Air-space-ground integrated public safety emergency command decision support system and method

Abstract

The invention relates to a space-air-ground integrated public safety emergency command decision support system and method, which can acquire multi-source data in the aspect of data acquisition and perception comprehensiveness, and can acquire event related information in an omnibearing and real-time manner. In the aspect of analysis and prediction accuracy, a traditional emergency command system lacks deep fusion and an advanced model and is combined with rich information after multi-source data deep fusion, so that a prediction result is more in line with an actual development trend. In the aspect of emergency decision optimization, traditional system decisions often give consideration to one another and are difficult to balance multiple targets. And the decision-making scientificity and effectiveness are greatly improved. In the aspect of command and dispatch collaboration, a traditional mode is unsmooth in communication and difficult in collaboration. The commanding, dispatching and collaboration layer of the new system realizes efficient information sharing and collaboration. And high efficiency and flexibility of public safety emergency command are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of public safety, and in particular to an air-space-ground integrated public safety emergency command and decision-making support system and method. Background Art

[0002] In the context of the current era of globalization and rapid technological development, the public safety field faces unprecedented complex challenges and opportunities. On the one hand, the acceleration of urbanization has led to a high population density, and the emergence of large urban complexes, transportation hubs and other places where people gather has greatly increased the probability of public safety incidents and the potential scope of influence. For example, once a fire or an emergency occurs in a bustling commercial center, the resulting casualties and property losses will be immeasurable.

[0003] On the other hand, with the popularization and progress of technology, various new technologies and new devices have been widely used, which, while bringing convenience to people's lives, have also triggered many new public safety problems. For example, the rapid development of the Internet and Internet of Things technologies, although realizing the interconnection of all things, has also led to frequent network security incidents such as cyber attacks and information leakage. These incidents may not only violate personal privacy, but also cause serious damage to critical infrastructure (such as power systems, financial systems, etc.), thereby threatening the stable operation of the entire society.

[0004] Furthermore, climate change and ecological environment changes have exacerbated the frequency and severity of natural disasters. Natural disasters such as strong typhoons, rainstorms and floods, earthquakes, forest fires, etc. frequently rage, bringing huge losses of life and property to human society. For example, in recent years, forest fires have occurred globally, burning large areas of forest resources, destroying the ecological balance, and at the same time threatening the lives and living orders of surrounding residents.

[0005] However, traditional public safety emergency command and decision-making support systems have exposed many limitations when dealing with these complex and changeable public safety incidents. Their data collection methods are relatively single, mainly relying on ground monitoring equipment and manual reports, and it is difficult to obtain comprehensive, real-time and accurate information. For example, in remote areas or large-scale disaster sites, ground monitoring cannot effectively cover, resulting in serious information gaps. At the same time, the analysis methods and mathematical models adopted by traditional systems are relatively simple, and it is impossible to deeply explore the complex relationships between multi-source data, making it difficult to accurately predict public safety incidents and make scientific decisions. In the face of public safety incidents with intertwined multi-factors and high uncertainty, traditional systems often seem powerless and cannot meet the requirements of modern society for the efficiency, accuracy and intelligence of public safety emergency management. Therefore, there is an urgent need to develop an innovative air-space-ground integrated public safety emergency command and decision-making support system to integrate multi-source data resources and use advanced mathematical models and algorithms to improve the effectiveness and level of public safety emergency command. Summary of the Invention

[0006] The present invention provides an integrated air, space, and ground public safety emergency command and decision-making support system, including:

[0007] A multi-source data acquisition layer for collecting public safety-related data from the air, ground, and space. The air data acquisition includes collecting data by drones equipped with monitoring devices and receiving satellite remote sensing data. The ground data acquisition includes collecting data from a network of surveillance cameras, a sensor array, and intelligent terminals. The space data acquisition relies on high-resolution observation satellites;

[0008] A data preprocessing and fusion layer for cleaning, denoising, format conversion, data alignment, and feature extraction of the collected multi-source heterogeneous data, and using a fusion algorithm based on the fusion of a hypergraph model and a deep neural network (HG-DNN) to construct a hypergraph structure from different modal feature vectors and perform deep fusion to obtain a fused feature representation;

[0009] An intelligent analysis and prediction layer, based on a quantum-inspired probabilistic graph model (QIPGM), regarding public safety-related factors as quantum states to construct a quantum probability graph and determine a quantum state transition probability matrix, and performing a measurement operation on the quantum probability graph through the quantum measurement principle to predict the occurrence probability, development trend, and possible results of public safety events;

[0010] An emergency decision-making optimization layer, based on the results of intelligent analysis and prediction, formulating emergency decision-making strategies in combination with an emergency decision-making optimization model based on a multi-objective evolutionary algorithm and dynamic game theory. The multi-objective evolutionary algorithm is used for searching multi-objective decision-making schemes, and the dynamic game theory is used for analyzing the interest conflicts and dynamic interactions among emergency rescue entities;

[0011] A command and dispatch and coordination layer that provides a command and dispatch platform for emergency command personnel, realizes unified dispatching of emergency resources and management of coordinated work among emergency rescue entities, and supports manually adjusting and optimizing command and dispatch orders according to actual situations, as well as information sharing and communication collaboration.

[0012] Furthermore, the devices carried by the drones in the multi-source data acquisition layer include one or more of a high-definition camera, an infrared thermal imager, and a gas sensor, and are used to obtain image, video, and environmental parameter data.

[0013] Furthermore, the preprocessing of satellite remote sensing image data in the data preprocessing and fusion layer includes image enhancement, geometric correction, band fusion processing, and extraction of image feature vectors such as ground object features and disaster features.

[0014] Furthermore, the quantum state transition probability matrix in the quantum-inspired probabilistic graph model (QIPGM) is dynamically updated and optimized based on historical data, expert knowledge, and real-time data.

[0015] On the other hand, the present invention also provides a method for integrated air, space and ground public safety emergency command and decision-making support, including the following steps:

[0016] Multi-source data collection step: using the multi-source data collection layer to widely collect public safety-related data from the air, ground and space; data preprocessing and fusion step: preprocessing the collected raw data and performing deep fusion of multi-source data using the HG-DNN fusion algorithm to obtain a fused feature representation;

[0017] Intelligent analysis and prediction step based on the quantum-inspired probabilistic graphical model (QIPGM): constructing a QIPGM model to predict the occurrence probability, development trend and possible results of public safety events;

[0018] Emergency decision-making optimization step: formulating an emergency decision-making strategy using an emergency decision-making optimization model based on the multi-objective evolutionary algorithm and dynamic game theory according to the results of intelligent analysis and prediction;

[0019] Command, dispatch and coordination step: through the command, dispatch and coordination layer, emergency command personnel issue command and dispatch orders based on the information, realize the allocation of emergency resources and the coordination of emergency rescue entities, and can adjust and optimize the decision-making in real time.

[0020] Furthermore, in the multi-source data collection step, different types of unmanned aerial vehicles are deployed according to the characteristics and requirements of the monitoring area by the unmanned aerial vehicles in the air to obtain corresponding data, and at the same time, different types of satellite data are received and a ground data collection network is established to collect data from monitoring cameras, sensors and intelligent terminals.

[0021] Furthermore, when constructing the hypergraph structure in the data preprocessing and fusion step, different modal feature vectors are used as nodes, and hyperedges are constructed according to the data logical relationship to connect multiple relevant feature vectors.

[0022] Furthermore, in the intelligent analysis and prediction step based on the quantum-inspired probabilistic graphical model (QIPGM), the quantum state includes relevant factors such as public safety event types, occurrence locations, environmental conditions, and social impact factors.

[0023] Furthermore, in the emergency decision-making optimization step, the decision-making objectives of the multi-objective evolutionary algorithm include minimizing casualties, minimizing property losses, maximizing emergency response efficiency, and minimizing resource consumption.

[0024] Furthermore, in the command, dispatch and coordination step, the command and dispatch platform supports location sharing, task assignment and collaborative operations, as well as information communication and resource allocation coordination among different emergency rescue entities.

[0025] Beneficial effects:

[0026] In terms of comprehensive data collection and perception, multi-source data collection can obtain event-related information in all directions and in real time. In terms of analysis and prediction accuracy, the traditional emergency command system lacks deep integration and advanced models. Combined with the rich information after deep integration of multi-source data, the prediction results are more in line with the actual development trend. In terms of emergency decision-making optimization, traditional system decisions often lose sight of one thing while focusing on another, and it is difficult to balance multiple goals. The scientificity and effectiveness of decision-making have been greatly improved. In terms of command and dispatch coordination, the traditional model has poor communication and difficulty in coordination. The command, dispatch and coordination layers of the new system realize efficient information sharing and collaboration. It ensures the efficiency and flexibility of public safety emergency command. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Schematic diagram of the algorithm flow. DETAILED DESCRIPTION

[0028] Example 1

[0029] An air-ground-integrated public safety emergency command decision support system mainly consists of a multi-source data acquisition layer, a data preprocessing and fusion layer, an intelligent analysis and prediction layer, an emergency decision optimization layer, and a command dispatch and coordination layer.

[0030] The multi-source data collection layer collects public safety-related data from the air, space and ground through various means. In the air, drones equipped with high-definition cameras, infrared thermal imagers, gas sensors and other equipment are used to conduct flexible low-altitude monitoring of specific areas to obtain high-resolution images, videos, environmental parameters and other data; with the help of satellite remote sensing technology, data from different types of satellites such as meteorological satellites, resource satellites, and reconnaissance satellites are received, including weather change information, geographic environment information, personnel gathering and movement information (obtained through specific algorithm analysis of satellite images), etc. On the ground, a large-scale surveillance camera network is deployed to cover key areas such as urban streets, public places, and transportation hubs to collect real-time video data; various sensors such as earthquake sensors, fire alarms, and environmental quality sensors are set up to monitor physical environmental parameters; at the same time, information uploaded from ground smart terminals (such as mobile phones, IoT devices, etc.) is integrated, such as abnormal events reported by the public, location information, etc. In space, high-resolution optical observation satellites and radar satellites are used to obtain more macro and accurate information on the earth's surface, such as the status of large infrastructure, the scope and extent of natural disasters, etc.

[0031] The data preprocessing and fusion layer performs preprocessing operations such as cleaning, denoising, format conversion, data alignment, and feature extraction on the collected massive multi-source heterogeneous data to improve data quality and convert it into a format suitable for model processing. Different preprocessing techniques are adopted for different types of data. For example, for satellite remote sensing image data, image enhancement, geometric correction, band fusion, etc. are performed to extract image feature vectors such as land feature and disaster features (such as flood inundation area, forest fire burned area, etc.); for data collected by drones, data calibration, target recognition (such as recognizing targets like people, vehicles, fire sources, etc.), trajectory analysis, etc. are performed to extract drone feature vectors such as target features and behavior features; for ground surveillance video data, video decoding, target detection and tracking, scene analysis, etc. are performed to extract video feature vectors such as human behavior features, object movement trajectory features, scene change features; for sensor data, data calibration, outlier detection and removal, data interpolation, etc. are performed to extract sensor feature vectors such as environmental parameter change features and equipment operating state features; for ground intelligent terminal data, information screening, location verification, data classification, etc. are performed to extract terminal feature vectors such as event description features and location features. After feature extraction, a fusion algorithm based on the fusion of hypergraph model and deep neural network (HG-DNN) is used to construct the feature vectors of different modalities into a hypergraph structure, where hyperedges connect multiple feature vectors with relevant relationships (breaking through the limitation that an edge in a traditional graph model can only connect two nodes and better reflecting the complex correlation relationships between multi-source data), and then the deep neural network is used to learn and fuse the hypergraph structure data to mine the potential correlation relationships between different modality data, realize the deep fusion of multi-source data, obtain the fused feature representation, and provide a rich and comprehensive data basis for subsequent intelligent analysis and prediction.

[0032] The intelligent analysis and prediction layer conducts intelligent analysis and prediction of public security events based on an innovative quantum-inspired probabilistic graph model (QIPGM). This model draws on the probability concept in quantum mechanics and the structural advantages of graph models, regarding public security-related factors (such as event types, event occurrence locations, environmental conditions, social impact factors, etc.) as quantum states, and describing the mutual relationships and uncertainties among these factors by constructing a quantum probability graph. In the model, a transition probability matrix between quantum states is defined, which is dynamically updated and optimized based on a large amount of historical data, expert knowledge, and real-time data to reflect the dynamic change laws in the development process of public security events. For example, in the prediction of fire events, factors such as the fire occurrence location, fire intensity, wind direction, surrounding building types, and population density are regarded as quantum states, and their mutual influence relationships are described by the quantum probability graph, such as the relationship between fire intensity and wind direction, surrounding building types, and the relationship between population density and fire occurrence location, fire intensity, etc., and the probabilities of these relationships are continuously adjusted according to real-time monitoring data. Using the quantum measurement principle, measurement operations are performed on the quantum probability graph to obtain prediction information such as the probability of public security events occurring, development trends (such as the expansion or contraction trend of event scale, the diffusion direction and speed of the impact range, etc.), and possible results (such as the evolution results of events under different response strategies). By introducing a quantum-inspired mechanism, this model can better handle the uncertainties and ambiguities in public security data, improve the accuracy and reliability of predictions, especially showing unique advantages when facing complex, changeable, and information-incomplete public security events.

[0033] The emergency decision-making optimization layer formulates emergency decision-making strategies based on the results output by the intelligent analysis and prediction layer and in combination with an emergency decision-making optimization model based on multi-objective evolutionary algorithms and dynamic game theory. The multi-objective evolutionary algorithm takes into account multiple conflicting decision-making objectives, such as minimizing casualties, minimizing property losses, maximizing emergency response efficiency, minimizing resource consumption, etc. By simulating genetic, mutation, and selection operations in the process of biological evolution, it searches for the optimal emergency decision-making plan in a vast decision-making space. For example, in the emergency decision-making for earthquake disasters, the decision variables may include the deployment direction and quantity of rescue teams, the distribution locations and types of rescue supplies, the layout and scheduling of medical resources, etc. The multi-objective evolutionary algorithm gradually optimizes each objective function by continuously evolving the decision-making plans in the population. At the same time, dynamic game theory is introduced to consider the interest conflicts and cooperation relationships among different emergency rescue entities (such as government departments, rescue teams, volunteer organizations, etc.), as well as the dynamic interaction between public safety incidents and emergency rescue operations. For example, in dealing with emergencies, there is a game relationship between the government department and the adversaries. The decisions of the government department (such as police force deployment, negotiation strategies, etc.) will affect the actions of the adversaries, and the actions of the adversaries will in turn affect the subsequent decisions of the government department. Through the dynamic game theory model, this complex dynamic interaction process can be analyzed, and more targeted, flexible, and robust emergency decision-making strategies can be formulated to cope with various possible situation changes.

[0034] The command, dispatch, and coordination layer provides an intuitive and convenient command and dispatch platform for emergency commanders, realizing the unified dispatch and command of various emergency resources (such as rescue teams, fire-fighting equipment, medical supplies, transportation tools, etc.), as well as the collaborative work management among different emergency rescue entities. Through this platform, commanders can view the situation awareness results of public safety incidents, intelligent analysis and prediction information, and emergency decision-making strategy suggestions in real time, make manual adjustments and optimizations according to the actual situation, and issue command and dispatch orders. At the same time, the platform supports information sharing and communication collaboration among different emergency rescue entities, such as location sharing, task assignment, and collaborative operations among rescue teams, and information communication and resource allocation collaboration between government departments and volunteer organizations, improving the overall efficiency and collaboration of emergency rescue work.

[0035] On the other hand, the present invention also provides a space-air-ground integrated public safety emergency command and decision-making support method, including the following steps:

[0036] Step 1: Multi-source data collection. Use the multi-source data collection layer to widely collect public safety-related data from multi-source channels such as air, ground, and space, ensuring the comprehensiveness, real-time nature, and accuracy of the data, and providing an adequate data foundation for subsequent analysis. Step 2: Data preprocessing and fusion. Conduct comprehensive preprocessing operations on the collected raw data, remove invalid data and noise interference, standardize and integrate data in different formats and types, extract key features of various types of data, and then use a fusion algorithm based on the integration of hypergraph model and deep neural network (HG-DNN) for in-depth multi-source data fusion to obtain the fused feature representation.

[0037] Step 3: Intelligent analysis and prediction based on the quantum-inspired probabilistic graph model (QIPGM). Construct the quantum-inspired probabilistic graph model, regard public safety-related factors as quantum states, determine the mutual relationships and transition probability matrices between quantum states, describe the development process of public safety events through the quantum probability graph, and use the quantum measurement principle for prediction operations to obtain prediction information such as the probability of public safety events occurring, development trends, and possible outcomes.

[0038] Step 4: Emergency decision-making optimization. Based on the results of intelligent analysis and prediction, use an emergency decision-making optimization model based on multi-objective evolutionary algorithm and dynamic game theory to formulate emergency decision-making strategies. The multi-objective evolutionary algorithm searches for the optimal decision-making plan under multi-objective constraints, and the dynamic game theory considers the interest conflicts and dynamic interactions between different parties, and comprehensively obtains the best emergency decision-making strategy. Step 5: Command, dispatch, and coordination. Through the command, dispatch, and coordination layer, emergency command personnel issue command and dispatch orders according to the intelligent analysis and prediction information and emergency decision-making strategy suggestions, combined with the actual situation, to achieve the unified allocation of emergency resources and the collaborative work management between different emergency rescue parties. At the same time, the decision-making can be adjusted and optimized in real time according to the on-site feedback.

[0039] By widely collecting multi-source heterogeneous data from air, ground, and space through the multi-source data collection layer and realizing in-depth data fusion with the fusion algorithm based on the integration of hypergraph model and deep neural network (HG-DNN), the potential correlation relationships between different data sources can be fully explored, and information resources in multiple aspects can be integrated, so as to achieve a comprehensive perception of public safety events. For example, in forest fire monitoring, satellite remote sensing data can provide information on the macroscopic scope of the fire and the direction of fire spread, UAV data can supplement details of the fire situation in local areas, smoke concentration, and surrounding terrain and landform information, and ground monitoring and sensor data can obtain information on the evacuation of people in nearby settlements and the status of fire-fighting facilities. Integrating these data can more comprehensively and accurately grasp the fire situation and provide a more complete information basis for emergency command.

[0040] Intelligent analysis and prediction using the Quantum-inspired Probabilistic Graphical Model (QIPGM) can effectively handle the uncertainty and ambiguity in public security data. By treating public security-related factors as quantum states and constructing a quantum probability graph, the complex relationships and dynamic change laws between factors can be described more precisely, and more accurate prediction information can be obtained using the principle of quantum measurement. For example, when dealing with sudden infectious disease outbreaks, the model can consider the uncertainty of virus transmission (such as the possibility of mutation, the diversity of transmission routes, etc.), the ambiguity of population susceptibility (the difference in resistance to the virus among different populations is difficult to precisely quantify), and the uncertainty of the effectiveness of prevention and control measures (such as the implementation intensity of isolation policies, the coverage rate and effectiveness of vaccination, etc.), so as to more accurately predict the spread trend of the epidemic, the peak number of infected people, and the development results of the epidemic under different prevention and control strategies, providing strong support for formulating scientific and reasonable prevention and control decisions.

[0041] The model based on multi-objective evolutionary algorithm and dynamic game theory in the emergency decision-making optimization layer can comprehensively consider multiple conflicting decision-making objectives and the interest conflicts and dynamic interaction relationships among different emergency rescue entities. The multi-objective evolutionary algorithm searches for the optimal solution in the vast decision-making space, balancing various factors such as casualties, property losses, emergency response efficiency, and resource consumption; dynamic game theory analyzes the strategy choices and mutual influences of different entities during the emergency process, making the decision more targeted, flexible, and robust. For example, in the emergency decision-making of urban floods, the multi-objective evolutionary algorithm can determine the best rescue material allocation plan, personnel evacuation route planning, and flood control facility activation strategy, etc. At the same time, dynamic game theory can consider the interaction relationships among government departments, rescue teams, affected people, and surrounding areas, such as the game between government departments and affected people in the distribution of rescue materials, the trade-off in flood control resource sharing and cooperation among different regions, etc., so as to formulate emergency decision-making strategies that are more in line with the actual situation and more capable of dealing with complex changes.

[0042] The command, dispatch, and coordination layer provides a convenient platform for emergency command personnel, realizing the unified dispatch of emergency resources and the efficient coordination among different emergency rescue entities. Through this platform, command personnel can grasp the event situation and decision-making information in real time, conduct precise command, and promote information sharing and cooperation among various entities. For example, in large-scale earthquake rescue, command personnel can quickly allocate rescue teams, medical supplies, and engineering equipment from all over according to the on-site situation, coordinate the work among different departments such as the military, fire, medical, and volunteer organizations, avoid resource waste and duplicate labor, improve rescue efficiency, and minimize disaster losses to the greatest extent.

[0043] Establish an integrated space-air-ground data acquisition network in different public safety monitoring areas, such as cities, mountainous areas, coastal areas, etc. In the air, deploy different types of unmanned aerial vehicles (UAVs) according to the characteristics and requirements of the monitoring area. For example, in urban areas, use small multi-rotor UAVs equipped with high-definition cameras and environmental sensors to conduct daily patrol and monitoring of urban blocks, large event venues, etc.; in mountainous or forest areas, dispatch long-endurance fixed-wing UAVs equipped with infrared thermal imagers and high-resolution optical cameras to monitor large areas for fires, geological disasters, etc. At the same time, establish a connection with satellite data receiving stations to receive various satellite data. According to the orbital parameters and data transmission protocols of the satellites, regularly receive meteorological data from meteorological satellites (such as cloud distribution, precipitation probability, temperature changes, etc.), geographical environment data from resource satellites (such as topography, vegetation cover, land use types, etc.), and specific target monitoring data from reconnaissance satellites (such as personnel gathering points, military facilities, etc. in the event of national security or major public safety incidents). On the ground, build a large-scale network of surveillance cameras, and reasonably set the positions, angles, and resolutions of the cameras according to the urban planning and key public safety areas to ensure full coverage of key areas. Deploy various sensors. According to the potential risk types of the area, such as earthquake sensors in earthquake-prone areas and harmful gas sensors in chemical industrial parks, etc., and establish a sensor data transmission network to transmit the collected data to the data processing center in real time. At the same time, develop intelligent terminal applications for the public, encourage the public to report information in a timely manner when discovering public safety incidents or abnormal situations, upload the data to the system through the mobile network, and verify and classify the reported information.

[0044] Preprocess the collected raw data. For satellite remote sensing image data, first perform image enhancement processing, such as histogram equalization, contrast stretching, etc., to improve the clarity and recognizability of the image; then perform geometric correction, and correct the geometric deformation of the image according to factors such as the satellite's orbital parameters and the curvature of the earth; then perform band fusion, fuse multi-band remote sensing images, and extract richer ground object feature information. For example, fusing the visible light band and the infrared band can simultaneously obtain the color, texture, and temperature information of ground objects, etc. Finally, extract the image feature vectors, such as ground object edge features, texture features, disaster features (such as the boundary of the flood inundation area, the high-temperature area of the forest fire, etc.). For the data collected by drones, perform data calibration, and correct the collected data according to factors such as the flight attitude of the drone and the installation position of the sensor; then perform target recognition, and use deep learning target detection algorithms (such as the YOLO algorithm or the Faster R-CNN algorithm based on convolutional neural networks) to identify targets such as people, vehicles, and fire sources in images and videos, and extract the features of the targets (such as the size, shape, color, and movement speed of the targets) and behavioral features (such as the aggregation, running direction of people, and driving trajectory of vehicles). For ground surveillance video data, perform video decoding to convert the video stream into an image sequence; then perform target detection and tracking. Similarly, use deep learning target detection algorithms to identify the targets in the video, and use target tracking algorithms (such as the Kalman filter algorithm or the particle filter algorithm) to track the movement trajectory of the targets, and extract human behavior features (such as behaviors such as people wandering, fighting, and abnormal aggregation), object movement trajectory features (such as illegal driving of vehicles, falling trajectories of objects, etc.), and scene change features (such as changes in light in the scene, addition or disappearance of objects, etc.). For sensor data, perform data calibration, and correct the collected data according to the calibration curve of the sensor and the standard reference value; then perform outlier detection and removal, and use statistical methods (such as the outlier judgment method based on the mean and standard deviation) or machine learning outlier detection models (such as the outlier detection model based on support vector machines) to identify and remove data points that significantly deviate from the normal range; then perform data interpolation, and for a small number of missing data points, use appropriate interpolation methods (such as linear interpolation or spline interpolation) for supplementation; finally, extract environmental parameter change features (such as sudden increase in temperature, change trend of harmful gas concentration, etc.), equipment operation state features (such as battery power and signal strength of the sensor). For ground intelligent terminal data, perform information screening to remove false information and duplicate information; then perform location verification, and verify the location accuracy of the reported event by comparing with geographic information system (GIS) data; finally, perform data classification, and classify and organize the data according to the type of the event (such as traffic accidents, fires, public security events, etc.), and extract event description features (such as the severity of the event, the number of people involved, etc.) and location features (such as the specific address where the event occurred, the area where it is located, etc.).After feature extraction, the feature vectors of different modalities are constructed into a hypergraph structure. For example, the satellite image feature vector, drone feature vector, ground surveillance video feature vector, sensor feature vector, and ground intelligent terminal feature vector are used as nodes of the hypergraph. According to the logical relationship between the data (such as the fire area in the satellite image is related to the fire details monitored by the drone, the increase in the concentration of harmful gases in the ground sensor is related to the abnormal smoke in the ground surveillance video, etc.), hyperedges are constructed to connect multiple related feature vectors to form hypergraph structure data. Then, a deep neural network is used to learn and fuse the hypergraph structure data. The deep neural network can adopt a multi-layer fully connected neural network or a combination of a convolutional neural network and a recurrent neural network. Through the training of the hypergraph structure data, the potential correlation between different modal data is mined, the deep fusion of multi-source data is realized, and the fused feature representation is obtained.

[0045] In terms of data collection and comprehensive perception, traditional systems mainly rely on limited ground monitoring and manual reporting, with a single source of information and a limited scope. The integrated air-ground-space system can obtain event-related information in a comprehensive and real-time manner through satellite, drone, and ground multi-source data collection. For example, in a natural disaster scenario, the traditional system may not know the local situation until some time after the disaster occurs, while the new system can quickly grasp the overall picture of the disaster-stricken area, the surrounding environment, and potential risk factors. The scope of data collection has been expanded several times, and the comprehensiveness of perception has been significantly improved.

[0046] In terms of analysis and prediction accuracy, traditional emergency command systems lack deep integration and advanced models, and are mostly based on experience and judgment, resulting in large errors. The quantum-inspired probabilistic graph model of the present invention can accurately handle uncertainty, and combined with the rich information after deep fusion of multi-source data, the prediction results are more in line with the actual development trend. For example, when responding to public health events, traditional systems find it difficult to predict the direction of transmission and the effect of prevention and control. The new system can accurately predict the peak of the epidemic, changes in the transmission path, etc., and the prediction accuracy is about 30%-50% higher than the traditional method.

[0047] In terms of emergency decision-making optimization, traditional system decisions often lose sight of one thing while focusing on another, and it is difficult to balance multiple objectives. The multi-objective evolutionary algorithm and dynamic game theory model of the integrated air-ground-space system can comprehensively consider multiple factors and take into account the conflicts of interest of all parties. In urban emergencies, traditional decision-making may lead to waste of resources or untimely rescue. The new system can reasonably allocate resources, increase rescue efficiency by about 40%, reduce resource waste by about 35%, and greatly improve the scientificity and effectiveness of decision-making.

[0048] In terms of command and dispatch coordination, there are problems of poor communication and difficult coordination in the traditional mode. The command, dispatch and coordination layer of the new system realizes efficient information sharing and collaboration. For example, in large-scale accident rescue, the cooperation among rescue teams in the traditional system is loose, while the new system enables all parties to cooperate closely, shortening the task completion time by about 30%, significantly accelerating the overall emergency response speed, and ensuring the efficiency and flexibility of public safety emergency command.

[0049] Embodiment 2

[0050] An integrated space-air-ground public safety emergency command and decision support system mainly consists of the following key parts: Multi-source data acquisition subsystem: In the air, unmanned aerial vehicle (UAV) swarms are used for flexible data acquisition. The UAVs are equipped with high-resolution optical cameras, infrared thermal imagers, multi-spectral sensors, and meteorological detection equipment, etc., which can conduct detailed image shooting, temperature monitoring, target recognition, and meteorological data collection for specific areas, and are applicable to scenarios such as fire monitoring, personnel search and rescue, and investigation of illegal activities. Data is transmitted to the ground control center in real time through satellite communication links to ensure data timeliness. At the same time, data from multiple satellites with different orbits and functions are received, including surface image information provided by high-resolution optical satellites for identifying abnormal changes in large-scale infrastructure, land use changes, etc.; surface deformation data obtained by synthetic aperture radar satellites, which can be used for early warning of geological disasters such as earthquakes and landslides; meteorological data of meteorological satellites, such as wind speed, precipitation, and temperature, providing a basis for environmental analysis of public safety events. On the ground, a wide network of surveillance cameras is deployed to cover key areas such as urban streets, public places, and transportation hubs to collect real-time video data; various sensors are set up, such as environmental sensors (detecting air quality, water quality, noise, etc.), safety sensors (fire alarms, intrusion detectors, etc.), and geographical information sensors (for monitoring crustal movement, groundwater level changes, etc.); and information from the public's mobile terminals, Internet of Things devices, etc. is integrated, such as emergencies reported by the public, location information, on-site photos or videos.

[0051] Data Preprocessing and Fusion Subsystem: Perform a series of preprocessing operations on the collected massive multi-source heterogeneous data. For satellite image data, perform radiation correction, geometric correction, image enhancement, etc., and extract feature points, edge information, texture features, and feature vectors such as the shape and size of target objects in the image; for data collected by drones, perform data calibration, denoising, target classification and recognition, and extract drone feature vectors such as the motion trajectory, speed, and behavior characteristics of the target; ground surveillance video data undergoes processing such as decoding, target detection and tracking, and scene analysis to obtain video feature vectors such as human behavior characteristics, object motion state characteristics, and scene change characteristics; sensor data undergoes operations such as data cleaning, outlier detection and repair, and data standardization, and extracts sensor feature vectors such as environmental parameter change characteristics and equipment operation state characteristics; ground intelligent terminal data undergoes processing such as information screening, authenticity verification, and data format conversion, and extracts terminal feature vectors such as event description characteristics and location characteristics. Then, adopt a fusion algorithm based on the integration of manifold learning and Bayesian network (ML-BN). First, use the manifold learning algorithm to mine the inherent low-dimensional manifold structure of different modal data, map the high-dimensional data to a low-dimensional space, reveal the essential features and potential relationships of the data, and then use the low-dimensional manifold data as the input nodes of the Bayesian network to construct a Bayesian network model. Through the probability inference ability of the Bayesian network, based on the conditional probability relationship between the data, achieve the deep fusion of multi-source data in the probability framework, obtain the fused feature representation, and provide a comprehensive and reliable data basis for subsequent analysis and decision-making.

[0052] Intelligent Analysis and Prediction Subsystem: An intelligent analysis and prediction model for public security events based on a hybrid of an innovative random forest and hidden Markov model (RF-HMM). First, the random forest algorithm is used to perform preliminary classification and feature importance assessment on the fused feature data. The random forest consists of multiple decision trees. Through the training of a large number of decision trees and the voting mechanism, it can effectively handle the complex relationships and noise interference in multi-source data, determine the importance weights of different features for judging the types of public security events, and conduct preliminary classification of events, such as distinguishing major categories like natural disasters (earthquakes, floods, fires, etc.), social security events (emergency incidents, mass incidents, etc.), public health events (infectious disease epidemics, food safety incidents, etc.). Then, the classification results and feature weight information of the random forest are used as part of the observation sequence and emission probability of the hidden Markov model (HMM) and input into the HMM. Based on the temporal characteristics of event development, the HMM constructs a state transition matrix and an observation probability matrix. By learning historical data and real-time data, it predicts the transition probabilities between different states of public security events and future development trends, such as the expansion or contraction of the event scale, changes in the affected area, and estimation of the duration. For example, in the prediction of infectious disease epidemics, the HMM can predict the transition probabilities at different stages (incubation period, outbreak period, control period, dissipation period) of the epidemic and key information such as the peak number of infections and changes in the transmission range in the future based on observation information such as the number of infected people, transmission area, and prevention and control measures at the initial stage of the epidemic, providing a forward-looking basis for emergency decision-making.

[0053] Emergency Decision Optimization Subsystem: Based on the results output by the Intelligent Analysis and Prediction Subsystem, an emergency decision-making strategy is formulated in combination with the emergency decision optimization model based on the particle swarm optimization algorithm and multi-agent system. The particle swarm optimization algorithm is used to search for the optimal emergency resource allocation plan in a complex decision space. Emergency resources (such as rescue teams, medical supplies, fire-fighting equipment, transport vehicles, etc.) are regarded as particles, and each particle represents a possible resource allocation combination plan. According to the emergency goals (such as minimizing casualties, minimizing property losses, maximizing rescue efficiency, etc.), the fitness function of the particle is set. The particle swarm continuously updates its own speed and position in the decision space, draws on the experience of individual and group optima, and gradually converges to the optimal resource allocation plan. At the same time, a multi-agent system is introduced, and different emergency rescue entities (such as government departments, professional rescue teams, volunteer organizations, medical units, etc.) are regarded as agents. Each agent has its own goals, capabilities, resources, and decision-making rules. Through information interaction, cooperation, and competition mechanisms among agents, the multi-party cooperation and game relationships in the actual emergency rescue process are simulated. For example, in earthquake disaster rescue, the government department agent is responsible for overall coordination and resource allocation decisions, the rescue team agent executes rescue tasks according to its own professional skills and equipment conditions, the medical unit agent provides medical treatment services and feedback on the situation of the wounded, and the volunteer organization agent assists in rescue work and material distribution, etc. Under the resource allocation framework determined by the particle swarm optimization algorithm, the multi-agent system further optimizes the action strategies and cooperation methods of each agent, improving the flexibility, adaptability, and overall effectiveness of emergency decision-making.

[0054] Command, Dispatch and Collaboration Subsystem: Provide a powerful and easy-to-operate command and dispatch platform for emergency commanders. The platform can intuitively display the real-time situation awareness results of public safety incidents, including information such as the incident location, impact scope, and development trend, presented in various visualization methods such as maps, charts, and 3D models. At the same time, it displays the event prediction information provided by the Intelligent Analysis and Prediction Subsystem and the decision-making strategy suggestions generated by the Emergency Decision Optimization Subsystem, facilitating commanders to quickly understand the overall situation of the incident and make scientific decisions. Commanders can perform real-time allocation operations of emergency resources on the platform, such as dispatching rescue teams, allocating materials, arranging transport vehicles, etc., and issue detailed task instructions to each emergency rescue entity. The platform supports information sharing and communication collaboration among different emergency rescue entities. By establishing unified communication standards and data interfaces, it realizes position sharing and task collaboration among rescue teams, instruction transmission and information feedback between government departments and other entities, and exchange of wounded information between medical units and rescue teams, etc., ensuring the efficient collaborative progress of emergency rescue work. In addition, the platform also has the functions of emergency drills and simulations, which can conduct simulation drills according to different public safety incident scenarios and plans, evaluate the feasibility and effectiveness of the plans, and provide experience and reference for actual emergency command.

[0055] On the other hand, the present invention also provides a method for integrated air, space and ground public safety emergency command decision-making support, comprising the following steps:

[0056] Step 1: Multi-source data collection. Collect public safety-related data from the air, ground and space in all directions through a multi-source data collection subsystem to ensure the comprehensiveness, real-time nature and accuracy of the data, laying a solid foundation for subsequent analysis and processing.

[0057] Step 2: Data preprocessing and fusion. Perform systematic preprocessing operations on the collected raw data, including data cleaning, feature extraction, etc., and then use a fusion algorithm based on the integration of manifold learning and Bayesian network (ML-BN). First, mine the internal manifold structure of the data, and then use the Bayesian network for probability fusion to obtain the fused feature representation.

[0058] Step 3: Intelligent analysis and prediction based on the RF-HMM model. Construct a hybrid model of random forest and hidden Markov model (RF-HMM). First, use the random forest for data classification and feature importance evaluation, and then input the results into the hidden Markov model to predict the development trend and key information of public safety events according to the temporal characteristics of the events.

[0059] Step 4: Emergency decision optimization. According to the prediction results, adopt an emergency decision optimization model based on the particle swarm optimization algorithm and multi-agent system. Search for the optimal resource allocation plan through the particle swarm optimization algorithm, and the multi-agent system optimizes the action strategies and cooperation methods of each subject to formulate a scientific and reasonable emergency decision-making strategy.

[0060] Step 5: Command, dispatch and coordination. With the help of the command, dispatch and coordination subsystem, emergency command personnel perform command and dispatch operations on the platform according to the situation awareness, prediction information and decision-making strategy suggestions, realize the efficient allocation of emergency resources and the coordinated work of each emergency rescue subject, and can use the exercise and simulation functions of the platform for plan evaluation and experience accumulation.

[0061] Deep integration and comprehensive perception of multi-source data: The multi-source data acquisition subsystem widely collects multi-source heterogeneous data from the air, space, and ground. With the fusion algorithm based on manifold learning and Bayesian network fusion (ML-BN), it can fully explore the internal structure and potential relationships of the data, achieving deep integration of multi-source data. For example, in forest fire monitoring, satellite data can provide the large-scale spread range of the fire and the direction of smoke diffusion, UAV data can obtain the fire intensity, fire source location, and surrounding terrain and geomorphic information in a local area, and ground monitoring and sensor data can real-time monitor the evacuation of nearby residents, the status of fire-fighting facilities, and changes in meteorological conditions. Integrating these data can build a comprehensive and detailed fire situation perception model, providing all-round information support for emergency command, and greatly improving the comprehensiveness and accuracy of public safety event perception. Precise prediction ability of the RF-HMM model: The intelligent analysis and prediction model that combines random forest and hidden Markov model (RF-HMM) gives full play to the advantages of random forest in dealing with complex relationships and feature selection of multi-source data and the accurate modeling ability of the hidden Markov model for the sequential changes of events. Random forest can quickly screen out the features that have an important impact on public safety events and conduct a preliminary classification of the events, providing more accurate input information for the hidden Markov model. The hidden Markov model, based on the sequence and internal logic of event development, predicts the transition probability and development trend of events at different stages. For example, in the prediction of infectious disease epidemics, it can estimate in advance the peak of the epidemic outbreak, the duration, and the epidemic trend under different prevention and control measures, providing a strong basis for formulating targeted prevention and control strategies, and effectively improving the accuracy and reliability of public safety event prediction. Particle swarm optimization and multi-agent system for optimized decision-making: The model based on particle swarm optimization algorithm and multi-agent system in the emergency decision-making optimization subsystem comprehensively considers the optimized allocation of emergency resources and the collaborative cooperation among emergency rescue entities. The particle swarm optimization algorithm efficiently searches for the optimal resource allocation plan in a huge decision space, avoiding the blindness and inefficiency of traditional decision-making methods. The multi-agent system makes the decision more in line with the actual emergency scenario by simulating the behaviors and interaction relationships of different emergency rescue entities, improving the flexibility and adaptability of the decision. For example, in urban flood rescue, the particle swarm optimization algorithm can determine the best distribution plan of rescue supplies, the deployment locations and action routes of rescue teams, etc. The multi-agent system further coordinates the work among government departments, rescue teams, medical units, and volunteer organizations to ensure that all parties cooperate closely and efficiently during the emergency rescue process, thereby minimizing disaster losses and improving the overall effect of emergency rescue. Efficient command and dispatch and collaborative operation platform: The command and dispatch and collaboration subsystem provides a comprehensive platform for emergency command personnel that integrates information display, decision support, resource allocation, and communication and collaboration.Through an intuitive visualization interface, commanders can quickly grasp the key information and development trend of public security incidents and make scientific decisions in a timely manner. The resource allocation function of the platform realizes the real-time and precise scheduling of emergency resources, improving the efficiency of resource utilization. At the same time, the information sharing and communication cooperation mechanism among various emergency rescue entities ensures that all parties can closely cooperate and fight together during the emergency rescue process, avoiding the occurrence of information islands and duplicate labor phenomena, and greatly enhancing the overall efficiency and coordination of emergency rescue work. In addition, the emergency drill and simulation function of the platform provides a powerful tool for the formulation and optimization of emergency plans, enabling problems to be discovered in advance and experience to be accumulated, further improving the scientificity and effectiveness of emergency command.

[0062] At the data perception level, traditional systems mainly rely on limited ground monitoring equipment and manual inspections, with a narrow and untimely data acquisition range. The system of the present invention can achieve real-time and full-scale monitoring of large areas through multi-source data collection from space, air, and ground. For example, in dealing with forest fires, the traditional method may only detect the fire after it has spread over a large area. This system can, at the initial stage of the fire, detect hot spots through the wide-area monitoring of satellites, quickly approach by drones to obtain precise details, and ground sensors simultaneously feedback information on the surrounding environment. The data perception range is several times larger than that of traditional systems, and the information acquisition time is significantly shortened, enabling early warnings several hours or even days in advance.

[0063] In terms of analysis and prediction, traditional emergency command systems lack advanced models to integrate multi-source data and mostly rely on experience to judge trends, with poor accuracy. The RF-HMM model of this system performs excellently. Random forests classify and screen effective features from multi-source data, and the hidden Markov model accurately deduces according to time series. For example, in predicting flood disasters, traditional predictions may have large errors. This system can combine meteorological satellite precipitation data, drone monitoring of river channel topography and water level changes, ground hydrological sensor data, etc., to accurately predict the arrival time of flood peaks, water level rises, etc. The prediction accuracy is increased by about 40%-60% compared with traditional methods, providing a reliable basis for evacuating the masses in advance and allocating rescue resources.

[0064] In the emergency decision-making link, traditional system decisions often lack systematic optimization, and resource allocation is prone to imbalance. The collaborative effect of the particle swarm optimization and multi-agent system of this system is significant. The particle swarm quickly finds an optimal resource allocation plan in the vast decision-making space, and the multi-agent system promotes efficient cooperation among all parties. In the rescue of major urban accidents, traditional decisions may lead to a concentration of rescue resources or a shortage of resources in key areas. This system can increase the rescue efficiency by about 50% and reduce resource waste by about 40%, achieving scientific and reasonable emergency decisions and making rescue operations more orderly and efficient.

[0065] In terms of command and dispatch collaboration, the information transmission in the traditional mode is not smooth and the collaboration efficiency is low. The command, dispatch and collaboration subsystem of this system provides a unified and efficient platform. The visual interface helps commanders make accurate decisions, and the information sharing and collaboration among all parties are smooth. For example, in the prevention and control of large-scale public health incidents, it is difficult for medical units, communities, transportation departments, etc. to collaborate under the traditional system. This system enables all parties to cooperate closely, shortening the task execution time by about 30%, greatly improving the overall collaboration and timeliness of emergency response, and ensuring the efficient development of public safety emergency work.

Claims

1. An air-ground integrated public safety emergency command decision support system, characterized in that: include: The multi-source data collection layer is used to collect public safety-related data from the air, ground and space. Aerial data collection includes data collection by drone-mounted monitoring equipment and satellite remote sensing data reception. Ground data collection includes surveillance camera networks, sensor arrays and smart terminal data collection. Space data collection relies on high-resolution observation satellites. The data preprocessing and fusion layer cleans, denoises, converts formats, aligns data, and extracts features from the collected multi-source heterogeneous data. It also uses a fusion algorithm based on the hypergraph model and deep neural network fusion (HG-DNN) to construct feature vectors of different modalities into a hypergraph structure and perform deep fusion to obtain a fused feature representation. The intelligent analysis and prediction layer, based on the quantum-inspired probabilistic graphical model (QIPGM), regards public safety-related factors as quantum states to construct a quantum probability graph and determine the quantum state transfer probability matrix. The quantum probability graph is measured through the quantum measurement principle to predict the probability of public safety incidents, development trends and possible results. The emergency decision optimization layer formulates emergency decision strategies based on intelligent analysis and prediction results, combined with an emergency decision optimization model based on a multi-objective evolutionary algorithm and dynamic game theory. The multi-objective evolutionary algorithm is used to search for multi-objective decision solutions, and the dynamic game theory is used to analyze the conflicts of interest and dynamic interactions between emergency rescue entities. The command and dispatch and collaboration layer provides a command and dispatch platform for emergency commanders, realizes unified dispatch of emergency resources and collaborative work management among emergency rescue entities, and supports manual adjustment and optimization of command and dispatch commands as well as information sharing and communication collaboration according to actual conditions.

2. According to claim 1, the air-ground integrated public safety emergency command decision support system is characterized in that: The drone-mounted equipment in the multi-source data acquisition layer includes one or more of a high-definition camera, an infrared thermal imager, and a gas sensor, which are used to obtain images, videos, and environmental parameter data.

3. The air-ground integrated public safety emergency command decision support system according to claim 1 is characterized in that: The preprocessing of satellite remote sensing image data in the data preprocessing and fusion layer includes image enhancement, geometric correction, band fusion processing and extraction of ground feature and disaster feature image feature vectors.

4. The air-ground integrated public safety emergency command decision support system according to claim 1 is characterized in that: The quantum state transfer probability matrix in the quantum-inspired probabilistic graphical model (QIPGM) is dynamically updated and optimized based on historical data, expert knowledge and real-time data.

5. An air-ground integrated public safety emergency command decision support method, characterized in that: The following steps are involved: Multi-source data collection step, using the multi-source data collection layer to extensively collect public safety-related data from the air, ground and space; Data preprocessing and fusion steps: preprocess the collected raw data and use the HG-DNN fusion algorithm to perform deep fusion of multi-source data to obtain fusion feature representation; Based on the intelligent analysis and prediction steps of the Quantum Inspired Probabilistic Graphical Model (QIPGM), a QIPGM model is constructed to predict the probability, development trend and possible results of public safety incidents; Emergency decision-making optimization step: formulate emergency decision-making strategies based on intelligent analysis and prediction results using emergency decision-making optimization models based on multi-objective evolutionary algorithms and dynamic game theory; Command, dispatch and coordination steps. Through the command, dispatch and coordination layer, emergency command personnel issue command and dispatch orders based on information, realize the coordinated work of emergency resource allocation and emergency rescue entities, and can adjust and optimize decisions in real time.

6. The air-ground integrated public safety emergency command decision support method according to claim 5 is characterized in that: In the multi-source data collection step, aerial drone collection deploys different types of drones according to the characteristics and needs of the monitoring area and obtains corresponding data, while receiving different types of satellite data and establishing a ground data collection network to collect monitoring camera, sensor and smart terminal data.

7. The air-ground integrated public safety emergency command decision support method according to claim 5 is characterized in that: When constructing a hypergraph structure in the data preprocessing and fusion step, different modal feature vectors are used as nodes, and hyperedges are constructed to connect multiple related feature vectors according to the logical relationship of the data.

8. The air-ground integrated public safety emergency command decision support method according to claim 5 is characterized in that: In the intelligent analysis and prediction step based on the quantum heuristic probabilistic graphical model (QIPGM), the quantum state includes relevant factors such as the type of public safety event, the location of occurrence, environmental conditions, and social influencing factors.

9. The air-ground integrated public safety emergency command decision support method according to claim 5, characterized in that: The decision objectives of the multi-objective evolutionary algorithm in the emergency decision optimization step include minimizing casualties, minimizing property losses, maximizing emergency response efficiency, and minimizing resource consumption.

10. The air-ground integrated public safety emergency command decision support method according to claim 5, characterized in that: In the command, dispatch and coordination steps, the command and dispatch platform supports location sharing, task allocation and collaborative operations as well as information communication and resource allocation coordination among different emergency rescue entities.

Images