Emergency scene content distribution method based on multi-modal data fusion and related device

Abstract

The invention discloses an emergency scene content distribution method based on multi-modal data fusion and a related device. The emergency scene content distribution method comprises the following steps: acquiring and preprocessing data of a data sensing layer; respectively constructing an intelligent analysis layer scene portrait and a user portrait; the content priority of the distribution decision layer is optimized, and an optimal distribution strategy is determined; the method further comprises the step of optimizing the distribution strategy by using the feedback data. According to the invention, by integrating disaster monitoring multi-source data and constructing an intelligent analysis layer and a distribution decision-making layer, accurate portraits of disaster scenes and dynamic division of user groups are realized, so that a content distribution strategy is optimized, and the transmission efficiency and coverage effect of emergency information are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of emergency management and information distribution, and in particular to a method and related device for distributing emergency scene content based on multimodal data fusion. Background Art

[0002] Guangdong, located in South China, is subject to frequent typhoons, and the power grid faces complex and diverse challenges in preventing typhoons and floods. There is an urgent need for digital online communication for emergency remote command and on-site collaboration, as well as online monitoring of emergency teams and emergency material deployment. In disaster response scenarios, timely and accurate information delivery is crucial. Traditional content distribution strategies often rely on a single data source, such as typhoon monitoring data or flood level data, and lack the integration and in-depth analysis of multi-source data, resulting in insufficient accuracy and timeliness in information distribution. Furthermore, existing methods fail to fully consider the differentiated information needs of different user groups, making it difficult to achieve personalized content push.

[0003] With the development of sensor technology, big data analysis technology and artificial intelligence technology, how to use multimodal data fusion technology to improve the efficiency and accuracy of emergency scene content distribution has become an urgent problem to be solved. Summary of the Invention

[0004] The present invention provides an emergency scenario content distribution method and related devices based on multimodal data fusion, which are used to integrate multi-source disaster monitoring data, build an intelligent analysis layer and a distribution decision layer, achieve accurate profiling of disaster scenarios and dynamic division of user groups, thereby optimizing content distribution strategies and improving the transmission efficiency and coverage of emergency information.

[0005] In view of this, a first aspect of the present invention provides an emergency scenario content distribution method based on multimodal data fusion, the method comprising:

[0006] Collecting disaster monitoring data and preprocessing the disaster monitoring data;

[0007] Based on the pre-processed disaster monitoring data, a multi-dimensional labeling system for disaster scenarios is constructed, and the labeling system for disaster scenarios is updated in real time using a machine learning algorithm. The labeling system includes disaster type, impact range, and vulnerability level;

[0008] Dynamically divide users into target groups based on their location information, behavior patterns, and historical interaction data, evaluate users' risk exposure by combining the tag system and location information, and match users' information needs based on the target group's characteristics and risk exposure;

[0009] Utilizing a reinforcement learning algorithm, the priority of content is dynamically adjusted according to the risk exposure and information needs of the user, and an optimal distribution strategy is determined based on the priority to distribute emergency scenario content.

[0010] Optionally, the method further includes: collecting user feedback data, and optimizing the optimal distribution strategy based on the feedback data.

[0011] Optionally, the method further includes constructing a multi-dimensional labeling system for disaster scenarios based on the pre-processed disaster monitoring data, and updating the labeling system for disaster scenarios in real time using a machine learning algorithm, and then further includes:

[0012] By introducing the weighted fusion algorithm to fuse the disaster monitoring data, the disaster intensity index is obtained. ;

[0013] Among them, the disaster intensity index The calculation expression is:

[0014] ;

[0015] Where, Indicates the typhoon intensity; T indicates the real-time temperature; d indicates the real-time flood peak water level; s indicates the typhoon-affected area; is the weighting coefficient, e is a constant; Represents the distance attenuation coefficient.

[0016] Optionally, the method for determining the optimal distribution strategy further includes:

[0017] By using a reinforcement learning algorithm, the priority of the content is dynamically adjusted according to the disaster intensity index, the risk exposure of the user and the information demand, and the optimal distribution strategy is determined according to the priority.

[0018] Optionally, the machine learning algorithm is a neural network machine learning algorithm;

[0019] Based on the neural network machine learning algorithm, historical data is trained to build a label prediction model for disaster scenarios. Multi-source data collected in real time is input into the trained label prediction model to predict the dynamic changes of disaster scenarios to obtain prediction results, and the label system is updated according to the prediction results.

[0020] Optionally, collecting disaster monitoring data and preprocessing the disaster monitoring data includes:

[0021] Collecting disaster monitoring data, the disaster monitoring data including: data including typhoon data and flood data;

[0022] Among them, the typhoon data is obtained by collecting typhoon range data, level, and typhoon eye center radius information in real time through a typhoon sensor network, and removing noise using data cleaning and preprocessing technology. The flood data is obtained by integrating water level data, rainfall data, and flow rate data from hydrological monitoring stations.

[0023] A second aspect of the present invention provides an emergency scenario content distribution system based on multimodal data fusion, the system comprising:

[0024] A collection unit, used for collecting disaster monitoring data and preprocessing the disaster monitoring data;

[0025] A construction unit is used to construct a multi-dimensional labeling system for disaster scenarios based on the pre-processed disaster monitoring data, and to update the labeling system for disaster scenarios in real time using a machine learning algorithm, wherein the labeling system includes disaster type, impact range, and vulnerability level;

[0026] A matching unit, configured to dynamically divide users into target groups based on their location information, behavior patterns, and historical interaction data, evaluate the risk exposure of users in combination with the tag system and the location information, and match the information needs of users based on the group characteristics of the target groups and the risk exposure;

[0027] The generation unit is used to use a reinforcement learning algorithm to dynamically adjust the priority of the content according to the user's risk exposure and information needs, and determine the optimal distribution strategy based on the priority to distribute emergency scenario content.

[0028] Optionally, it further includes: an optimization unit, configured to collect user feedback data and optimize the optimal distribution strategy according to the feedback data.

[0029] A third aspect of the present invention provides an emergency scenario content distribution device based on multimodal data fusion, the device comprising a processor and a memory:

[0030] The memory is used to store program code and transmit the program code to the processor;

[0031] The processor is used to execute the steps of the emergency scenario content distribution method based on multimodal data fusion as described in the first aspect according to the instructions in the program code.

[0032] A fourth aspect of the present invention provides a computer-readable storage medium, which is used to store program code, and the program code is used to execute the emergency scene content distribution method based on multimodal data fusion described in the first aspect above.

[0033] It can be seen from the above technical solutions that the present invention has the following advantages:

[0034] The present invention integrates multi-source disaster monitoring data to perceive the dynamic changes of disaster scenes in real time. It uses machine learning algorithms and real-time data access technology to dynamically update the label system of disaster scenes to ensure the timeliness and accuracy of information. It combines multiple channels such as text messages, social media, and emergency broadcasts, and selects the optimal distribution channel based on user portraits and scene portraits to ensure that information covers all users. Through user feedback data, the distribution strategy is dynamically adjusted and continuously optimized to ensure that the effect of information transmission is maximized. In an emergency, key information can be quickly and accurately delivered to all users, significantly improving the transmission efficiency and coverage of emergency information, and providing strong support for disaster emergency management and rescue work. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0036] Figure 1 A flowchart of a method for distributing emergency scene content based on multimodal data fusion provided by an embodiment of the present invention;

[0037] Figure 2 A schematic structural diagram of an emergency scenario content distribution system based on multimodal data fusion provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose, features, and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0039] See also Figure 1 , an embodiment of the present invention provides an emergency scenario content distribution method based on multimodal data fusion, comprising:

[0040] Step 101: Collect disaster monitoring data and pre-process the disaster monitoring data.

[0041] It should be noted that, in one embodiment, the method includes: collecting disaster monitoring data, the disaster monitoring data including: data including typhoon data and flood data;

[0042] Specifically, a typhoon sensor network is deployed in typhoon-prone areas to collect typhoon data in real time. Data cleaning and preprocessing techniques are then used to remove noise. Hydrological monitoring stations are deployed in key areas, such as rivers and reservoirs, to collect real-time data on water levels, rainfall, and other areas. Flood risk maps are generated using GIS (Geographic Information System, a specific and crucial spatial information system that, supported by computer hardware and software, collects, stores, manages, calculates, analyzes, displays, and describes geographically distributed data on the entire or partial Earth's surface (including the atmosphere)).

[0043] Furthermore, real-time typhoon data and flood data can be analyzed through drone remote sensing images. Drones are equipped with sensing equipment to collect data; typhoon wind speeds can be monitored by equipped sensors, and flood peak water level information can be captured.

[0044] Among them, the collected multi-source disaster monitoring data are cleaned, denoised and standardized to ensure the quality and consistency of the data; time series analysis and spatial interpolation technology are used to supplement and repair missing data.

[0045] Furthermore, mobile communication base station data and satellite positioning data can be used to generate real-time population heat maps to analyze the density and mobility trends of people within the disaster area. Base station signaling data can also be used to dynamically monitor the number of users in the area and their behavior patterns (such as the frequency of emergency calls and text messages).

[0046] We use crawler technology to capture disaster-related information on social media platforms in real time, and use natural language processing technology to extract public opinion hotspots, public sentiment and needs as disaster monitoring data. I will not go into details here.

[0047] Step 102: Based on the pre-processed disaster monitoring data, a multi-dimensional labeling system for disaster scenarios is constructed, and the labeling system for disaster scenarios is updated in real time using a machine learning algorithm. The labeling system includes disaster type, impact range, and vulnerability level.

[0048] It's important to note that a multi-dimensional labeling system for disaster scenarios is constructed based on disaster monitoring data, including disaster type (typhoon, flood, tornado, etc.), impact area (accurate to the block or community level), and vulnerability level (assessed based on population density, infrastructure conditions, etc.). Subsequently, a machine learning algorithm is used to update the disaster scenario labeling system in real time to ensure the timeliness and accuracy of the scenario portraits. This neural network machine learning algorithm is trained on historical data to construct a disaster scenario label prediction model. Real-time multi-source data is fed into the trained model to predict dynamic changes in disaster scenarios, and the labeling system is updated based on the predicted results.

[0049] Specifically, real-time multi-source data is fed into a trained model to predict the dynamics of disaster scenarios. Based on the predictions, the labeling system is updated, for example, by adjusting the disaster type, impact area, or vulnerability level. The accuracy of the labeling system is evaluated through user feedback and verification in actual disaster scenarios. Finally, the model and labeling system are further optimized based on this feedback. Furthermore, time series data (such as rainfall and water level changes) can be used to predict the dynamic development of disasters and adjust the labeling system in real time. Incorporating a reinforcement learning framework, the labeling system's update strategy can be dynamically adjusted, for example, adjusting the update frequency based on the severity and impact area of ​​the disaster.

[0050] Through the above steps, the disaster scene labeling system can be updated in real time, ensuring the timeliness and accuracy of the scene portrait, and providing reliable support for subsequent content distribution strategies.

[0051] In one embodiment, after step 102, the following further includes: fusing the disaster monitoring data by introducing a weighted fusion algorithm to obtain a disaster intensity index ;

[0052] Among them, the disaster intensity index The calculation expression is:

[0053] ;

[0054] Where, Indicates the typhoon intensity; T indicates the real-time temperature; d indicates the real-time flood peak water level; s indicates the typhoon-affected area; is the weighting coefficient, e is a constant; Represents the distance attenuation coefficient.

[0055] It should be noted that, assuming that in an actual use, the sensor measures the following data:

[0056] Typhoon intensity q Level 3.4, temperature ℃, peak water level m, typhoon affected area m²;

[0057] And, by assuming the weighting coefficient is: , , , , , l = 15 m, then we can substitute these values ​​into the formula to calculate the disaster intensity index .

[0058] First, calculate each term:

[0059]

[0060]

[0061]

[0062]

[0063] Then, the composite calculation:

[0064]

[0065] The disaster intensity index calculated at this time , indicating the comprehensive disaster intensity in the current area. This value serves as a key indicator for experts to determine whether to allocate relief supplies and rescue teams.

[0066] Through this step, the present invention innovatively combines multiple factors such as typhoon intensity, temperature, distance, flood peak water level, typhoon affected area, etc., and proposes a weighted fusion calculation method for the comprehensive disaster impact degree, which effectively solves the problem that when encountering multiple disaster problems, traditional methods cannot comprehensively consider the current situation and accurately reflect the severity of user disasters. As a core indicator, it will provide an accurate reference for disaster monitoring modeling in the current area and subsequent judgment on whether rescue materials and rescue teams need to be allocated, ensuring that the entire monitoring area can adaptively adjust the allocation of rescue materials and teams according to real-time conditions, and can effectively distribute content using multimodal data fusion technology based on the current situation, so that the public can understand the current regional situation in a timely manner.

[0067] Step 103: Dynamically divide the user's target group based on the user's location information, behavior pattern and historical interaction data, evaluate the user's risk exposure by combining the tag system and location information, and match the user's information needs based on the group characteristics and risk exposure of the target group.

[0068] It should be noted that target groups, such as enterprise safety officers, community residents, and frontline rescue teams, are dynamically divided based on users' location information, behavioral patterns, and historical interaction data. The system combines the disaster scenario tagging system with the user's location information to assess the user's risk exposure. Information needs are then matched based on their risk exposure and group characteristics. For example, if an enterprise safety officer's company is located in a coastal area prone to typhoons and their historical interaction data indicates a high level of attention to typhoon warnings, their risk exposure will be assessed as high. Therefore, the system prioritizes key information such as typhoon intensity, expected landfall time, and potential damage to enable them to prepare for disasters. For community residents, the system may prioritize practical information such as shelter locations, emergency evacuation routes, and disaster prevention knowledge to enhance their self-protection capabilities. Frontline rescue teams may require more detailed disaster site data and maps of relief supply distribution to support efficient rescue operations.

[0069] Step 104: Using a reinforcement learning algorithm, dynamically adjust the priority of the content based on the user's risk exposure and information needs, and determine the optimal distribution strategy based on the priority to distribute emergency scenario content.

[0070] In one embodiment, the method for determining the optimal distribution strategy further includes: using a reinforcement learning algorithm to dynamically adjust the priority of the content according to the disaster intensity index, the user's risk exposure and information needs, and determining the optimal distribution strategy based on the priority.

[0071] It should be noted that the reinforcement learning algorithm is used to dynamically adjust the priority of content based on the severity of the disaster scenario (i.e., the disaster intensity index), the user's risk exposure and information needs.

[0072] For example:

[0073] When a disaster scenario reaches the red alert level, key information will be pushed first to ensure that users in high-risk areas can obtain emergency evacuation guidelines in a timely manner.

[0074] Voice alarms are forcibly broadcast through the emergency broadcast system to cover the area as much as possible, and the traffic light system is linked to open the "rescue green wave" channel; high-risk groups receive encrypted text messages with evacuation routes (national encryption SM4 algorithm), and the on-board terminals of rescue vehicles synchronize high-precision waterlogging maps, equipped with AR navigation (AR navigation is augmented reality navigation, which is a navigation method that combines augmented reality (AR) technology with navigation functions) and a real-time water level monitoring system; access is made to the online car-hailing platform to dispatch vehicles to assist in transfer, and social personnel cooperate in rescue (demand matching degree ≥ 90%, drivers with high credit), and volunteer recruitment information is pushed through the integrated media APP.

[0075] Association rule recommendation: Based on user historical data, we mine the relevant associations behind user data, analyze the user's potential needs, and recommend information that may be of interest to the user.

[0076] Adapt the content format based on the user's device type (e.g., mobile, PC) (e.g., SMS pushes short messages, social media pushes detailed text and image information).

[0077] Optimize content push time based on user behavior patterns and time preferences. For example, technical content is pushed during the day on weekdays, while decision-making content is pushed in the evening.

[0078] Combine multiple channels, including SMS, social media, and emergency broadcasts, and select the optimal distribution channel based on user and scenario profiles. For example, prioritize broadcasting and television for the elderly and social media for younger groups.

[0079] Through a dynamic distribution mechanism driven by intelligent algorithms, we connect the contact resources scattered across social media, mobile applications and offline scenarios, and achieve a full-link closed loop of content production, distribution and optimization.

[0080] Further:

[0081] Provide feedback on customer touchpoints, accurately quantify the contribution of each content touchpoint to the final conversion, and dynamically optimize content delivery strategies through continuous monitoring of key indicators such as user stay time and secondary dissemination rate.

[0082] Specifically, by analyzing user behavior data (such as click-through rate, dwell time, bounce rate, etc.), we can identify user preferences and pain points regarding content. For example, if the click-through rate of a certain type of content is consistently below the benchmark, the content reorganization process will be automatically triggered.

[0083] Through intelligent tracking and streaming computing technologies, user behavior data (such as click heatmaps and bounce points) is captured in real time and synchronized with the optimization model. This immediacy enables the operations team to quickly identify high-potential content variations and terminate inefficient testing branches.

[0084] Combined with an automated workflow engine, content strategies can be quickly iterated to ensure content adaptability and consistency across different channels. For example, content formatting and distribution logic can be dynamically adjusted for different user groups.

[0085] By building a global data collection network, we can centralize user behavior data scattered across various channels to create a complete user journey map. For example, we can combine user behavior data from social media, email push, and mobile apps to optimize cross-channel distribution strategies.

[0086] Adapt to channel characteristics: Dynamically adjust the content presentation format based on the user behavior characteristics of different channels. For example, short video platforms are suitable for interactive content, while email marketing is more suitable for in-depth graphic information.

[0087] Through the above steps, the present invention can quickly and accurately deliver key information to all users in an emergency, significantly improving the transmission efficiency and coverage of emergency information, and providing strong support for disaster emergency management and rescue work.

[0088] In one embodiment, after step 104, the emergency scenario content distribution method based on multimodal data fusion provided by the present invention further includes: collecting user feedback data, and optimizing the optimal distribution strategy based on the feedback data.

[0089] It should be noted that, specifically, typhoon sensor networks, hydrological monitoring stations, detectors, and other equipment will be deployed in disaster-prone areas to ensure real-time data collection. A data processing and analysis platform will be built to integrate multi-source data and dynamically update scenario and user profiles. A content distribution system will be deployed to support multi-channel distribution and dynamic grading strategies. Disaster scenarios will be simulated to verify the system's data collection, analysis, and distribution capabilities. Finally, user feedback will be collected to evaluate the system's effectiveness and optimize the system based on this feedback data. User behavior data, including click-through rates, reading time, and feedback, will be collected in real time through tracking technology to form a complete user feedback dataset. User feedback data distributed across different channels (such as text messages, social media, and emergency broadcasts) will be unified to ensure data integrity and consistency.

[0090] Embed feedback mechanisms within distributed content so users can submit feedback directly when they encounter issues. Establish automated response templates or set up automatic confirmation mechanisms on user feedback platforms to ensure that users receive immediate system confirmation when leaving feedback. Use user feedback management mechanisms to categorize feedback, track trends, and identify frequently occurring issues. Leverage data analysis tools to extract valuable insights from massive amounts of data, providing clear guidance for product improvement.

[0091] Through the above mechanism, user feedback can be effectively collected, analyzed and applied to ensure the continuous optimization of emergency scenario content distribution strategies and improve user experience and information delivery effects.

[0092] One feasible implementation approach is to compare the effectiveness of different content strategies through A / B testing, combined with multivariate optimization (MVT) techniques to pinpoint the key factors influencing user experience. Based on the test results, traffic allocation between the experimental and control groups can be dynamically adjusted to ensure that high-potential content receives more exposure.

[0093] It's important to note that A / B testing is a comparative experimentation method, also known as split testing or bucket testing. It involves simultaneously displaying two or more versions (version A and version B, or even more) of a target variable (such as a webpage, product feature, or marketing strategy) to different user groups. By collecting and analyzing user feedback (such as click-through rate, conversion rate, and dwell time) on each version, the method determines which version performs better in achieving a specific goal (such as increasing sales or user engagement).

[0094] The present invention integrates multi-source disaster monitoring data to perceive the dynamic changes of disaster scenes in real time. It uses machine learning algorithms and real-time data access technology to dynamically update the label system of disaster scenes to ensure the timeliness and accuracy of information. It combines multiple channels such as text messages, social media, and emergency broadcasts, and selects the optimal distribution channel based on user portraits and scene portraits to ensure that information covers all users. Through user feedback data, the distribution strategy is dynamically adjusted and continuously optimized to ensure that the effect of information transmission is maximized. In an emergency, key information can be quickly and accurately delivered to all users, significantly improving the transmission efficiency and coverage of emergency information, and providing strong support for disaster emergency management and rescue work.

[0095] The present invention is described below by means of specific embodiments:

[0096] Example 1: Content distribution in a typhoon disaster scenario:

[0097] Data collection and preprocessing:

[0098] Deploy a typhoon sensor network in typhoon-prone areas to collect real-time data on typhoon coverage, level, typhoon eye center radius and other information.

[0099] Generate population heat maps using mobile communication base station data to analyze the distribution and flow trends of people in the disaster area.

[0100] Use social media crawler technology to capture public opinion data and extract public needs and emotions.

[0101] Scenario portrait and user portrait construction:

[0102] Construct a multi-dimensional labeling system for typhoon disaster scenarios, including disaster type (typhoon), impact range (accurate to the block), and vulnerability level (based on population density, building structure, etc.).

[0103] Dynamically divide target groups, assess users' risk exposure, and match their information needs.

[0104] Content distribution and optimization:

[0105] Use reinforcement learning algorithms to optimize content priority and ensure that red alert information is pushed first.

[0106] Distribute emergency evacuation guides through various channels including text messages, social media, and emergency broadcasts.

[0107] Collect user feedback, optimize distribution strategies, and improve distribution effects.

[0108] Example 2: Content distribution in flood disaster scenarios:

[0109] Data collection and preprocessing:

[0110] Hydrological monitoring stations are deployed in key areas such as rivers and reservoirs to collect data such as water level and rainfall in real time.

[0111] Combine GIS technology to generate flood risk maps and analyze the inundation range and risk areas.

[0112] Use social media public opinion data to extract public needs and emotions.

[0113] Scenario portrait and user portrait construction

[0114] Build a multi-dimensional labeling system for flood disaster scenarios, including disaster type (flood), impact range (accurate to the community), and vulnerability level (based on population density, infrastructure conditions, etc.).

[0115] Dynamically divide target groups, assess users' risk exposure, and match their information needs.

[0116] Content distribution and optimization:

[0117] Reinforcement learning algorithms are used to optimize content priorities to ensure that users in high-risk areas can obtain risk avoidance information in a timely manner.

[0118] Distribute flood warning information through various channels such as text messages, social media, and emergency broadcasts.

[0119] Collect user feedback, optimize distribution strategies, and improve distribution effects.

[0120] The above is an emergency scene content distribution method based on multimodal data fusion provided in an embodiment of the present invention. The following is an emergency scene content distribution system based on multimodal data fusion provided in an embodiment of the present invention.

[0121] See also Figure 2 , an emergency scenario content distribution system based on multimodal data fusion provided in an embodiment of the present invention includes:

[0122] The collection unit 201 is used to collect disaster monitoring data and pre-process the disaster monitoring data;

[0123] A construction unit 202 is configured to construct a multi-dimensional labeling system for disaster scenarios based on the pre-processed disaster monitoring data, and to update the labeling system for disaster scenarios in real time using a machine learning algorithm. The labeling system includes disaster type, impact range, and vulnerability level.

[0124] Matching unit 203 is used to dynamically divide users into target groups based on their location information, behavior patterns, and historical interaction data, evaluate users' risk exposure by combining the tag system and location information, and match users' information needs based on the group characteristics and risk exposure of the target groups;

[0125] The generation unit 204 is used to use a reinforcement learning algorithm to dynamically adjust the priority of the content according to the user's risk exposure and information needs, and determine the optimal distribution strategy based on the priority to distribute the emergency scenario content.

[0126] Furthermore, in one embodiment, the emergency scenario content distribution system based on multimodal data fusion provided by the present invention also includes: an optimization unit, which is used to collect user feedback data and optimize the optimal distribution strategy according to the feedback data.

[0127] Furthermore, an embodiment of the present invention also provides an emergency scenario content distribution device based on multimodal data fusion, the device including a processor and a memory:

[0128] The memory is used to store program code and transmit the program code to the processor;

[0129] The processor is used to execute the steps of the emergency scenario content distribution method based on multimodal data fusion as described in the above method embodiment according to the instructions in the program code.

[0130] Furthermore, an embodiment of the present invention also provides a computer-readable storage medium, which is used to store program code, and the program code is used to execute the emergency scenario content distribution method based on multimodal data fusion described in the above method embodiment.

[0131] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0132] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interface, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0133] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0134] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0135] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0136] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for distributing emergency scene content based on multimodal data fusion, characterized in that: include: Collecting disaster monitoring data and preprocessing the disaster monitoring data; Based on the pre-processed disaster monitoring data, a multi-dimensional labeling system for disaster scenarios is constructed, and the labeling system for disaster scenarios is updated in real time using a machine learning algorithm. The labeling system includes disaster type, impact range, and vulnerability level; Dynamically divide users into target groups based on their location information, behavior patterns, and historical interaction data, evaluate users' risk exposure by combining the tag system and location information, and match users' information needs based on the target group's characteristics and risk exposure; Utilizing a reinforcement learning algorithm, the priority of content is dynamically adjusted according to the risk exposure and information needs of the user, and an optimal distribution strategy is determined based on the priority to distribute emergency scenario content.

2. The emergency scenario content distribution method based on multimodal data fusion according to claim 1 is characterized in that: Also includes: Collect user feedback data, and optimize the optimal distribution strategy based on the feedback data.

3. The emergency scenario content distribution method based on multimodal data fusion according to claim 1 is characterized in that: The method further includes: constructing a multi-dimensional labeling system for disaster scenarios based on the pre-processed disaster monitoring data, and updating the labeling system for disaster scenarios in real time using a machine learning algorithm; and further including: By introducing the weighted fusion algorithm to fuse the disaster monitoring data, the disaster intensity index is obtained. ; Among them, the disaster intensity index The calculation expression is: ； Where, Indicates the typhoon intensity; T indicates the real-time temperature; d indicates the real-time flood peak water level; s indicates the typhoon-affected area; is the weighting coefficient, e is a constant; Represents the distance attenuation coefficient.

4. The emergency scenario content distribution method based on multimodal data fusion according to claim 3 is characterized in that: The method for determining the optimal distribution strategy further includes: By using a reinforcement learning algorithm, the priority of the content is dynamically adjusted according to the disaster intensity index, the risk exposure of the user and the information demand, and the optimal distribution strategy is determined according to the priority.

5. The emergency scenario content distribution method based on multimodal data fusion according to claim 1 is characterized in that: The machine learning algorithm is a neural network machine learning algorithm; Based on the neural network machine learning algorithm, historical data is trained to build a label prediction model for disaster scenarios. Multi-source data collected in real time is input into the trained label prediction model to predict the dynamic changes of disaster scenarios to obtain prediction results, and the label system is updated according to the prediction results.

6. The emergency scenario content distribution method based on multimodal data fusion according to any one of claims 1 to 5, characterized in that: The collecting of disaster monitoring data and pre-processing of the disaster monitoring data include: Collecting disaster monitoring data, the disaster monitoring data including: data including typhoon data and flood data; Among them, the typhoon data is obtained by collecting typhoon range data, level, and typhoon eye center radius information in real time through a typhoon sensor network, and removing noise using data cleaning and preprocessing technology. The flood data is obtained by integrating water level data, rainfall data, and flow rate data from hydrological monitoring stations.

7. An emergency scenario content distribution system based on multimodal data fusion, characterized in that: include: A collection unit, used for collecting disaster monitoring data and preprocessing the disaster monitoring data; A construction unit is configured to construct a multi-dimensional labeling system for disaster scenarios based on the pre-processed disaster monitoring data, and to update the labeling system for disaster scenarios in real time using a machine learning algorithm, wherein the labeling system includes disaster type, impact range, and vulnerability level; A matching unit, configured to dynamically divide users into target groups based on their location information, behavior patterns, and historical interaction data, evaluate the risk exposure of users in combination with the tag system and the location information, and match the information needs of users based on the group characteristics of the target groups and the risk exposure; The generation unit is used to use a reinforcement learning algorithm to dynamically adjust the priority of the content according to the user's risk exposure and information needs, and determine the optimal distribution strategy based on the priority to distribute emergency scenario content.

8. The emergency scenario content distribution system based on multimodal data fusion according to claim 7 is characterized in that: Also includes: The optimization unit is used to collect user feedback data and optimize the optimal distribution strategy according to the feedback data.

9. An emergency scene content distribution device based on multimodal data fusion, characterized in that: The device includes a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the emergency scene content distribution method based on multimodal data fusion as described in any one of claims 1-6 according to the instructions in the program code.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium is used to store program code, and the program code is used to execute the emergency scene content distribution method based on multimodal data fusion as described in any one of claims 1-6.

Images