacity and the current pollution pressure. In this way, the abnormal event linkage attribute vector generated by the system may point to an area where the air quality reaches a dangerous level or a water source where excessive chemical substances are detected in the environmental protection scenario. This comprehensive indicator guides the environmental protection department to take effective measures to control environmental deterioration in a timely manner.

[0032] Finally, in the disaster warning scenario, the smart IoT sensor monitoring data to be analyzed for early warning integrates real-time information from multiple devices such as earthquake monitoring stations, meteorological satellites, and hydrological monitoring stations. These data cover the dynamics of natural phenomena such as earthquake fluctuations, meteorological changes, and water level fluctuations. Sensor monitoring attention data quickly captures abnormal signals that may indicate the occurrence of disasters, such as earthquake precursors, the formation of extreme weather patterns, and abnormal rises in river water levels. The target abnormal event label information here refers to natural disaster events that require early warning, such as earthquakes, floods, and extreme weather. The target multimodal state trend representation reveals the changing laws of natural disasters, such as the periodicity of earthquake activity and the frequent occurrence trend of extreme climate events. The target city interface energy level representation evaluates the disaster resistance level of different geographical regions, taking into account the vulnerability and resilience under factors such as terrain, building density, and population distribution. In this way, the abnormal event linkage attribute vector generated by the system may clearly indicate in the disaster warning scenario that a strong earthquake is about to occur in a certain area, or that a river basin is about to usher in a severe flood. This comprehensive indicator guides emergency management departments and relevant local departments to take prompt action to effectively reduce the losses caused by disasters.

[0033] It can be seen that through the examples of the three application scenarios of traffic management, environmental protection and disaster warning, the early warning event distribution system can accurately identify abnormal events in different scenarios and achieve rapid early warning and effective distribution.

[0034] The present invention obtains the smart IoT sensor monitoring data to be analyzed for early warning, processes it with attention, effectively extracts key information, and forms sensor monitoring attention data. This step significantly enhances the system's sensitivity to abnormal signals, making subsequent analysis more accurate. Next, the debugged AI decision analysis algorithm is used to deeply analyze the target abnormal event label information in the monitoring data, which not only reveals the multimodal state trend of the event, but also evaluates the urban interface energy level, thereby fully grasping the development trend and impact range of the abnormal event. By integrating these two aspects of characterization, the present invention innovatively constructs an abnormal event linkage attribute vector, and this comprehensive indicator provides a scientific basis for early warning distribution. Finally, based on rich data support and accurate vector analysis, the present invention accurately marks the early warning distribution keywords, and performs efficient early warning event distribution processing accordingly, greatly improving the speed and accuracy of emergency response. In summary, the rapid identification, accurate early warning and effective distribution of abnormal events are achieved, providing strong support for urban safety management and disaster prevention and control.

[0035] In some optional embodiments, the calling of the debugged AI decision analysis algorithm respectively analyzes the multimodal state trend and the city interface energy level of the target abnormal event label information in the smart IoT sensor monitoring data to obtain a target multimodal state trend representation and a target city interface energy level representation, including: disassembling the sensor monitoring data set involved in the target abnormal event label information in the smart IoT sensor monitoring data to obtain a target sensor monitoring data record; calling the debugged AI decision analysis algorithm to mine the multimodal state trend elements of the target sensor monitoring data record to obtain a target multimodal state trend representation; disassembling the spatiotemporal description vector in the smart IoT sensor monitoring data based on the target abnormal event label information to obtain target city interface energy level data; calling the debugged AI decision analysis algorithm to mine the interface energy level element vector of the target city interface energy level data to obtain a target city interface energy level representation.

[0036] It can be understood that the process of calling the debugged AI decision analysis algorithm to analyze the multimodal state trend and urban interface energy level of the target abnormal event label information in the smart IoT sensor monitoring data to obtain the target multimodal state trend representation and the target urban interface energy level representation is a complex process that deeply integrates data analysis and artificial intelligence technology. This process not only requires refined processing of massive, multi-dimensional sensor monitoring data, but also requires the use of relevant algorithm models to reveal the laws and trends hidden behind the data.

[0037] In the application scenario of traffic management, smart IoT sensor monitoring data covers real-time information from multiple devices such as road monitoring cameras, traffic flow sensors, and vehicle GPS. These data record key traffic indicators such as vehicle flow, vehicle speed, and road occupancy in detail, forming the data basis for traffic management. In this scenario, the target abnormal event label information may include events that require special attention, such as traffic congestion and traffic accidents. In order to warn and analyze these events, the warning event distribution system needs to first disassemble the sensor monitoring data set involved in these label information, and extract the sensor monitoring data records that are closely related to the target event. For example, in the case of traffic congestion, the warning event distribution system will focus on data records that show a significant decrease in vehicle speed and a continuous increase in road occupancy.

[0038] Next, the debugged AI decision analysis algorithm is called to mine the multimodal state trend elements recorded in these target sensor monitoring data. The purpose of this step is to reveal the changing patterns of traffic flow, such as the periodicity of morning and evening peaks, special flow patterns during holidays, etc. The algorithm analyzes traffic flow data in different time periods and sections, identifies the pattern features, and generates a target multimodal state trend representation based on this. This representation is a highly abstract data model that can accurately describe the dynamic changes in traffic flow and provide an important reference for subsequent warning distribution.

[0039] At the same time, the early warning event distribution system also needs to disassemble the spatiotemporal description vector of the target abnormal event label information in the smart IoT sensor monitoring data to obtain the target city interface energy level data. In the traffic management scenario, the city interface energy level data mainly reflects the carrying capacity and saturation of the urban road interface. Through comprehensive analysis of traffic flow data on different sections and at different time points, the early warning event distribution system can evaluate the performance of different areas in dealing with traffic flow, that is, their interface energy level. For example, some sections of road are often congested during peak hours, indicating that their interface energy level is low and measures need to be taken to improve it.

[0040] Finally, the debugged AI decision analysis algorithm is called to mine the interface energy level element vector of the target city interface energy level data to obtain the target city interface energy level representation. This representation is a comprehensive evaluation model that takes into account multiple factors such as road design, traffic flow, and traffic rules, and can accurately reflect the actual carrying capacity and potential risks of the urban road interface. In the practice of traffic management, this representation can help the early warning event distribution system identify those sections of the road that are prone to congestion or accidents, so as to take measures in advance to intervene and prevent them.

[0041] In the application scenario of environmental protection, smart IoT sensor monitoring data covers real-time data from sensors such as air quality monitoring stations, water quality monitoring stations, and noise monitoring equipment. These data reflect environmental indicators such as air quality index, water pollutant concentration, and noise decibel level, and constitute the data basis for environmental protection. In this scenario, the target abnormal event label information may include events that require special attention, such as environmental pollution and ecological damage. In order to warn and analyze these events, the warning event distribution system needs to first disassemble the sensor monitoring data set involved in these label information, and extract the sensor monitoring data records that are closely related to the target event. For example, in the case of air pollution, the warning event distribution system will focus on data records that show a sudden increase in the air quality index and excessive levels of specific chemicals.

[0042] Next, the debugged AI decision analysis algorithm is called to mine the multi-modal state trend elements recorded in these target sensor monitoring data. The purpose of this step is to reveal the changing patterns of environmental indicators, such as seasonal changes in air quality and the improvement trend of water quality after rain. The algorithm analyzes environmental data in different time periods and regions, identifies the pattern characteristics, and generates a target multi-modal state trend representation based on this. This representation can accurately describe the dynamic changes of environmental indicators and provide an important reference for subsequent warning distribution. For example, when predicting air pollution events, the warning event distribution system can infer changes in air quality in the next few days based on historical data and current trends.

[0043] At the same time, the early warning event distribution system also needs to disassemble the spatiotemporal description vector of the target abnormal event label information in the smart IoT sensor monitoring data to obtain the target city interface energy level data. In the scenario of environmental protection, the urban interface energy level data mainly reflects the environmental carrying capacity and self-recovery capacity of different areas of the city. Through comprehensive analysis of environmental data in different regions and at different time points, the early warning event distribution system can assess the vulnerability and resilience of different regions in the face of environmental pollution or ecological damage. For example, some industrial areas often have excessive air quality during peak emission periods, indicating that their environmental carrying capacity is low and measures need to be taken to improve it.

[0044] Finally, the debugged AI decision analysis algorithm is called to mine the interface energy level element vector of the target city interface energy level data to obtain the target city interface energy level representation. The target city interface energy level representation is a comprehensive evaluation model that takes into account the impact of multiple factors such as terrain, climate, and human activities on the environment, and can accurately reflect the actual environmental conditions and potential risks in different areas of the city. In the practice of environmental protection, the target city interface energy level representation can help the early warning event distribution system identify areas that are prone to environmental pollution or ecological damage, so that measures can be taken in advance for protection and restoration.

[0045] In the application scenario of disaster warning, smart IoT sensor monitoring data covers real-time information from a variety of equipment such as earthquake monitoring stations, meteorological satellites, and hydrological monitoring stations. These data cover the dynamics of natural phenomena such as earthquake fluctuations, meteorological changes, and water level fluctuations, and constitute the data basis for disaster warning. In this scenario, the target abnormal event label information may include earthquakes, floods, extreme weather, and other events that require special attention. In order to warn and analyze these events, the warning event distribution system needs to first disassemble the sensor monitoring data set involved in these label information, and extract the sensor monitoring data records that are closely related to the target event. For example, in the case of earthquakes, the warning event distribution system will focus on data records that show abnormal earthquake fluctuations and crustal deformation.

[0046] Next, the debugged AI decision analysis algorithm is called to mine the multi-modal state trend elements recorded in these target sensor monitoring data. The purpose of this step is to reveal the changing patterns of natural disasters, such as the periodicity of seismic activity and the frequent occurrence of extreme climate events. The algorithm analyzes natural disaster data from different time periods and locations, identifies the pattern features, and generates a target multi-modal state trend representation based on this. This representation can accurately describe the dynamic changes of natural disasters and provide an important reference for subsequent warning distribution. For example, when predicting earthquake events, the early warning event distribution system can infer the possibility of earthquake activity in the future based on historical earthquake data and current crustal deformation.

[0047] At the same time, the early warning event distribution system also needs to disassemble the spatiotemporal description vector of the target abnormal event label information in the smart IoT sensor monitoring data to obtain the target city interface energy level data. In the disaster warning scenario, the city interface energy level data mainly reflects the disaster resistance and recovery capabilities of different areas of the city. Through comprehensive analysis of disaster data in different regions and at different time points, the early warning event distribution system can assess the vulnerability and resilience of different regions in the face of natural disasters. For example, some cities located in mountainous areas are prone to disasters such as landslides and mudslides during the rainy season, indicating that their disaster resistance is low and measures need to be taken to strengthen and protect them.

[0048] Finally, the debugged AI decision analysis algorithm is called to mine the interface energy level element vector of the target city's interface energy level data to obtain the target city's interface energy level representation. The target city's interface energy level representation is a comprehensive evaluation model that takes into account the impact of multiple factors such as terrain, building density, and population distribution on disasters, and can accurately reflect the actual conditions and potential risks of different areas of the city when facing natural disasters. In the practice of disaster warning, the target city's interface energy level representation can help the warning event distribution system identify areas that are prone to disasters or severely affected by disasters, so as to take measures in advance to prevent and respond. For example, in flood warnings, the warning event distribution system can predict which areas may be flooded based on the city's interface energy level representation, and promptly notify relevant departments and residents to evacuate and transfer.

[0049] It can be seen that by calling the debugged AI decision analysis algorithm to deeply analyze the smart IoT sensor monitoring data, the early warning event distribution system can obtain the target multi-mode state trend representation and the target city interface energy level representation. These two representations respectively reveal the dynamic change law of abnormal events and the actual conditions and potential risks of different areas of the city when facing abnormal events. In application scenarios such as traffic management, environmental protection and disaster warning, these two representations play an important role, providing scientific and accurate early warning information for the early warning event distribution system, helping the early warning event distribution system to better respond to various emergency risks.

[0050] In the next step, the sensor monitoring data set involved in the target abnormal event label information in the smart IoT sensor monitoring data is disassembled to obtain the target sensor monitoring data record, including: determining the early warning distribution keywords of all sensor monitoring tasks in the smart IoT sensor monitoring data; determining the task division identifier of the target sensor monitoring data set in the smart IoT sensor monitoring data based on the early warning distribution keywords of the sensor monitoring task; sampling the target sensor monitoring data set according to the determined task division identifier to obtain the target sensor monitoring data record.

[0051] Based on this embodiment, the sensor monitoring data set involved in the target abnormal event label information in the smart IoT sensor monitoring data is disassembled to obtain the target sensor monitoring data record, which is an important process involving data processing and task division. This process aims to extract data records closely related to the target abnormal event from the massive sensor monitoring data, providing an accurate data basis for subsequent analysis and early warning.

[0052] In the application scenario of traffic management, smart IoT sensor monitoring data covers real-time information from various traffic monitoring devices, such as road monitoring cameras, traffic flow sensors, vehicle GPS, etc. These data together constitute the data basis for traffic management, recording key traffic indicators such as vehicle flow, vehicle speed, and road occupancy rate. In this scenario, the goal of the early warning event distribution system is to identify sensor monitoring data records related to abnormal events such as traffic congestion and traffic accidents. In order to achieve this goal, the early warning event distribution system first needs to determine the early warning distribution keywords of all sensor monitoring tasks in the smart IoT sensor monitoring data. These keywords are the key to identifying the target sensor monitoring data set, and they may include "traffic congestion", "traffic accident", "vehicle speed reduction", etc. Through these keywords, the early warning event distribution system can preliminarily screen out sensor monitoring data sets related to traffic management. Next, based on the early warning distribution keywords of the sensor monitoring task, the early warning event distribution system needs to determine the task division identifier of the target sensor monitoring data set in the smart IoT sensor monitoring data. These identifiers are the inherent attributes of the data set, and they help the early warning event distribution system to further subdivide the data set so as to more accurately find data records related to the target abnormal event. In the scenario of traffic management, the task division identifier may include time periods (such as morning and evening rush hours), geographical locations (such as specific road sections or intersections), traffic types (such as cars, buses), etc. Finally, based on the determined task division identifier, the early warning event distribution system samples the target sensor monitoring data set to obtain the target sensor monitoring data record. This process is a refined processing of the data set, which ensures that the data records extracted by the early warning event distribution system are both comprehensive and accurate, and can truly reflect the status and trend of the target abnormal events. In the practice of traffic management, these data records will provide reference traffic flow information for the early warning event distribution system, helping the early warning event distribution system to better respond to traffic congestion, traffic accidents, etc.

[0053] In the application scenario of environmental protection, smart IoT sensor monitoring data covers real-time data from sensors such as air quality monitoring stations, water quality monitoring stations, and noise monitoring equipment. These data reflect environmental indicators such as air quality index, water pollutant concentration, and noise decibel level, and constitute the data basis for environmental protection. In this scenario, the goal of the early warning event distribution system is to identify sensor monitoring data records related to abnormal events such as environmental pollution and ecological damage. In order to achieve this goal, the early warning event distribution system also needs to first determine the early warning distribution keywords for all sensor monitoring tasks in the smart IoT sensor monitoring data. In the scenario of environmental protection, these keywords may include "air pollution", "water quality deterioration", "noise exceeding the standard", etc. Through these keywords, the early warning event distribution system can preliminarily screen out sensor monitoring data sets related to environmental protection. Next, based on the early warning distribution keywords of the sensor monitoring task, the early warning event distribution system needs to determine the task division identification of the target sensor monitoring data set in the smart IoT sensor monitoring data. In the scenario of environmental protection, the task division identifier may include the geographical location of the monitoring point (such as industrial area, residential area), monitoring time (such as specific season, time period), environmental factors (such as temperature, humidity), etc. These identifiers help the early warning event distribution system to more accurately locate the data records related to the target abnormal event. Finally, based on the determined task division identifier, the early warning event distribution system samples the target sensor monitoring data set to obtain the target sensor monitoring data record. In the practice of environmental protection, these data records will provide reference environmental quality information for the early warning event distribution system, helping the early warning event distribution system to respond more effectively to environmental pollution, ecological damage, etc. For example, by analyzing these data records, the early warning event distribution system can find that the air quality in certain industrial areas continues to deteriorate within a specific time period, so as to take timely measures to intervene and govern.

[0054] In the application scenario of disaster warning, smart IoT sensor monitoring data covers real-time information from a variety of equipment such as earthquake monitoring stations, meteorological satellites, and hydrological monitoring stations. These data cover the dynamics of natural phenomena such as earthquake fluctuations, meteorological changes, and water level fluctuations, and constitute the data basis for disaster warning. In this scenario, the goal of the warning event distribution system is to identify sensor monitoring data records related to abnormal events such as earthquakes, floods, and extreme weather. In order to achieve this goal, the warning event distribution system first needs to determine the warning distribution keywords for all sensor monitoring tasks in the smart IoT sensor monitoring data. In the disaster warning scenario, these keywords may include "earthquake", "flood", "extreme weather", etc. Through these keywords, the warning event distribution system can preliminarily screen out sensor monitoring data sets related to disaster warning. Next, based on the warning distribution keywords of the sensor monitoring task, the warning event distribution system needs to determine the task division identification of the target sensor monitoring data set in the smart IoT sensor monitoring data. In the disaster warning scenario, the task division identifier may include geographical location (such as earthquake zones, flood-prone areas), time (such as earthquake active period, rainy season), disaster type (such as earthquake, flood, hurricane), etc. These identifiers help the warning event distribution system to more accurately locate the data records related to the target abnormal event. Finally, based on the determined task division identifier, the warning event distribution system samples the target sensor monitoring data set to obtain the target sensor monitoring data record. In the practice of disaster warning, these data records will provide reference disaster information for the warning event distribution system, helping the warning event distribution system to respond more effectively to earthquakes, floods, extreme weather and other problems. For example, by analyzing these data records, the warning event distribution system can find that certain earthquake zones have abnormally frequent seismic activities in a specific time period, so as to timely issue earthquake warning information and reduce potential casualties and property losses.

[0055] By disassembling the sensor monitoring data set involved in the target abnormal event label information in the smart IoT sensor monitoring data, the early warning event distribution system can obtain the sensor monitoring data records closely related to the target abnormal event. This process involves multiple steps such as determining the early warning distribution keywords, determining the task division identifier, and sampling to obtain the target data records. In application scenarios such as traffic management, environmental protection, and disaster warning, this embodiment provides accurate and comprehensive data support for the early warning event distribution system, helping the early warning event distribution system to better respond to various emergency risks.

[0056] Under other possible design ideas, the spatiotemporal description vector in the smart IoT sensor monitoring data based on the target abnormal event label information is disassembled to obtain the target city interface energy level data, including: using a deep structured semantic network to disassemble the smart IoT sensor monitoring data, and obtaining a sensor monitoring key task identifier based on the disassembly result; determining the spatiotemporal correlation between each sensor monitoring task of the smart IoT sensor monitoring data and the sensor monitoring key task identifier to obtain a spatiotemporal description vector set; extracting the spatiotemporal description vector belonging to the target abnormal event label information from the spatiotemporal description vector set based on the target sensor monitoring data set to obtain the target city interface energy level data.

[0057] In the actual application process, in order to extract valuable information for key areas such as urban management, environmental protection and disaster warning, it is necessary to implement it based on the above design ideas. The following will use three specific application scenarios of traffic management, environmental protection and disaster warning to elaborate on how to disassemble the spatiotemporal description vector in the smart IoT sensor monitoring data based on the target abnormal event label information, and then obtain the technical solution of the target city interface energy level data.

[0058] First, let's focus on the application scenario of traffic management. In this scenario, smart IoT sensor monitoring data covers all aspects of urban traffic, including traffic flow, vehicle speed, traffic light status, traffic violations, etc. These data together constitute a real-time portrait of urban traffic, providing reference decision support for traffic managers. However, it is not easy to extract information related to target abnormal events (such as traffic congestion, traffic accidents, etc.) from these massive data. This requires the early warning event distribution system to use deep structured semantic networks to disassemble smart IoT sensor monitoring data. Through the disassembly of deep structured semantic networks, the early warning event distribution system can reveal the key task identifiers of sensor monitoring hidden behind the data. These identifiers are important clues to understand the state of urban traffic and identify abnormal events. For example, they may include specific traffic flow patterns, vehicle speed change characteristics, traffic light switching rules, etc. Once the early warning event distribution system determines these key task identifiers, it can further analyze their spatiotemporal correlation with each sensor monitoring task in the smart IoT sensor monitoring data. Spatiotemporal correlation analysis is a bridge connecting sensor monitoring data with actual application scenarios. It helps the early warning event distribution system understand the temporal and spatial relationships between different sensor monitoring tasks, thereby revealing the abnormal state of urban traffic. For example, by analyzing the data of traffic flow sensors and vehicle speed sensors, the early warning event distribution system can find that the traffic flow of certain sections of roads has increased abnormally and the vehicle speed has dropped significantly in a specific time period, which is likely to be a sign of traffic congestion. Through such analysis, the early warning event distribution system can obtain a set of spatiotemporal description vectors that accurately describe the abnormal state of urban traffic. Finally, based on the target sensor monitoring data set, the early warning event distribution system can extract the spatiotemporal description vectors belonging to the target abnormal event label information from the spatiotemporal description vector set. These vectors not only contain basic information such as the time and location of the abnormal event, but also contain deep information such as the severity and development trend of the event. Through further analysis of these vectors, the early warning event distribution system can obtain the target city interface energy level data, which provides accurate and comprehensive information on traffic abnormal events for traffic managers, helping them to better formulate traffic management strategies, optimize traffic signal control, and alleviate traffic congestion.

[0059] Turning to the application scenario of environmental protection. In this scenario, the smart IoT sensor monitoring data mainly covers the real-time monitoring data of environmental indicators such as air quality, water quality, and noise. These data are important bases for evaluating environmental quality and discovering environmental pollution problems. However, to extract information related to target abnormal events (such as air pollution, water quality deterioration, etc.) from these data, the early warning event distribution system also needs to use deep structured semantic networks for disassembly. By disassembling the smart IoT sensor monitoring data, the early warning event distribution system can obtain a series of key sensor monitoring task identifiers related to environmental protection. These identifiers may include specific pollutant concentration change patterns, abnormal fluctuations in water quality indicators, and excessive noise decibel levels. After determining these key task identifiers, the early warning event distribution system can further analyze their spatiotemporal correlation with each sensor monitoring task. In the scenario of environmental protection, spatiotemporal correlation analysis also plays an important role. It helps the early warning event distribution system understand the mutual influence between different environmental monitoring points and reveal the spatiotemporal distribution characteristics of environmental pollution problems. For example, by analyzing the air quality data at different monitoring points, the early warning event distribution system can find that the air pollution problem in some industrial areas continues to worsen within a specific period of time and may affect the surrounding residential areas. Through such analysis, the early warning event distribution system can obtain a set of spatiotemporal description vectors, which accurately describe the spatiotemporal characteristics of environmental pollution problems. Finally, based on the target sensor monitoring data set, the early warning event distribution system can extract the spatiotemporal description vectors belonging to the target abnormal event label information from the spatiotemporal description vector set. These vectors not only contain basic information such as the time and location of environmental pollution problems, but also contain deep information such as the severity of the problem and the possible scope of impact. Through further analysis of these vectors, the early warning event distribution system can obtain the interface energy level data of the target city, which provides accurate and comprehensive information on environmental pollution problems for environmental protection departments, helping them to better formulate environmental protection strategies, strengthen environmental monitoring and governance, and protect public health. Finally, let the early warning event distribution system explore the application scenarios of disaster warning. In this scenario, the smart IoT sensor monitoring data covers real-time monitoring data of various natural disasters such as earthquakes, floods, and extreme weather. These data are an important basis for disaster warning and emergency response. However, it is not easy to extract information related to target abnormal events (such as earthquakes, floods, etc.) from these data. This also requires the early warning event distribution system to use deep structured semantic networks for disassembly. By disassembling the smart IoT sensor monitoring data, the early warning event distribution system can obtain a series of key sensor monitoring task identifiers related to disaster warning. These identifiers may include specific earthquake waveform characteristics, abnormal rise in flood water levels, changes in meteorological parameters of extreme weather, etc. After determining these key task identifiers, the early warning event distribution system can further analyze their spatiotemporal correlation with each sensor monitoring task.

[0060] In the disaster warning scenario, spatiotemporal correlation analysis is particularly important. It helps the warning event distribution system understand the mutual influence between different disaster monitoring points and reveal the spatiotemporal laws of disaster occurrence and development. For example, by analyzing the data of earthquake monitoring stations, the warning event distribution system can find that certain earthquake zones have abnormally frequent seismic activities in a specific time period, which may indicate that a major earthquake is about to occur. Through such analysis, the warning event distribution system can obtain a set of spatiotemporal description vectors that accurately describe the spatiotemporal characteristics of the disaster. Finally, based on the target sensor monitoring data set, the warning event distribution system can extract the spatiotemporal description vectors belonging to the target abnormal event label information from the spatiotemporal description vector set. These vectors not only contain basic information such as the time and location of the disaster, but also contain deep information such as the intensity of the disaster and the possible impact range. Through further analysis of these vectors, the warning event distribution system can obtain the interface energy level data of the target city, which provides accurate and comprehensive disaster warning information for emergency management departments, helping them to better formulate emergency response plans, issue warning information in a timely manner, and reduce disaster losses.

[0061] With this design, by using a deep structured semantic network to disassemble the smart IoT sensor monitoring data and analyzing the spatiotemporal correlation between the disassembly results and the sensor monitoring tasks, the early warning event distribution system can effectively extract information related to the target abnormal events.

[0062] In a preferred embodiment, the spatiotemporal description vector in the smart IoT sensor monitoring data based on the target abnormal event label information is disassembled to obtain the target city interface energy level data, including: using a deep structured semantic network to disassemble the smart IoT sensor monitoring data, and obtaining a sensor monitoring key task identifier according to the disassembly result; extracting the sensor monitoring tasks belonging to the target abnormal event label information from the smart IoT sensor monitoring data based on the target sensor monitoring data set to obtain a sensor monitoring task set; determining the spatiotemporal correlation between each sensor monitoring task in the sensor monitoring task set and the sensor monitoring key task identifier to obtain the target city interface energy level data.

[0063] When it comes to the fine management and emergency response of smart cities, in order to extract valuable information in key areas such as urban traffic management, environmental protection and disaster warning from massive data, this embodiment adopts a technical solution based on the spatiotemporal description vector of target abnormal event label information in smart IoT sensor monitoring data. This solution finally obtains the target city interface energy level data through the disassembly of deep structured semantic networks, the extraction of sensor monitoring task sets, and the analysis of spatiotemporal correlation, providing strong support for urban management.

[0064] In the traffic management scenario, smart IoT sensor monitoring data is like the blood of the city, recording key information such as traffic flow, vehicle speed, and traffic light status in real time. These data together constitute a real-time portrait of urban traffic, providing a reference for traffic managers to make decisions. However, it is not easy to extract information related to target abnormal events (such as traffic congestion, traffic accidents, etc.) from these data. This requires the early warning event distribution system to use a deep structured semantic network to disassemble the smart IoT sensor monitoring data.

[0065] The deep structured semantic network reveals the key task identifiers of sensor monitoring hidden behind the data by deeply disassembling the smart IoT sensor monitoring data. These identifiers are important clues for understanding the urban traffic status and identifying abnormal events. For example, they may include specific traffic flow patterns, vehicle speed change characteristics, traffic light switching rules, etc. These key task identifiers are like the traffic fingerprints of the city, providing a basis for the subsequent abnormal event identification.

[0066] After obtaining the key sensor monitoring task identification, the early warning event distribution system needs to further extract the sensor monitoring tasks that belong to the target abnormal event label information from the smart IoT sensor monitoring data. This process is like fishing for a specific school of fish in a vast ocean of data. The early warning event distribution system selects the sensor monitoring tasks related to the target abnormal event (such as traffic congestion) from the sensor monitoring data based on the label information of the target abnormal event. These tasks may include traffic flow monitoring and vehicle speed monitoring on a specific road section. Through this step, the early warning event distribution system obtains a set of sensor monitoring tasks related to the target abnormal event.

[0067] Next, the early warning event distribution system needs to determine the spatiotemporal correlation between each task in the sensor monitoring task set and the key task identification of sensor monitoring. This process is like drawing a dynamic traffic map of a city. The early warning event distribution system analyzes the temporal and spatial relationships between each sensor monitoring task, as well as the degree of correlation between them and the key task identification. For example, the early warning event distribution system finds that the traffic flow monitoring task of a certain section of road is highly correlated with the key task identification of traffic congestion within a specific time period, indicating that traffic congestion is likely to occur on this section of road during that time period. Through such analysis, the early warning event distribution system can obtain a set of spatiotemporal description vectors that accurately describe the abnormal state of urban traffic and its spatiotemporal distribution characteristics.

[0068] Finally, based on these spatiotemporal description vectors, the early warning event distribution system can further analyze and obtain the target city interface energy level data. These data not only contain basic information such as the time and location of traffic abnormalities, but also contain in-depth information such as the severity and development trend of the incident. Through in-depth analysis of these data, traffic managers can more accurately understand the operating status of urban traffic, timely discover and solve problems such as traffic congestion, thereby optimizing traffic management strategies and improving the overall efficiency of urban traffic.

[0069] In other examples, the early warning event distribution system turns to the application scenario of environmental protection. In this scenario, the smart IoT sensor monitoring data is like nature's stethoscope, recording the monitoring data of environmental indicators such as air quality, water quality, and noise in real time. These data are an important basis for evaluating environmental quality and discovering environmental pollution problems. However, to extract information related to target abnormal events (such as air pollution, water quality deterioration, etc.) from these data, the early warning event distribution system also needs to use a deep structured semantic network for disassembly.

[0070] By disassembling the deep structured semantic network, the early warning event dispatching system can obtain a series of key sensor monitoring task identifiers related to environmental protection. These identifiers may include specific pollutant concentration change patterns, abnormal fluctuations in water quality indicators, and excessive noise decibel levels. These key task identifiers are like environmental health indicators, providing a basis for subsequent abnormal environmental event identification.

[0071] After obtaining the key sensor monitoring task identification, the early warning event distribution system needs to further extract the sensor monitoring tasks that belong to the target abnormal event label information from the smart IoT sensor monitoring data. For example, if the early warning event distribution system focuses on air pollution, the early warning event distribution system will filter out sensor monitoring tasks related to air quality monitoring. These tasks may include PM2.5 concentration monitoring in a specific area, harmful gas emission monitoring, etc. Through this step, the early warning event distribution system obtains a set of sensor monitoring tasks related to the target abnormal environmental event.

[0072] Next, the early warning event distribution system needs to determine the spatiotemporal correlation between each task in the sensor monitoring task set and the key task identification of sensor monitoring. This process is like drawing a map of the dynamic changes of the environment. The early warning event distribution system analyzes the temporal and spatial relationships between each sensor monitoring task, as well as the degree of correlation between them and the key task identification. For example, the early warning event distribution system found that the PM2.5 concentration monitoring task in a certain area was highly correlated with the key task identification of air pollution within a specific time period, indicating that an air pollution incident was likely to have occurred in the area during that time period. Through such analysis, the early warning event distribution system can obtain a set of spatiotemporal description vectors that accurately describe the spatiotemporal characteristics of environmental pollution problems.

[0073] Finally, based on these spatiotemporal description vectors, the early warning event distribution system can further analyze and obtain the target city interface energy level data. These data not only contain basic information such as the time and location of environmental pollution problems, but also contain in-depth information such as the severity of the problem and the possible scope of impact. Through in-depth analysis of these data, environmental protection departments can understand the environmental quality status more accurately, discover and solve environmental pollution problems in a timely manner, and thus formulate more effective environmental protection strategies.

[0074] When it comes to disaster warning application scenarios, smart IoT sensor monitoring data records real-time monitoring data of various natural disasters such as earthquakes, floods, and extreme weather. These data are an important basis for disaster warning and emergency response. However, it is not easy to extract information related to target abnormal events (such as earthquakes, floods, etc.) from these data. This also requires the warning event distribution system to use a deep structured semantic network for disassembly.

[0075] By disassembling the deep structured semantic network, the early warning event distribution system can obtain a series of key sensor monitoring task identifiers related to disaster early warning. These identifiers may include specific earthquake waveform characteristics, abnormal rise in flood water levels, changes in meteorological parameters of extreme weather, etc. These key task identifiers are like early warning signals of disasters, providing a basis for subsequent disaster event identification.

[0076] After obtaining the key sensor monitoring task identification, the early warning event distribution system needs to further extract the sensor monitoring tasks that belong to the target abnormal event label information from the smart IoT sensor monitoring data. For example, if the early warning event distribution system focuses on earthquake disasters, the early warning event distribution system will filter out sensor monitoring tasks related to earthquake monitoring. These tasks may include seismic waveform monitoring and crustal movement monitoring in a specific earthquake zone. Through this step, the early warning event distribution system obtains a set of sensor monitoring tasks related to the target abnormal disaster event.

[0077] Next, the early warning event distribution system needs to determine the temporal and spatial correlation between each task in this sensor monitoring task set and the key task identification of sensor monitoring. This process is like drawing a dynamic development map of the disaster. The early warning event distribution system analyzes the temporal and spatial relationships between each sensor monitoring task, as well as the degree of correlation between them and the key task identification. For example, the early warning event distribution system found that the seismic waveform monitoring task of a certain seismic belt was highly correlated with the key task identification of earthquake occurrence within a specific time period, indicating that an earthquake event was likely to have occurred in the seismic belt within that time period. Through such analysis, the early warning event distribution system can obtain a set of spatiotemporal description vectors that accurately describe the spatiotemporal characteristics of the disaster and its development trend.

[0078] Finally, based on these spatiotemporal description vectors, the early warning event distribution system can further analyze and obtain the target city interface energy level data. These data not only contain basic information such as the time and location of the disaster, but also contain in-depth information such as the intensity of the disaster, possible impact range, and development trend. Through in-depth analysis of these data, emergency management departments can more accurately understand the development of disasters, issue early warning information in a timely manner, and formulate effective emergency response plans, thereby reducing disaster losses.

[0079] In this way, the smart IoT sensor monitoring data is disassembled by using a deep structured semantic network, and the sensor monitoring key task identification is obtained based on the disassembly results; then the sensor monitoring task set belonging to the target abnormal event label information is extracted based on the target sensor monitoring data set; finally, the spatiotemporal correlation between each task in the sensor monitoring task set and the key task identification is determined, and the early warning event distribution system can effectively extract information related to the target abnormal event.

[0080] In other exemplary embodiments, the smart IoT sensor monitoring data is subjected to attention processing to obtain sensor monitoring attention data, including: using a multi-head attention mechanism to determine the sensor monitoring characteristic variables of each data unit in the smart IoT sensor monitoring data; and configuring attention weights for the smart IoT sensor monitoring data based on the sensor monitoring characteristic variables to obtain sensor monitoring attention data.

[0081] In the actual implementation process, facing massive and complex data streams, how to efficiently extract valuable information has become a major technical barrier in the development of smart IoT. Based on this, this embodiment innovatively uses a multi-head attention mechanism to deeply analyze the smart IoT sensor monitoring data, thereby obtaining more valuable sensor monitoring attention data.

[0082] First of all, the core of this embodiment is the multi-head attention mechanism. In traditional data processing methods, data is often simply aggregated or filtered, but this method is difficult to capture the complex relationships and intrinsic characteristics between data. The multi-head attention mechanism is a more intelligent way of processing data. It can start from multiple dimensions and pay attention to different aspects of the data at the same time, so as to understand the meaning of the data more comprehensively. In the smart IoT sensor monitoring data, each data unit contains rich information, such as temperature, humidity, pressure, vibration, etc., which together constitute the overall state of the device or environment. The multi-head attention mechanism can accurately capture the sensor monitoring feature variables in these data units, whether it is a small change or a significant trend, it can be captured by it.

[0083] Next, determining these sensor monitoring feature variables requires the powerful ability of the multi-head attention mechanism. The multi-head attention mechanism will conduct an in-depth analysis of each data unit and extract the characteristics of the data from multiple angles. For example, in temperature data, it will pay attention to the temperature change trend, fluctuation range, abnormal values, etc.; in vibration data, it will capture the frequency, amplitude, mode, etc. of the vibration. These feature variables can uniquely identify the state and changes of the data.

[0084] After determining the characteristic variables of sensor monitoring, the next step is to configure the attention weights for the smart IoT sensor monitoring data. This step is a key part of the technical solution, which determines which data is truly valuable and which data needs to be focused on. In traditional data processing, all data are often treated equally, but in fact, different data have different importance in decision-making and judgment. Some data may contain key information, which is crucial to understanding the operating status of the equipment or the changing trend of the environment; while some data may just be noise or insignificant details.

[0085] Therefore, it is particularly important to configure the attention weights for smart IoT sensor monitoring data based on sensor monitoring characteristic variables. This step is equivalent to assigning different "attention levels" to the data. Data containing important information will be given higher weights, which means they will play a greater role in subsequent analysis and decision-making; while relatively minor data will be given lower weights to reduce attention and processing. Such weight configuration not only improves the efficiency of data processing, but also makes the analysis results more accurate and reliable.

[0086] Finally, through the processing of the multi-head attention mechanism and the configuration of attention weights, the early warning event distribution system obtains sensor monitoring attention data. These data are no longer raw massive data streams, but valuable information that has been carefully refined and screened. They can more accurately reflect the status of the device or environment and reveal potential problems and trends more promptly. For the application of smart IoT, this means higher efficiency, lower costs and better decision support.

[0087] In detail, in smart cities, it can be used to monitor and predict traffic flow, helping urban planners to arrange traffic signals and road layouts more reasonably; in industrial production, it can be used for equipment fault warning and maintenance management, improving the stability and efficiency of production lines; in the field of environmental protection, it can be used for air quality monitoring and pollution source tracking, providing a scientific basis for the formulation of environmental protection strategies.

[0088] In this way, the multi-head attention mechanism is used to deeply process the sensor monitoring data of the smart Internet of Things, and more valuable sensor monitoring attention data is obtained. It not only improves the efficiency and accuracy of data processing, but also provides strong support for the widespread application of smart Internet of Things.

[0089] In a replaceable embodiment, the warning distribution keywords of the target abnormal event label information are labeled based on the smart IoT sensor monitoring data, the sensor monitoring attention data and the abnormal event linkage attribute vector to obtain the labeling result, including: calling the debugged AI decision analysis algorithm to perform knowledge mining on the smart IoT sensor monitoring data, the sensor monitoring attention data and the abnormal event linkage attribute vector to obtain a target event discrimination knowledge vector; analyzing the warning keyword confidence weight of the target abnormal event label information based on the target event discrimination knowledge vector; and generating the labeling result of the smart IoT sensor monitoring data according to the warning keyword confidence weight of the target abnormal event label information.

[0090] In actual application, how to accurately capture and label the warning distribution keywords related to the target abnormal event is the key to improving the system's intelligent warning and response capabilities. On this basis, this embodiment integrates the smart IoT sensor monitoring data, sensor monitoring attention data, and abnormal event linkage attribute vectors, and finally achieves accurate labeling of the warning distribution keywords of the target abnormal event label information through corresponding calculations and processing.

[0091] First of all, the core of this technical solution is to call the debugged AI decision analysis algorithm, which can perform in-depth knowledge mining on the input smart IoT sensor monitoring data, sensor monitoring attention data and abnormal event linkage attribute vectors. In this process, the algorithm will carefully analyze each piece of data, find the relationship and rules between them, and then extract the target event discrimination knowledge vector.

[0092] As the basis of this technical solution, smart IoT sensor monitoring data provides rich and comprehensive real-time information. These data cover the monitoring results of various sensor devices, such as temperature, humidity, pressure, flow, etc., which together constitute a multi-dimensional data space. The sensor monitoring attention data is the result of further refinement of this data space. Through the processing of the multi-head attention mechanism, the early warning event distribution system can more accurately capture the key features in the data, which are the information that is crucial to judging abnormal events.

[0093] The abnormal event linkage attribute vector is another important component of this technical solution. It contains various attribute information related to the target abnormal event, such as event type, location, impact range, etc. Based on this, the debugged AI decision analysis algorithm will conduct in-depth cross-analysis and fusion processing on these three, and extract the target event discrimination knowledge vector from them.

[0094] Next, based on the target event discriminant knowledge vector, the warning event distribution system begins to analyze the warning keyword confidence weights of the target abnormal event label information. In this step, the warning event distribution system needs to determine which keywords are most relevant to the target abnormal event based on the information and rules in the target event discriminant knowledge vector, and assign them corresponding confidence weights.

[0095] Finally, based on the warning keyword confidence weights of the target abnormal event label information, the warning event distribution system generates the annotation results of the smart IoT sensor monitoring data. The annotation results of the smart IoT sensor monitoring data clearly list all the warning distribution keywords related to the target abnormal event and give their importance ranking. Therefore, the annotation results of the smart IoT sensor monitoring data can provide strong data support for subsequent warning distribution and emergency response.

[0096] In this way, by integrating smart IoT sensor monitoring data, sensor monitoring attention data and abnormal event linkage attribute vectors, and calling the debugged AI decision analysis algorithm for knowledge mining and keyword confidence weight analysis, the accurate labeling of warning distribution keywords for target abnormal event label information is finally achieved. In this way, not only the warning and response capabilities of the smart IoT system are improved, but also a solid foundation for future intelligent development is laid.

[0097] In other possible embodiments, the AI ​​decision analysis algorithm includes a first mining algorithm branch, an integrated algorithm branch, and a second mining algorithm branch. The calling of the debugged AI decision analysis algorithm performs knowledge mining on the smart IoT sensor monitoring data, the sensor monitoring attention data, and the abnormal event linkage attribute vector to obtain a target event discrimination knowledge vector, including: using the first mining algorithm branch to perform knowledge mining on the smart IoT sensor monitoring data and the sensor monitoring attention data respectively to obtain a scene event discrimination knowledge vector and a local event discrimination knowledge vector; using the integrated algorithm branch to integrate the scene event discrimination knowledge vector, the local event discrimination knowledge vector, and the abnormal event linkage attribute vector to obtain an integrated event discrimination knowledge vector; and performing knowledge mining on the integrated event discrimination knowledge vector through the second mining algorithm branch to obtain a target event discrimination knowledge vector.

[0098] Based on this embodiment, the AI ​​decision analysis algorithm is designed to include three core branches: the first mining algorithm branch, the integration algorithm branch, and the second mining algorithm branch. These three branches each undertake different tasks and work together to complete the deep knowledge mining of smart IoT sensor monitoring data, sensor monitoring attention data, and abnormal event linkage attribute vectors, and finally obtain the target event discrimination knowledge vector.

[0099] First, the first mining algorithm branch is responsible for preliminary knowledge mining of smart IoT sensor monitoring data and sensor monitoring attention data. As the cornerstone of IoT technology, smart IoT sensor monitoring data contains massive real-time data collected from various sensors. Sensor monitoring attention data is specially processed data that highlights information that is critical for decision analysis. The first mining algorithm branch obtains scene event discrimination knowledge vectors and local event discrimination knowledge vectors by deeply analyzing and mining these data. These two vectors represent the understanding and judgment of events from the macro scene level and the micro local level, respectively.

[0100] Next, the integration algorithm branch will come into play. It integrates the previously obtained scene event discrimination knowledge vector, local event discrimination knowledge vector, and abnormal event linkage attribute vector. The abnormal event linkage attribute vector is a comprehensive description of various related attributes of the abnormal event, which records key information such as the type, nature, and scope of influence of the event. The task of the integration algorithm branch is to organically integrate these three vectors to form a more comprehensive and in-depth event discrimination knowledge vector. This process involves comprehensive analysis and processing of different vectors to ensure that the final event discrimination knowledge vector can fully reflect the characteristics and attributes of the target abnormal event.

[0101] Finally, the second mining algorithm branch conducts further knowledge mining on the integrated event discrimination knowledge vector. This step aims to extract a more accurate and in-depth target event discrimination knowledge vector from the integrated event discrimination knowledge vector. The second mining algorithm branch uses its powerful analytical ability to conduct in-depth analysis and mining on the integrated event discrimination knowledge vector to obtain the most accurate and comprehensive description and judgment of the target abnormal event.

[0102] In this way, through the collaborative work of the three algorithm branches, deep knowledge mining of smart IoT sensor monitoring data, sensor monitoring attention data, and abnormal event linkage attribute vectors is achieved. This process involves preliminary data mining, vector integration, and further knowledge mining, and finally obtains the target event discrimination knowledge vector. The implementation of this technical solution not only improves the discrimination and early warning capabilities of target abnormal events, but also lays a solid foundation for the further development of the smart city IoT field.

[0103] Under some other preferred design ideas, the AI ​​decision analysis algorithm also includes a residual processing branch. Before analyzing the warning keyword confidence weight of the target abnormal event label information based on the target event discrimination knowledge vector, it also includes: obtaining latent space event derivative features through the residual processing branch of the AI ​​decision analysis algorithm; the warning keyword confidence weight of the target abnormal event label information based on the target event discrimination knowledge vector includes: analyzing the warning keyword confidence weight of the target abnormal event label information based on the latent space event derivative features and the target event discrimination knowledge vector.

[0104] Based on this design idea, the AI ​​decision analysis algorithm demonstrates a more comprehensive and in-depth analysis capability. Its unique structural design makes the confidence weight analysis of the warning keywords of the target abnormal events more accurate. Specifically, the algorithm not only includes the first mining algorithm branch, the integrated algorithm branch, and the second mining algorithm branch mentioned earlier, but also adds an additional residual processing branch. This additional branch is responsible for more detailed data processing to extract key information hidden deep in the data.

[0105] Before analyzing the warning keyword confidence weights of the target abnormal event label information based on the target event discrimination knowledge vector, the algorithm first uses the residual processing branch to obtain the latent space event derivative features.

[0106] The working principle of the residual processing branch can be simply described as follows: it performs multiple iterative calculations on the input data, each calculation focuses on a different part of the data, and tries to extract information that was ignored or not fully expressed in the previous calculation.

[0107] Through such processing, the residual processing branch can capture data details that are easily overlooked in traditional analysis methods. These details often contain key features of events. These features are called "latent space event-derived features". They are mined from the deep structure of the data and are of great significance for understanding the nature of events and predicting their development trends.

[0108] After obtaining the derived features of latent space events, the algorithm will then analyze the warning keyword confidence weights of the target abnormal event label information based on these features and the previously obtained target event discriminant knowledge vector. This step is the core of the algorithm analysis process, which is equivalent to "diagnosing" the event to determine which keywords can most accurately describe and warn the event.

[0109] In this process, the algorithm will comprehensively consider the information contained in the latent space event derivative features and the target event discriminant knowledge vector, and obtain the confidence weight of each warning keyword through complex calculations and analysis. These weights reflect the degree of association between the keyword and the target abnormal event. The higher the weight, the more accurately the keyword can describe and warn the event.

[0110] Through such analysis, the algorithm can generate a detailed warning keyword confidence weight report, which clearly lists the weight of each keyword and provides strong data support for subsequent warnings and responses.

[0111] In this way, by adding a residual processing branch and comprehensively considering the latent space event derivative features and the target event discrimination knowledge vector when analyzing the warning keyword confidence weight, a more accurate and comprehensive analysis of the target abnormal event is achieved. The implementation of this technical solution not only improves the ability to discriminate and warn target abnormal events, but also provides strong technical support for the further development of the smart city Internet of Things field.

[0112] In some optional embodiments, the labeling result of the smart IoT sensor monitoring data is generated based on the warning keyword confidence weight of the target abnormal event label information, including: obtaining event state change data of the target abnormal event label information; determining that the warning keyword confidence weight of the target abnormal event label information is greater than the warning distribution keyword with a set weight; and generating the labeling result of the smart IoT sensor monitoring data based on the determined warning distribution keyword according to the event state change data.

[0113] In actual application, the process of how to generate the labeling results of smart IoT sensor monitoring data based on the warning keyword confidence weight of the target abnormal event label information involves multiple links, and each link is closely linked to ensure the accuracy and effectiveness of the labeling results.

[0114] First, the system needs to obtain the event state change data of the target abnormal event label information. This step is the basis for generating annotation results, because the event state change data reflects the changes in each state of the abnormal event during its occurrence and development, and is an important basis for understanding the overall picture of the event and determining the warning keywords. After obtaining this data, the system can have a comprehensive understanding of the abnormal event and provide strong support for subsequent steps.

[0115] Secondly, the system needs to determine that the warning keyword confidence weight of the target abnormal event label information is greater than the warning distribution keyword with the set weight. This step is the key to generating annotation results, because the warning keyword confidence weight reflects the degree of correlation and accuracy between the keyword and the abnormal event. By setting a weight threshold, the system can screen out those warning keywords that are highly relevant to abnormal events and have high accuracy as candidate words for subsequent generation of annotation results. This step ensures the accuracy and pertinence of the annotation results and avoids interference of irrelevant or misleading information on the annotation results.

[0116] Finally, the system needs to generate the annotation results of the smart IoT sensor monitoring data with the determined warning distribution keywords based on the event state change data. This step is the final link in generating the annotation results, and it is also the key to converting the results of the previous steps into practical applications. In this step, the system will associate and match the warning distribution keywords with the sensor monitoring data according to the specific situation of the event state change data, and generate annotation results with practical significance. These annotation results can clearly reflect the performance and characteristics of abnormal events in the sensor monitoring data, and provide strong data support for subsequent monitoring, warning and response.

[0117] Specifically, the system can use adaptive algorithms and technologies to achieve automation and intelligence of the process. For example, the system can use machine learning algorithms to analyze and mine event state change data, extract key features and patterns, and then use these features and patterns to filter and sort warning keywords to determine warning distribution keywords with higher confidence weights; finally, associate and match these keywords with sensor monitoring data to generate annotation results.

[0118] In addition, in order to ensure the accuracy and reliability of the annotation results, the system can also adopt some additional measures and methods. For example, the system can manually review and verify the generated annotation results to ensure that they meet the actual situation and business needs; at the same time, the system can also use some feedback mechanisms and optimization algorithms to continuously optimize and improve the annotation process to improve the accuracy and efficiency of annotation.

[0119] In this way, generating the annotation results of smart IoT sensor monitoring data based on the warning keyword confidence weight of the target abnormal event label information is a complex and delicate process. By obtaining event state change data, determining warning distribution keywords and generating annotation results based on these data, the system can achieve accurate monitoring and warning of abnormal events, providing strong technical support for the development of smart city IoT.

[0120] In other possible embodiments, the calling of the debugged AI decision analysis algorithm to perform knowledge mining on the smart IoT sensor monitoring data, the sensor monitoring attention data and the abnormal event linkage attribute vector to obtain a target event discrimination knowledge vector also includes: collecting multiple historical smart IoT sensor monitoring data containing target abnormal event label information as an algorithm debugging sample set; debugging the debugged AI decision analysis algorithm based on the algorithm debugging sample set to obtain an AI decision analysis algorithm.

[0121] It is understandable that the process of calling the debugged AI decision analysis algorithm to analyze the smart IoT sensor monitoring data, sensor monitoring attention data, target sensor monitoring data set and target city interface energy level representation to obtain the warning keyword confidence weights related to the target abnormal event label information is not straightforward. Before this, a series of preparations are required to ensure that the AI ​​decision analysis algorithm can accurately and effectively complete the analysis task.

[0122] First, it is necessary to collect multiple historical smart IoT sensor monitoring data containing target abnormal event label information as algorithm debugging sample sets. This step is the basis for building and debugging AI decision analysis algorithms, because only with enough sample data can an algorithm model with generalization ability be trained. These historical data contain various characteristics and manifestations of target abnormal events, and are an important basis for algorithm learning and understanding abnormal events. When collecting data, it is necessary to ensure the diversity and representativeness of the data to cover as many abnormal event types and situations as possible.

[0123] Secondly, based on the collected algorithm debugging sample set, the debugged AI decision analysis algorithm is further debugged and optimized. This step is a key link in the algorithm development process, which determines the performance and accuracy of the algorithm in practical applications. During the debugging process, the algorithm needs to be repeatedly tested and verified to ensure that it can correctly parse the sensor monitoring data and accurately output the warning keyword confidence weights related to the target abnormal event label information. At the same time, the algorithm also needs to adjust and optimize parameters to improve its parsing speed and accuracy.

[0124] When debugging an AI decision analysis algorithm, multiple factors need to be considered. First, it is necessary to ensure that the algorithm can accurately identify the characteristics of abnormal events in sensor monitoring data, which requires feature extraction and selection operations on the algorithm. Second, it is necessary to ensure that the algorithm can correctly calculate the confidence weights of warning keywords, which requires weight calculation and optimization operations on the algorithm. Finally, the algorithm needs to be tested and verified as a whole to ensure that it can maintain stable performance and accuracy under various circumstances.

[0125] To achieve these goals, a variety of techniques and methods can be used during the debugging process. For example, machine learning algorithms can be used to extract features and classify sensor monitoring data to identify abnormal events. Statistical methods can be used to calculate and optimize the confidence weights of warning keywords to improve the accuracy and reliability of the weights. Methods such as cross-validation can also be used to test and verify the algorithm as a whole to ensure its performance and accuracy.

[0126] After repeated debugging and optimization, the AI ​​decision analysis algorithm has gradually become stable and accurate. At this point, the algorithm can be used to parse smart IoT sensor monitoring data, sensor monitoring attention data, target sensor monitoring data set, and target city interface energy level characterization. During the parsing process, the algorithm will first preprocess and feature extract the input data to extract key information related to the target abnormal event. The algorithm will then use this key information to calculate the confidence weights of the warning keywords, and sort and filter the keywords according to the weights. Finally, the algorithm will output a list of warning keywords and their confidence weights as the parsing result.

[0127] The analysis results are of great significance for the application of smart city IoT. Through the analysis results, we can clearly understand the performance and characteristics of the target abnormal event in the sensor monitoring data, as well as which keywords can accurately describe and warn the event. This information has important guiding significance for subsequent monitoring, warning and response work, and can assist in more accurately identifying and responding to various abnormal events, and improving the security and reliability of the smart IoT system.

[0128] Therefore, before calling the debugged AI decision analysis algorithm to analyze the smart IoT sensor monitoring data, etc., sufficient preparation is required, including collecting algorithm debugging sample sets and debugging and optimizing the algorithm. These works ensure the performance and accuracy of the algorithm in practical applications and lay a solid foundation for subsequent analysis work. Through the application of this technical solution, target abnormal events can be identified and warned more accurately, improving the overall performance and security of the smart IoT system.

[0129] In a further embodiment, the debugged AI decision analysis algorithm is debugged based on the algorithm debugging sample set to obtain the AI ​​decision analysis algorithm, including: performing attention processing on the historical smart IoT sensor monitoring data to obtain historical sensor monitoring attention data; disassembling the target sensor monitoring data set in the historical smart IoT sensor monitoring data to obtain a target sensor monitoring sample data set; determining the city interface energy level of the target abnormal event label information in the smart IoT sensor monitoring data to obtain a target city interface energy level sample data set; calling the debugged AI decision analysis algorithm to parse the historical smart IoT sensor monitoring data, the historical sensor monitoring attention data, the target sensor monitoring sample data set and the target city interface energy level sample data set to obtain a sensor monitoring training vector related to the historical smart IoT sensor monitoring data; obtaining the sensor monitoring prior vector of the historical smart IoT sensor monitoring data, and optimizing the AI ​​decision analysis algorithm based on the sensor monitoring prior vector and the sensor monitoring training vector of the historical smart IoT sensor monitoring data to obtain the AI ​​decision analysis algorithm.

[0130] It can be understood that the process of debugging the debugged AI decision analysis algorithm based on the algorithm debugging sample set to obtain a more complete and accurate AI decision analysis algorithm involves multiple links, and each link is closely linked to ensure the accuracy and effectiveness of the final algorithm.

[0131] First, the historical smart IoT sensor monitoring data needs to be processed with attention to obtain historical sensor monitoring attention data. This step is critical because attention processing can help the algorithm better focus on the key information in the data and ignore irrelevant or redundant parts. By performing attention processing on historical data, a more refined and valuable set of historical sensor monitoring attention data can be obtained, providing more targeted input for subsequent algorithm debugging.

[0132] Secondly, the target sensor monitoring data set in the historical smart IoT sensor monitoring data is disassembled to obtain the target sensor monitoring sample data set. The purpose of this step is to decompose the complex data set into smaller and easier to handle units so that the algorithm can better learn and understand the inherent laws and characteristics of the data. By disassembling, a set of representative and diverse target sensor monitoring sample data sets can be obtained, providing rich samples for debugging and optimization of the algorithm.

[0133] Then, the city interface energy level of the target abnormal event label information in the smart IoT sensor monitoring data is determined to obtain a target city interface energy level sample data set. This step is to associate the target abnormal event with the city interface energy level so that the algorithm can better understand the impact and propagation law of the event in the city interface. By determining the city interface energy level of the target abnormal event label information, a set of target city interface energy level sample data sets with clear labels and energy level information can be obtained, providing strong support for the subsequent analysis and optimization of the algorithm.

[0134] Next, the debugged AI decision analysis algorithm is called to parse the historical smart IoT sensor monitoring data, historical sensor monitoring attention data, target sensor monitoring sample data set, and target city interface energy level sample data set to obtain sensor monitoring training vectors related to the historical smart IoT sensor monitoring data. This step is the core of algorithm debugging. By parsing these four sets of data, the algorithm can extract key features and patterns related to target abnormal events and convert these features and patterns into sensor monitoring training vectors. These training vectors are the basis for subsequent optimization and learning of the algorithm.

[0135] At the same time, the sensor monitoring prior vector of the historical smart IoT sensor monitoring data is obtained, and the AI ​​decision analysis algorithm is optimized based on the sensor monitoring prior vector and sensor monitoring training vector of the historical smart IoT sensor monitoring data to obtain the final AI decision analysis algorithm. This step is the final stage of algorithm debugging and is also a crucial step. By obtaining the sensor monitoring prior vector, a set of known and accurate vectors can be provided for the algorithm as a reference. Then, the sensor monitoring training vector is compared and analyzed with the sensor monitoring prior vector, and the algorithm can continuously adjust and optimize its own parameters and structure to improve the accuracy and efficiency of the analysis. Finally, after repeated optimization and debugging, a more complete and accurate set of AI decision analysis algorithms can be obtained.

[0136] Therefore, the process of debugging the debugged AI decision analysis algorithm based on the algorithm debugging sample set is a complex and delicate process. Through the close cooperation and collaboration of multiple links such as attention processing, data decomposition, urban interface energy level determination, data analysis and algorithm optimization, a more complete and accurate AI decision analysis algorithm can be obtained. This set of algorithms not only has higher analysis accuracy and efficiency, but also can better adapt to and cope with various complex and changing application scenarios.

[0137] In some independent embodiments, after the warning event distribution processing is performed based on the annotation results, the method further includes: generating a warning event processing strategy based on feedback information from the warning event response end; wherein the warning event response end is the object end of the warning event distribution processing.

[0138] When it comes to technical solutions for distributing and processing warning events based on annotation results, we can start from the three key application scenarios of traffic management, environmental protection and disaster warning, and elaborate on the application of the solution in actual operations and the positive impact it brings.

[0139] 1. Application in Traffic Management

[0140] In the modern urban traffic management system, the warning event dispatching and processing technology based on the labeling results can automatically identify potential traffic congestion, accident risks or violations by real-time monitoring and analyzing traffic flow, road conditions and other data, and accurately label these events. This process mainly relies on relevant data collection equipment, such as high-definition cameras, radar detectors and various sensors.

[0141] Secondly, after the marking is completed, the system will immediately start the warning event distribution and processing mechanism. This means that according to the nature and urgency of the event, the system will automatically send relevant information to the corresponding response end, such as traffic police, road maintenance teams or emergency rescue services. This process greatly shortens the time interval from event discovery to response, effectively improving traffic management efficiency.

[0142] Then, based on the feedback from the early warning event response end, the system will further generate specific processing strategies. For example, if a traffic accident is detected, the system will not only notify the traffic police and ambulance, but also calculate the best diversion route based on the accident location, time and current traffic conditions, and guide vehicles to detour through traffic signal control or information release systems to reduce the impact of the accident on overall traffic.

[0143] Finally, this series of steps is based on powerful data processing and analysis capabilities: through continuous learning and optimization of historical data, the system can gradually improve its warning accuracy and strategy effectiveness, thereby significantly improving urban traffic conditions and reducing congestion and accidents in the long run.

[0144] 2. Application in Environmental Protection

[0145] In this scenario, the first task is to monitor and mark those events that may have a negative impact on the natural environment, such as air pollution, water pollution, illegal discharge of pollutants, etc. This relies on a three-dimensional monitoring network composed of satellite remote sensing, ground monitoring stations, drone inspections, and public reports.

[0146] Once the system identifies and marks such events, the early warning event dispatching and processing mechanism will be quickly activated. Depending on the nature of the event, the system will send an alert to the relevant emergency management department. For example, when a factory is found to be illegally discharging pollutants, the system will not only notify environmental monitoring personnel, but also automatically calculate the spread of pollution, predict possible ecological impacts, and provide this information to decision makers.

[0147] Then, based on the feedback from the responding end, the system will generate a specific treatment strategy. This may include emergency closure of pollution sources, evacuation of residents in affected areas, initiation of water or air purification measures, etc. At the same time, the system will track the progress of treatment, evaluate the effect, and adjust the strategy according to the actual situation to ensure that environmental damage is effectively controlled.

[0148] In this process, the application of technology not only improves the speed and efficiency of environmental response, but also helps environmental protection departments identify pollution trends and formulate more scientific and reasonable long-term governance plans through big data analysis.

[0149] 3. Application in Disaster Early Warning

[0150] In the field of disaster warning, warning event distribution and processing technology based on annotation results plays an irreplaceable role. Whether it is earthquakes, floods, typhoons or forest fires, the occurrence of these natural disasters is often accompanied by huge losses of life and property. Therefore, quickly and accurately identifying and responding to disaster warning signals is the key to reducing the impact of disasters.

[0151] First, the technology integrates data sources from multiple departments such as meteorology, geology, and hydrology, and uses the algorithm model of the present invention to predict and mark possible disasters. This process not only involves real-time analysis of large amounts of data, but also requires consideration of the interaction of multiple natural and social factors to ensure the accuracy of the warning.

[0152] Secondly, once the system issues a disaster warning, the warning event dispatching and processing mechanism will take immediate action. According to the type of disaster and the expected impact range, the system will send warning information to relevant departments, emergency rescue teams, and residents in the affected areas. For example, in earthquake warning, the system will not only notify relevant departments, but also issue earthquake warnings to the public through multiple channels such as mobile phone text messages, chat platforms, and broadcasts, guiding them to take correct risk avoidance measures.

[0153] Then, based on the feedback from the responding end, the system will generate a specific disaster response strategy, including mobilizing rescue resources, evacuating people, repairing infrastructure, etc. At the same time, the system will continue to monitor the development of the disaster, adjust the strategy according to the actual situation, and ensure the effective implementation of the rescue work.

[0154] Finally, it is worth mentioning that the application of technology in disaster warning is not limited to the warning and response stages. Through in-depth analysis of historical disaster data, the system can also help relevant departments identify disaster risk areas and formulate targeted disaster prevention and mitigation plans, thereby improving disaster resistance capabilities over a longer time scale.

[0155] In other independent embodiments, generating a warning event processing strategy based on feedback information from a warning event response terminal includes: extracting target warning response interaction semantics of the feedback information; determining core feedback behavior elements of the feedback information based on the feedback information and the target warning response interaction semantics of the feedback information; extracting the warning response interaction semantics of the core feedback behavior elements of the feedback information; acquiring an event matching description vector of the feedback information based on the target warning response interaction semantics of the feedback information and the warning response interaction semantics of the core feedback behavior elements of the feedback information; generating a warning event processing strategy for the feedback information based on the target warning response interaction semantics of the feedback information, the warning response interaction semantics of the core feedback behavior elements of the feedback information, and the event matching description vector of the feedback information.

[0156] It can be understood that the process of generating a warning event handling strategy based on the feedback information from the warning event response end is a complex and delicate task, which involves in-depth understanding, analysis and strategy generation of feedback information. This process can be specifically explained through three application scenarios: traffic management, environmental protection and disaster warning.

[0157] Let's first look at the traffic management scenario. In this scenario, warning events may include traffic accidents, traffic congestion, road construction, etc. When these events occur, the warning system will dispatch these events to relevant responders, such as traffic police, road maintenance teams, etc. After receiving the warning information, these responders will perform a series of processing and return the processing results to the warning system in the form of feedback information. At this time, the warning system needs to generate the next processing strategy based on these feedback information. Specifically, the warning system will first extract the target warning response interaction semantics in the feedback information, that is, the specific response content of the responder to the warning event. For example, for a traffic accident, the feedback information of the traffic police may include the accident location, accident type, casualties, etc. Then, the warning system will determine the core feedback behavior elements of the feedback information based on this information and the target warning response interaction semantics, that is, the actual actions taken by the responder, such as whether police force was dispatched, whether traffic diversion was carried out, etc. Next, the warning system will extract the warning response interaction semantics of these core feedback behavior elements, that is, convert these actual actions into a language that the system can understand. For example, "police force was dispatched" is converted into a semantic description such as "police force is in place". Then, the early warning system will obtain the event matching description vector of the feedback information based on the target early warning response interaction semantics of the feedback information and the early warning response interaction semantics of the core feedback behavior elements. This vector is actually a numerical representation of the feedback information, which can help the system understand the meaning of the feedback information more accurately. Finally, the early warning system will generate an early warning event processing strategy for the feedback information based on the target early warning response interaction semantics of the feedback information, the early warning response interaction semantics of the core feedback behavior elements, and the event matching description vector. This process is actually a decision-making process. The system will decide what action to take next based on the current feedback information and historical learning experience. For example, if the feedback information from the traffic police indicates that the accident has been handled, the system may generate a strategy to end the early warning; if the accident is still being handled, the system may generate a strategy to continue monitoring and wait for subsequent feedback.

[0158] Next, let's look at the environmental protection scenario. In this scenario, warning events may include air pollution, water pollution, illegal discharge of pollutants, etc. When these events occur, the warning system will dispatch these events to relevant environmental protection departments or law enforcement agencies. After receiving the warning information, these agencies will conduct a series of investigations and processing, and return the processing results to the warning system in the form of feedback information. Similarly, the warning system needs to generate the next processing strategy based on these feedback information. In this process, the warning system will first extract the target warning response interaction semantics in the feedback information, that is, the specific response content of the environmental protection department or law enforcement agency to the warning event. For example, for an air pollution incident, the feedback information of the environmental protection department may include the pollution source, pollution degree, affected area, etc. Then, the warning system will determine the core feedback behavior elements of the feedback information based on this information and the target warning response interaction semantics, that is, the actual actions taken by the environmental protection department or law enforcement agency, such as whether an on-site investigation was conducted, whether the pollution source was dealt with, etc. Next, the warning system will extract the warning response interaction semantics of these core feedback behavior elements and convert these actual actions into a language that the system can understand. Then, the early warning system will obtain the event matching description vector of the feedback information based on the target early warning response interaction semantics of the feedback information and the early warning response interaction semantics of the core feedback behavior elements. This vector will help the system understand the feedback information from the environmental protection department or law enforcement agency more accurately. Finally, the early warning system will generate an early warning event processing strategy for the feedback information based on the target early warning response interaction semantics of the feedback information, the early warning response interaction semantics of the core feedback behavior elements, and the event matching description vector. This process is also a decision-making process. The system will decide what action to take next based on the current feedback information and historical learning experience. For example, if the feedback information from the environmental protection department indicates that the pollution source has been controlled, the system may generate a strategy to end the early warning; if the pollution source is still continuously emitting pollutants, the system may generate a strategy to continue monitoring and notify more relevant departments.

[0159] Finally, let's look at the disaster warning scenario. In this scenario, warning events may include natural disasters such as earthquakes, floods, and typhoons. When these events occur, the warning system will distribute these events to relevant departments or emergency rescue agencies. After receiving the warning information, these agencies will carry out a series of emergency preparation and response work, and return the work results to the warning system in the form of feedback information. Similarly, the warning system needs to generate the next processing strategy based on these feedback information. In this process, the warning system will first extract the target warning response interaction semantics in the feedback information, that is, the specific response content of the department or emergency rescue agency to the warning event. For example, for an earthquake warning, the department's feedback information may include the magnitude of the earthquake, the epicenter location, and the expected affected area. Then, based on this information and the target warning response interaction semantics, the warning system will determine the core feedback behavior elements of the feedback information, that is, the actual actions taken by the department or emergency rescue agency, such as whether the emergency plan has been activated, whether personnel evacuation has been carried out, etc. Next, the warning system will extract the warning response interaction semantics of these core feedback behavior elements and convert these actual actions into a language that the system can understand. Then, the early warning system will obtain the event matching description vector of the feedback information based on the target early warning response interaction semantics of the feedback information and the early warning response interaction semantics of the core feedback behavior elements. This vector will help the system understand the feedback information of the department or emergency rescue agency more accurately. Finally, the early warning system will generate an early warning event processing strategy for the feedback information based on the target early warning response interaction semantics of the feedback information, the early warning response interaction semantics of the core feedback behavior elements, and the event matching description vector. This process is also a decision-making process. The system will decide what action to take next based on the current feedback information and historical learning experience. For example, if the feedback information from the department indicates that the earthquake has occurred and caused damage, the system may generate a strategy to initiate a comprehensive emergency response; if the earthquake has not occurred but the warning is still ongoing, the system may generate a strategy to continue monitoring and prepare emergency resources.

[0160] In summary, whether in traffic management, environmental protection or disaster warning scenarios, generating warning event processing strategies based on feedback information from the warning event response end is a complex and delicate task. It requires in-depth understanding and analysis of feedback information, extracting key information and elements, and making decisions based on these information and elements. Through this process, the warning system can better respond to various warning events and improve the efficiency and accuracy of emergency response.

[0161] Further, Figure 2 FIG. 2 is a schematic diagram of a warning event distribution system 200 provided in an embodiment of the present invention. Figure 2The warning event dispatching system 200 shown includes a processor 210. The processor 210 can call and run a computer program from a memory to implement the method in the embodiment of the present invention.

[0162] Alternatively, if Figure 2 As shown, the warning event dispatching system 200 may further include a memory 230. The processor 210 may call and run a computer program from the memory 230 to implement the method in the embodiment of the present invention.

[0163] The memory 230 may be a separate device independent of the processor 210 , or may be integrated into the processor 210 .

[0164] Alternatively, if Figure 2 As shown, the warning event dispatching system 200 may further include a transceiver 220, and the processor 210 may control the transceiver 220 to interact with other devices, specifically, may send information or data to other devices, or receive information or data sent by other devices.

[0165] Optionally, the warning event dispatching system 200 may implement the corresponding processes corresponding to the storage engine or components in the storage engine (such as a processing module) or a device deployed with a storage engine in each method of the embodiments of the present invention, which will not be described in detail here for the sake of brevity.

[0166] It should be understood that the processor of the embodiment of the present invention may be an integrated circuit chip with signal processing capability.

[0167] It is understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memory. It should be noted that the memory of the systems and methods described herein is intended to include but is not limited to suitable types of memory.

[0168] Based on the above, a readable storage medium is provided, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the above method are implemented.

[0169] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the purpose of the present invention and the scope of protection of the present invention, all of which are within the protection of the present invention.

Claims

1. A method for automatically distributing early warning events for smart cities, characterized in that: The method is applied to a warning event distribution system, and the method comprises: Obtain smart IoT sensor monitoring data for early warning analysis; Performing attention processing on the smart IoT sensor monitoring data to obtain sensor monitoring attention data; Calling the debugged AI decision analysis algorithm to analyze the multi-mode state trend and the city interface energy level of the target abnormal event label information in the smart IoT sensor monitoring data, respectively, to obtain the target multi-mode state trend representation and the target city interface energy level representation; Integrate the target multi-mode state trend representation and the target city interface energy level representation to obtain an abnormal event linkage attribute vector; Based on the smart IoT sensor monitoring data, the sensor monitoring attention data and the abnormal event linkage attribute vector, the warning distribution keywords of the target abnormal event label information are marked to obtain a marking result, and a warning event distribution process is performed based on the marking result; The step of labeling the warning distribution keywords of the target abnormal event label information based on the smart IoT sensor monitoring data, the sensor monitoring attention data and the abnormal event linkage attribute vector to obtain the labeling results includes: Calling the debugged AI decision analysis algorithm to perform knowledge mining on the smart IoT sensor monitoring data, the sensor monitoring attention data and the abnormal event linkage attribute vector to obtain a target event discrimination knowledge vector; Analyzing the warning keyword confidence weight of the target abnormal event label information based on the target event discrimination knowledge vector; Generate the labeling result of the smart IoT sensor monitoring data according to the warning keyword confidence weight of the target abnormal event label information; The AI ​​decision analysis algorithm includes a first mining algorithm branch, an integrated algorithm branch, and a second mining algorithm branch. The calling of the debugged AI decision analysis algorithm performs knowledge mining on the smart IoT sensor monitoring data, the sensor monitoring attention data, and the abnormal event linkage attribute vector to obtain a target event discrimination knowledge vector, including: Using a first mining algorithm branch to perform knowledge mining on the smart IoT sensor monitoring data and the sensor monitoring attention data, respectively, to obtain a scene event discrimination knowledge vector and a local event discrimination knowledge vector; The scene event discrimination knowledge vector, the local event discrimination knowledge vector and the abnormal event linkage attribute vector are integrated by using an integration algorithm branch to obtain an integrated event discrimination knowledge vector; Performing knowledge mining on the integrated event discrimination knowledge vector through a second mining algorithm branch to obtain a target event discrimination knowledge vector; The AI ​​decision analysis algorithm also includes a residual processing branch. Before analyzing the warning keyword confidence weight of the target abnormal event label information based on the target event discrimination knowledge vector, it also includes: obtaining latent space event derivative features through the residual processing branch of the AI ​​decision analysis algorithm; the warning keyword confidence weight of the target abnormal event label information based on the target event discrimination knowledge vector includes: analyzing the warning keyword confidence weight of the target abnormal event label information based on the latent space event derivative features and the target event discrimination knowledge vector; The method of generating the annotation results of the smart IoT sensor monitoring data based on the warning keyword confidence weight of the target abnormal event label information includes: obtaining event state change data of the target abnormal event label information; determining that the warning keyword confidence weight of the target abnormal event label information is greater than the warning distribution keyword with a set weight; and generating the annotation results of the smart IoT sensor monitoring data based on the determined warning distribution keyword according to the event state change data.

2. The method according to claim 1, characterized in that The calling of the debugged AI decision analysis algorithm respectively analyzes the multi-mode state trend and the city interface energy level of the target abnormal event label information in the smart IoT sensor monitoring data to obtain the target multi-mode state trend representation and the target city interface energy level representation, including: Decomposing the sensor monitoring data set involved in the target abnormal event label information in the smart IoT sensor monitoring data to obtain the target sensor monitoring data record; Calling the debugged AI decision analysis algorithm to mine the multi-mode state trend elements recorded in the target sensor monitoring data to obtain the target multi-mode state trend representation; Based on the target abnormal event label information, the spatiotemporal description vector in the smart IoT sensor monitoring data is disassembled to obtain the target city interface energy level data; The debugged AI decision analysis algorithm is called to mine the interface energy level element vectors of the target city interface energy level data to obtain the target city interface energy level representation.

3. The method according to claim 2, characterized in that The sensor monitoring data set involved in the target abnormal event label information in the smart IoT sensor monitoring data is disassembled to obtain the target sensor monitoring data record, including: Determine the early warning distribution keywords of all sensor monitoring tasks in the smart IoT sensor monitoring data; Determine the task division identifier of the target sensor monitoring data set in the smart IoT sensor monitoring data based on the early warning distribution keywords of the sensor monitoring task; The target sensor monitoring data set is sampled according to the determined task division identifier to obtain the target sensor monitoring data record.

4. The method according to claim 2, characterized in that: The step of disassembling the spatiotemporal description vector in the smart IoT sensor monitoring data based on the target abnormal event label information to obtain the target city interface energy level data includes: The smart IoT sensor monitoring data is disassembled using a deep structured semantic network, and key sensor monitoring task identifiers are obtained based on the disassembly results; Determine the spatiotemporal correlation between each sensor monitoring task of the smart IoT sensor monitoring data and the sensor monitoring key task identifier to obtain a spatiotemporal description vector set; Based on the target sensor monitoring data set, the spatiotemporal description vector belonging to the target abnormal event label information is extracted from the spatiotemporal description vector set to obtain the target city interface energy level data.

5. The method according to claim 2, characterized in that: The step of disassembling the spatiotemporal description vector in the smart IoT sensor monitoring data based on the target abnormal event label information to obtain the target city interface energy level data includes: The smart IoT sensor monitoring data is disassembled using a deep structured semantic network, and key sensor monitoring task identifiers are obtained based on the disassembly results; Extracting sensor monitoring tasks belonging to target abnormal event label information from the smart IoT sensor monitoring data based on the target sensor monitoring data set to obtain a sensor monitoring task set; Determine the spatiotemporal correlation between each sensor monitoring task in the sensor monitoring task set and the sensor monitoring key task identifier, and obtain the target city interface energy level data.

6. The method according to claim 1, characterized in that The step of performing attention processing on the smart IoT sensor monitoring data to obtain sensor monitoring attention data includes: A multi-head attention mechanism is used to determine the sensor monitoring feature variables of each data unit in the smart IoT sensor monitoring data; Based on the sensor monitoring characteristic variables, the attention weight of the smart IoT sensor monitoring data is configured to obtain sensor monitoring attention data.

7. The method according to claim 1, characterized in that The calling of the debugged AI decision analysis algorithm performs knowledge mining on the smart IoT sensor monitoring data, the sensor monitoring attention data and the abnormal event linkage attribute vector to obtain a target event discrimination knowledge vector, and further includes: Collect multiple historical smart IoT sensor monitoring data containing target abnormal event label information as algorithm debugging sample sets; Debugging the debugged AI decision analysis algorithm based on the algorithm debugging sample set to obtain an AI decision analysis algorithm; The step of debugging the debugged AI decision analysis algorithm based on the algorithm debugging sample set to obtain the AI ​​decision analysis algorithm includes: Performing attention processing on the historical intelligent IoT sensor monitoring data to obtain historical sensor monitoring attention data; Disassembling the target sensor monitoring data set in the historical smart IoT sensor monitoring data to obtain a target sensor monitoring sample data set; Determine the city interface energy level of the target abnormal event label information in the smart IoT sensor monitoring data, and obtain a target city interface energy level sample data set; Calling the debugged AI decision analysis algorithm to parse the historical smart IoT sensor monitoring data, the historical sensor monitoring attention data, the target sensor monitoring sample data set, and the target city interface energy level sample data set to obtain a sensor monitoring training vector related to the historical smart IoT sensor monitoring data; The sensor monitoring prior vector of the historical smart IoT sensor monitoring data is obtained, and the AI ​​decision analysis algorithm is optimized based on the sensor monitoring prior vector and the sensor monitoring training vector of the historical smart IoT sensor monitoring data to obtain the AI ​​decision analysis algorithm.

8. A warning event distribution system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 7.

Images