Satellite-borne ai disaster emergency response method and system

Abstract

The application provides a satellite-borne AI disaster emergency response method and system, which is applied to a satellite-borne AI disaster emergency response module carried on a satellite. Disaster monitoring data from at least two sources of ground sensors, user terminals or satellite-borne remote sensing devices is received, a lightweight disaster emergency AI model is called to analyze and fuse the disaster monitoring data, the type of disaster and the risk level are identified, and an emergency guidance text is generated. The text is encoded into a response short message and transmitted to the user terminals in the coverage area through the downlink. The whole process of multi-source fusion, disaster analysis and text generation is placed in the satellite end, solving the problem that multi-source data cannot be fused on the satellite. Relying on the on-orbit processing of the satellite-borne AI, the response period is shortened to the satellite transit period. At the same time, the downlink is directly connected and does not depend on ground communication facilities, so that the instruction can still be quickly reached in the case of communication interruption or network congestion, significantly improving the disaster response timeliness and risk avoidance reliability in the area without ground network coverage.

Description

Technical Field

[0001] This invention relates to the fields of satellite IoT communication and disaster emergency management technology, and in particular to a spaceborne AI disaster emergency response method and system. Background Technology

[0002] With the increasing frequency of extreme weather and geological disasters, the demand for disaster monitoring and early warning in remote mountainous areas, near-shore waters, and areas without terrestrial communication network coverage is growing. Currently, in remote, near-shore waters, and areas without terrestrial communication network coverage, existing satellite-based disaster emergency response methods typically employ a standard process of data collection, transparent satellite relay, centralized ground processing, and command transmission and distribution. For example, disaster monitoring data collected by ground sensors or user terminals is transmitted to a ground gateway station via satellite relay, then transmitted back to the data center via the core network. The ground server then performs centralized analysis and comprehensive assessment to generate early warning or guidance information, which is then distributed to end users via satellite link.

[0003] However, existing response methods have several shortcomings in real-world disaster scenarios. First, heterogeneous data from multiple sources, such as sensor readings received on-board, terminal-reported text, and remote sensing images, cannot be fused and analyzed on-board and must be completely transmitted back to the ground system for processing, resulting in a lengthy data processing chain. Second, the generation of emergency guidance texts relies entirely on the computing power of ground servers and centralized computing processes. The entire process, from data upload, cross-network transmission, ground parsing to instruction generation, is time-consuming, making rapid response difficult. Finally, due to the long response chain and dispersed processing nodes, emergency information received by end users is often significantly delayed, failing to meet the timeliness requirements of disaster emergency response.

[0004] It is evident that existing disaster emergency response methods suffer from technical problems such as the inability to integrate and process multi-source monitoring data on satellite in real time, the heavy reliance on ground computing links for emergency guidance text generation leading to high response latency, and the difficulty in generating and distributing emergency guidance text in a timely manner. Summary of the Invention

[0005] This invention provides a spaceborne AI disaster emergency response method and system, aiming to overcome the technical problems of existing disaster emergency response methods, such as the inability to integrate and process multi-source monitoring data on the satellite in real time, the heavy reliance on ground computing links for emergency guidance text generation leading to high response latency, and the difficulty in generating and distributing emergency guidance text in a timely manner.

[0006] This invention provides a spaceborne AI disaster emergency response method, applied to a spaceborne AI disaster emergency response module mounted on a satellite, the method comprising: Receive disaster monitoring data from at least two sources: ground sensors, user terminals, or spaceborne remote sensing equipment; The lightweight disaster emergency AI model deployed in the emergency response module is invoked to perform fusion analysis on the disaster monitoring data, identify the disaster type and assess the risk level, and generate disaster emergency guidance text as a guide for evacuation actions for the disaster type. The disaster emergency guidance text is encoded into a response short message and sent to user terminals within the coverage area via the downlink, so that the user terminals receive and output the emergency guidance text.

[0007] According to a spaceborne AI disaster emergency response method provided by the present invention, the step of fusing and analyzing the disaster monitoring data includes: The received disaster monitoring data is spatiotemporally aligned and features are extracted to construct a unified multi-source disaster feature vector; The multi-source disaster feature vector is input into the lightweight disaster emergency AI model for analysis, and the disaster type and risk level are output.

[0008] According to a spaceborne AI disaster emergency response method provided by the present invention, the lightweight disaster emergency AI model is invoked to generate evacuation action guidelines for the disaster type as the disaster emergency guidance text, including: Invoke the onboard geographic information system and evacuation road network; Based on the disaster type, the risk level, and the disaster location information extracted from the disaster monitoring data, and in conjunction with the geographic information system and the evacuation road network, the optimal evacuation route and the location of the target shelter are planned. The disaster type, risk level, optimal evacuation route, and target shelter location are combined to generate the evacuation action guide, which is then output as the disaster emergency guidance text.

[0009] The spaceborne AI disaster emergency response method provided by the present invention further includes: The system invokes the onboard historical disaster case database and matches historical risk avoidance strategies corresponding to the disaster type and risk level from the database. The historical risk avoidance strategies are analyzed to extract safety precautions, which are then added to the disaster emergency guidance text to generate an updated disaster emergency guidance text.

[0010] The spaceborne AI disaster emergency response method provided by the present invention further includes: Based on the urgency level identifier carried in the disaster monitoring data, or based on the pre-set regional priority on the satellite, a processing priority label is generated; When on-board computing resources are limited, the lightweight disaster emergency AI model is prioritized to perform fusion analysis on high-priority disaster monitoring data and generate emergency guidance text, according to the processing priority tags.

[0011] According to a spaceborne AI disaster emergency response method provided by the present invention, if there are multiple response short messages to be sent within the coverage area, the step of encoding the disaster emergency guidance text into response short messages and sending them to all user terminals within the coverage area via downlink includes: Generate a corresponding distribution priority label based on the risk level corresponding to each response short message, and bind the distribution priority label to the corresponding response short message; If the on-board communication bandwidth is limited or the risk levels are all below the preset broadcast threshold, the multiple response short messages are queued according to the distribution priority label, and the corresponding response short messages are sent to the user terminals in high-risk areas via the downlink. If any of the aforementioned risk levels reaches or exceeds the preset broadcast threshold, the system switches to regional broadcast mode, and the corresponding response short message is simultaneously sent to all user terminals within the coverage area via the downlink broadcast channel.

[0012] The spaceborne AI disaster emergency response method provided by the present invention further includes: The disaster monitoring data, the disaster type, the risk level, and the disaster emergency guidance text are packaged into a disaster response data package; The disaster response data packet is stored in the satellite's onboard non-volatile memory; When the satellite flies over the ground station's communication coverage area, it will transmit the disaster response data packets in batches to the ground disaster assessment system for disaster assessment and rescue dispatch.

[0013] The present invention also provides a spaceborne AI disaster emergency response device, applied to a spaceborne AI disaster emergency response module mounted on a satellite, the device comprising: The data receiving unit is used to receive disaster monitoring data from at least two sources, including ground sensors, user terminals, or spaceborne remote sensing equipment. The data processing unit is used to call the lightweight disaster emergency AI model deployed in the emergency response module to perform fusion analysis on the disaster monitoring data, identify the disaster type and assess the risk level, and generate disaster emergency guidance text as a guide for evacuation actions for the disaster type. The instruction issuing unit is used to encode the disaster emergency guidance text into a response short message and send it to all user terminals within the coverage area via the downlink, so that the user terminals receive and output the emergency guidance text.

[0014] The present invention also provides an Internet of Things (IoT) satellite, comprising: an onboard memory, an onboard processor, and a computer program stored on the onboard memory and running on the onboard processor, wherein the onboard processor, when executing the computer program, implements any of the above-described onboard AI disaster emergency response methods.

[0015] The present invention also provides a spaceborne AI disaster emergency response system, comprising: a user terminal and the aforementioned Internet of Things satellite; The user terminal is connected to the IoT satellite via a store-and-forward communication link and is used to receive disaster emergency guidance texts issued by the IoT satellite.

[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the spaceborne AI disaster emergency response method as described above.

[0017] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the spaceborne AI disaster emergency response method as described above.

[0018] This invention provides a spaceborne AI disaster emergency response method and system. The method is applied to a spaceborne AI disaster emergency response module mounted on a satellite. First, it receives disaster monitoring data from at least two sources: ground sensors, user terminals, or spaceborne remote sensing equipment. Then, it invokes a lightweight disaster emergency AI model deployed in the emergency response module to fuse and analyze the disaster monitoring data, identify the disaster type and assess the risk level, and generate disaster-specific evacuation action guidelines as disaster emergency guidance text. Finally, the disaster emergency guidance text is encoded into a short response message and sent to user terminals within the coverage area via a downlink, enabling user terminals to receive and output the emergency guidance text. This invention executes the entire process of multi-source data fusion, disaster assessment, and text generation on the satellite, eliminating the need for centralized processing links through ground gateways, core networks, and data centers. This overcomes the technical limitation of multi-source data fusion and analysis on-board. By relying on the spaceborne AI model to complete identification, assessment, and guidance generation in orbit, it avoids the time bottlenecks of cross-network data transmission and ground computation, shortening the emergency response cycle from hours to within the satellite's transit cycle, thereby overcoming the technical problem of high response latency. Meanwhile, the downlink direct connection mechanism eliminates the need for terrestrial communication facilities, ensuring rapid delivery of instructions even during communication interruptions or network congestion. This addresses the shortcomings of timely emergency guidance text delivery and significantly improves the timeliness of response and the reliability of disaster avoidance in areas without terrestrial network coverage during the initial stages of a disaster. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is one of the flowcharts of the spaceborne AI disaster emergency response method provided by the present invention.

[0021] Figure 2 This is the second flowchart illustrating the spaceborne AI disaster emergency response method provided by the present invention.

[0022] Figure 3 This is a time-series comparison diagram of the spaceborne AI disaster emergency response method provided by the present invention.

[0023] Figure 4 This is one of the structural schematic diagrams of the spaceborne AI disaster emergency response device provided by the present invention.

[0024] Figure 5 This is the second structural schematic diagram of the spaceborne AI disaster emergency response device provided by the present invention.

[0025] Figure 6 This is a schematic diagram of the structure of the Internet of Things satellite provided by the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0027] Existing satellite-based disaster emergency response methods typically employ a standard process: data acquisition, transparent satellite relay, centralized ground processing, and command feedback and dissemination. For example, disaster monitoring data collected by ground sensors or user terminals is relayed to a ground gateway station via satellite, then transmitted back to the data center via the core network. The ground server performs centralized analysis and comprehensive assessment to generate early warning or guidance information, which is then disseminated to end users via satellite link. However, this response method has several shortcomings in real-world disaster scenarios. For instance, multi-source heterogeneous data, such as sensor values ​​received on-board, terminal-reported text, or remote sensing images, cannot be fused and analyzed on-board and must be completely transmitted back to the ground system for processing, resulting in a lengthy data processing chain. Furthermore, the generation of emergency guidance text relies entirely on the computing power and centralized computing processes of the ground server. The entire process, from data upload, cross-network transmission, ground analysis to command generation, is time-consuming, making rapid response difficult. Moreover, due to the long response chain and dispersed processing nodes, emergency information received by end users often experiences significant delays, failing to meet the timeliness requirements of disaster emergency response.

[0028] To address the aforementioned problems in existing technologies, this invention provides a spaceborne AI-based disaster emergency response method and system. The inventive concept lies in deploying a lightweight disaster emergency AI model on a satellite to directly perform on-orbit fusion analysis of disaster monitoring data from various sources, such as ground sensors, user terminals, or spaceborne remote sensing equipment. This identifies disaster types and assesses risk levels, instantly generating targeted evacuation action guidelines. The process of multi-source data fusion analysis, disaster type identification, risk level assessment, and emergency guidance text generation is pre-executed on the satellite and ultimately transmitted directly to user terminals within the coverage area via downlink. This completely changes the traditional model of transparent satellite relay and centralized ground computing, eliminating the need for ground gateway stations, core networks, and data centers, forming a closed-loop disaster emergency response link on the satellite. This significantly reduces response time and improves emergency guidance capabilities in areas without terrestrial network coverage during the initial stages of a disaster.

[0029] The spaceborne AI disaster emergency response method and system provided by the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] Figure 1 This is one of the flowcharts illustrating the spaceborne AI disaster emergency response method provided by the present invention. This method is applied to a spaceborne AI disaster emergency response module mounted on a satellite. Figure 1 As shown, the method includes: S101. Receive disaster monitoring data from at least two sources, including ground sensors, user terminals, or spaceborne remote sensing equipment.

[0031] In this embodiment of the invention, a satellite refers to an on-orbit spacecraft with short message communication capabilities between satellite and ground, as well as on-board data storage and processing capabilities. For example, it could be an IoT satellite supporting store-and-forward mode, a low-Earth orbit communication satellite with on-board processing capabilities, a navigation satellite supporting short message functionality, or a geostationary orbit communication satellite equipped with an onboard processor. All of the above satellite platforms can receive data from ground sensors, user terminals, or onboard remote sensing equipment, and complete storage, processing, and distribution onboard. The onboard AI disaster emergency response module is a dedicated functional payload independently mounted on this satellite platform. Its physical architecture may include, for example, an onboard processor, onboard memory, and a short message communication interface. A lightweight disaster emergency AI model is deployed on it, and it is pre-installed with a geographic information system, an evacuation road network, and a historical disaster case database.

[0032] When a satellite passes over a disaster-prone area, it receives disaster monitoring data from at least two different sources. These sources may include numerical data such as seismic waves, water levels, smoke concentrations, and wind speeds collected by ground sensors deployed in disaster-prone areas, which can be sent to the satellite in short message format; or descriptions of the emergency reported by user terminals, such as transceiver modules integrated into emergency equipment, in short message format, such as strong tremors, shaking houses, water levels rising to the warning line, or the discovery of forest fire smoke points; or optical or infrared images and spectral data acquired by spaceborne remote sensing equipment to help identify the scope and intensity of the disaster.

[0033] It receives data from at least two sources for multi-source data fusion analysis, comprehensively utilizes the complementary advantages of different types of data, improves the accuracy of disaster type identification and risk level assessment, and enables disaster monitoring data to converge on the satellite without passing through ground gateway stations or data centers, providing a prerequisite for forming an on-board closed-loop emergency response link.

[0034] S102. Call the lightweight disaster emergency AI model deployed in the emergency response module to perform fusion analysis on disaster monitoring data, identify disaster types and assess risk levels, and generate disaster avoidance action guidelines as disaster emergency guidance texts.

[0035] The lightweight disaster emergency AI model deployed in the onboard AI disaster emergency response module is invoked to perform fusion analysis on disaster monitoring data from at least two sources. This fusion analysis integrates the advantages of different data types, such as spatiotemporal alignment and joint analysis of data reported by ground sensors (e.g., water levels, seismic waves), text descriptions uploaded by user terminals (e.g., strong tremors or visible fire points), and images and spectral data collected by onboard remote sensing equipment. This allows for accurate identification of the current disaster type, such as earthquakes, floods, wildfires, and typhoons, and assessment of its risk level (e.g., general, moderate, severe, or extremely severe). Furthermore, based on the identified disaster type and risk level, the lightweight disaster emergency AI model generates evacuation action guidelines for that disaster type, which serve as the final disaster emergency guidance text. This text may include specific details such as evacuation directions, shelter locations, and safety precautions. The entire process of fusion analysis of disaster monitoring data, identification of disaster types, assessment of risk levels, and generation of evacuation action guidelines for disaster types as disaster emergency guidance text is completed independently onboard, without relying on ground stations or data centers, ensuring the timeliness of disaster response.

[0036] It should be noted that the lightweight disaster emergency response AI model is a multimodal disaster analysis model trained based on historical disaster data, geographic information, and evacuation route planning models. It can integrate numerical data, text descriptions, and image / spectral data. Furthermore, the model undergoes lightweight processing through compression techniques such as pruning, quantization, and knowledge distillation to adapt to the computing power and storage constraints of satellite payloads. It supports embedded real-time inference and possesses functions such as disaster type identification, risk level assessment, evacuation route planning, and safety command generation. Optionally, the lightweight disaster emergency response AI model can employ, for example, a lightweight convolutional neural network or a lightweight Transformer architecture, but this invention is not limited to these.

[0037] In some embodiments, the fusion analysis of disaster monitoring data in step S102 may include: The received disaster monitoring data is spatiotemporally aligned and features are extracted to construct a unified multi-source disaster feature vector. The multi-source disaster feature vector is then input into a lightweight disaster emergency AI model for analysis, outputting disaster type and risk level to identify the disaster type and assess the risk level.

[0038] Specifically, when receiving at least two types of data, such as water level data reported by ground sensors (e.g., water levels surging to warning levels), text descriptions reported by user terminals (e.g., torrential rain in mountainous areas, rapid rise in river levels), or satellite-borne optical remote sensing images showing an expansion of water area in the region), the satellite-borne AI emergency response module first performs spatiotemporal alignment on the aforementioned multi-source data. For example, it unifies the timestamps of sensor values, the sending location of user text, and the pixel coordinates of remote sensing images to the same spatiotemporal reference. Then, it performs feature extraction, such as extracting numerical features from water level data, extracting keyword features from text descriptions through natural language processing, and extracting water area change features from remote sensing images. These features are then concatenated into a unified multi-source disaster feature vector, which is input into the lightweight disaster emergency AI model. The lightweight disaster emergency AI model performs forward inference, for example, outputting a disaster type of flash flood and a risk level of severe.

[0039] It is evident that by aligning and extracting features, a unified multi-source disaster feature vector can be constructed. This can integrate heterogeneous and scattered ground sensor values, user text descriptions, or spaceborne remote sensing image data into a structured model input on satellite, thereby significantly improving the accuracy of disaster type identification and risk level assessment. It can also avoid misjudgments or omissions caused by incomplete information from a single data source, providing a reliable basis for the generation of subsequent evacuation guidelines.

[0040] In some embodiments, step S102 involves calling a lightweight disaster emergency AI model to generate evacuation action guidelines for specific disaster types as disaster emergency guidance text. Possible implementation methods include... Figure 2 As shown, Figure 2 This is the second flowchart illustrating the spaceborne AI disaster emergency response method provided by the present invention, as shown below. Figure 2 As shown, the method includes: S201, invoke the onboard geographic information system and evacuation road network.

[0041] For example, the onboard AI disaster emergency response module calls upon the pre-installed Geographic Information System (GIS) and evacuation road network in the satellite's non-volatile memory. The GIS contains data such as digital elevation models, settlement distribution, water systems, and terrain slopes, while the evacuation road network records the various levels of roads, bridges, emergency access routes, and their accessibility within the area. It's worth noting that the GIS and evacuation road network data are loaded into the onboard memory before satellite launch, eliminating the need to retrieve them from the ground during a disaster. This provides fundamental geographic information support for autonomous evacuation route planning onboard.

[0042] S202. Based on the disaster type, risk level, and disaster location information extracted from disaster monitoring data, and combined with geographic information systems and evacuation road networks, plan the optimal evacuation routes and target shelter locations.

[0043] Based on the disaster types identified in the aforementioned steps, such as earthquakes, floods, and wildfires, and the assessed risk levels, such as severe, as well as the disaster location information extracted from disaster monitoring data, such as epicenter coordinates, flood inundation boundaries, and fire point latitude and longitude, combined with the invoked geographic information system and evacuation road network, the optimal evacuation routes and target shelter locations are planned.

[0044] For example, in the case of flood disasters, a lightweight disaster emergency AI model combines digital elevation models to analyze water flow direction and low-lying areas, avoiding roads that are already flooded or may be flooded, and using road network data to retrieve the shortest and safest routes to higher ground or shelters. In the case of earthquake disasters, it avoids bridges, areas with dense high-rise buildings, and sections of road prone to collapse, prioritizing shelters in open areas. The planning results can include optimal evacuation routes and target shelter locations. The optimal evacuation route could be a specific route description, such as traveling northeast along Heping Road to the stadium, and the target shelter location could be the coordinates or name of the shelter.

[0045] S203. Combine the disaster type, risk level, optimal evacuation route and target shelter location to generate evacuation action guidelines, and output the evacuation action guidelines as a disaster emergency guidance text.

[0046] Furthermore, the output disaster type and risk level are combined with the planned optimal evacuation route and target shelter location to generate a structured evacuation action guide, which is then output as a disaster emergency guidance text. For example, the output text could be: Earthquake warning, magnitude estimated 5.5, severe risk level. Please immediately evacuate along Heping Road towards the stadium, avoiding buildings. Target shelter: Stadium (approximately 300 meters northeast).

[0047] As described in the above embodiments, by combining pre-installed onboard geographic information and road networks with real-time disaster location information, the system can autonomously plan optimal evacuation routes and shelter locations without relying on ground communication. The generated emergency guidance text contains specific and actionable route instructions, addressing the pain points of traditional solutions that only send simple warnings and leave users unsure of where to take shelter. This significantly improves the scientific rigor and success rate of evacuation operations. Furthermore, the planning process is completed onboard, eliminating the need for data transmission back to the ground, thus compressing the response time to within the satellite's transit cycle and providing valuable evacuation windows for affected individuals.

[0048] In some embodiments, after planning the optimal evacuation route and target shelter location and generating emergency guidance text, in order to further enhance the comprehensiveness and practicality of the guidance, it also includes supplementing safety precautions by incorporating a historical disaster case database.

[0049] For example, the system first calls the onboard historical disaster case database, matches historical risk avoidance strategies corresponding to disaster types and risk levels from the database, then analyzes the historical risk avoidance strategies to extract safety precautions, and then adds the safety precautions to the disaster emergency guidance text to generate an updated disaster emergency guidance text.

[0050] The historical disaster case database stores historical avoidance strategies for different disaster types and risk levels, including summaries of successful experiences and lessons learned. For example, for earthquake disasters, the database records safety precautions such as avoiding glass curtain walls, staying away from dangerous walls, protecting the head, and not using elevators; for flood disasters, it records experiences such as not wading through water, being wary of manhole covers being washed away, staying away from utility poles, and moving to higher ground. The lightweight disaster emergency AI model uses the pre-installed historical disaster case database to match historical avoidance strategies corresponding to the current disaster type and risk level. For example, for an identified earthquake disaster with a severe risk level, it matches earthquake-related avoidance strategy entries. Further analysis of the historical avoidance strategies extracts key safety precautions, such as staying away from buildings, glass curtain walls, and dangerous walls, protecting the head, and being aware of aftershocks. Finally, these safety precautions are added to the generated disaster emergency guidance text, resulting in an updated and more comprehensive disaster emergency guidance text.

[0051] For example, the original disaster emergency guidance text was: Earthquake Warning: Magnitude estimated at 5.5, risk level severe. Please evacuate immediately along Heping Road towards the stadium. After adding safety precautions, it was updated to: Earthquake Warning: Magnitude estimated at 5.5, risk level severe. Please evacuate immediately along Heping Road towards the stadium, avoid glass curtain walls, stay away from dangerous walls, protect your head, and be alert for aftershocks.

[0052] By accessing a historical disaster case database and matching it with safety precautions, valuable disaster avoidance experience accumulated in past disasters can be added to the current emergency guidance text in real time. This makes the final disaster emergency guidance text more comprehensive and closer to the actual scenario, helping users avoid typical secondary risks and further enhancing the practical value and risk avoidance effect of the emergency guidance.

[0053] S103. Encode the disaster emergency guidance text into a response short message and send it to the user terminal in the coverage area through the downlink, so that the user terminal can receive and output the emergency guidance text.

[0054] The generated disaster emergency guidance text can be encoded according to a preset short message protocol and encapsulated into one or more response short messages for distribution.

[0055] In some embodiments, the response short message format needs to be adapted to the satellite-to-ground link bandwidth and the receiving capability of the user terminal. For example, a compact binary encoding or compressed text format can be used to ensure reliable transmission within a limited transit communication window.

[0056] After encoding, the response short message can be sent to user terminals within the satellite coverage area via downlink, such as IoT communication bands. Upon receiving the response short message, the user terminal automatically decodes and extracts the emergency guidance text, which is then output to the user through the terminal display or voice broadcast, enabling the user to obtain emergency guidance information including disaster type, risk level, optimal evacuation route, location of target shelter, and safety precautions.

[0057] In this step, the response short message is sent directly to the user terminal via the downlink, without passing through ground gateway stations, core networks, or any ground communication facilities. Therefore, even if the ground communication network in the disaster area is completely interrupted, such as when an earthquake causes damage to base stations or severe network congestion, the emergency guidance text can still be delivered quickly and stably. This eliminates the dependence on ground communication facilities and effectively solves the shortcomings of existing methods in the timely delivery of emergency guidance texts during communication interruptions or network congestion. It significantly improves the timeliness of response and the reliability of disaster avoidance in the early stages of disasters in areas without ground network coverage.

[0058] In some embodiments, the method for sending response short messages can be either point-to-point mode, such as sending to a designated terminal, or area broadcast mode, such as sending to all terminals within the coverage area simultaneously.

[0059] When there are multiple short response messages to be sent within the coverage area, such as emergency guidance texts for multiple disaster events such as earthquakes, wildfires, and floods, the spaceborne AI disaster emergency response method provided by this invention also includes a dynamic distribution strategy based on risk level.

[0060] Specifically, the onboard AI disaster emergency response module first generates a corresponding distribution priority label based on the risk level of each response short message. For example, a risk level of particularly severe or severe corresponds to a high priority label, while a risk level of moderate or relatively mild corresponds to a low priority label, and then binds the distribution priority label to the corresponding response short message.

[0061] Subsequently, the current status of the onboard communication link is determined. If the onboard communication bandwidth is limited, such as the downlink being occupied by other tasks or the risk level of all pending response short messages being lower than the preset broadcast threshold (e.g., all risk levels are moderate or mild), then multiple response short messages are queued and sorted according to the distribution priority label. The corresponding response short messages are then sent to user terminals in high-risk areas via the downlink to ensure that users in high-risk and urgent areas can receive emergency instructions first.

[0062] Conversely, if the risk level corresponding to any short response message reaches or exceeds a preset broadcast threshold, such as a severe or extremely severe earthquake disaster risk level, the system automatically switches to regional broadcast mode. The high-risk short response message is simultaneously sent to all user terminals within the coverage area via the downlink broadcast channel, eliminating the need for individual addressing. For example, when a magnitude 6.5 earthquake (severe risk level) occurs in a certain area, the earthquake emergency guidance text is immediately broadcast to all terminals within its beam coverage area, achieving second-level coverage.

[0063] This invention employs a dynamic distribution strategy based on risk levels. When onboard communication bandwidth is limited or the risk level is low, multiple response short messages are queued and prioritized for distribution to users in high-risk areas, improving resource utilization efficiency. When any risk level reaches or exceeds a preset broadcast threshold, the system automatically switches to regional broadcast mode, simultaneously distributing response short messages to all user terminals within the coverage area via a broadcast channel. This solves the inefficiency problem of existing satellites only supporting point-to-point communication and unable to simultaneously broadcast emergency instructions to all users in disaster-stricken areas. Furthermore, the entire distribution process relies on the satellite downlink, eliminating the need for terrestrial communication facilities. Even in scenarios where extreme disasters such as earthquakes and typhoons cause complete paralysis of terrestrial communications, it ensures that emergency guidance texts are delivered quickly and reliably to affected users. This overcomes the technical bottleneck of traditional solutions where instructions cannot be delivered due to damage to ground facilities or network congestion, significantly improving the timeliness, coverage, and reliability of emergency guidance information distribution.

[0064] The spaceborne AI disaster emergency response method provided by this invention is applied to a spaceborne AI disaster emergency response module mounted on a satellite. First, it receives disaster monitoring data from at least two sources: ground sensors, user terminals, or spaceborne remote sensing equipment. Then, it invokes a lightweight disaster emergency AI model deployed in the emergency response module to fuse and analyze the disaster monitoring data, identify the disaster type and assess the risk level, and generate disaster-specific evacuation action guidelines as disaster emergency guidance text. Finally, the disaster emergency guidance text is encoded into a short response message and sent to user terminals within the coverage area via a downlink, enabling user terminals to receive and output the emergency guidance text. This invention executes the entire process of multi-source data fusion, disaster assessment, and text generation on the satellite end, eliminating the need for centralized processing links through ground gateways, core networks, and data centers. This solves the technical deficiency of multi-source data fusion and analysis on-board. By relying on the spaceborne AI model to complete identification, assessment, and guidance generation in orbit, it avoids the time bottlenecks of cross-network data transmission and ground computation, shortening the emergency response cycle from hours to within the satellite's transit cycle, thereby overcoming the technical problem of high response latency. Meanwhile, the downlink direct connection mechanism eliminates the need for terrestrial communication facilities, ensuring rapid delivery of instructions even during communication interruptions or network congestion. This addresses the shortcomings of timely emergency guidance text delivery and significantly improves the timeliness of response and the reliability of disaster avoidance in areas without terrestrial network coverage during the initial stages of a disaster.

[0065] Figure 3 This is a time-series comparison diagram of the spaceborne AI disaster emergency response method provided by the present invention, which compares the time differences between traditional disaster response modes and the spaceborne AI real-time response mode implemented by the spaceborne AI disaster emergency response method provided by the present invention. Figure 3 As shown in the upper part, in traditional disaster response models, after ground sensors or user terminals send disaster monitoring data, the satellite must wait for it to pass over a ground station before transmitting the data to the ground gateway station. The data is then forwarded to the data center via the core network, and analyzed by personnel or ground systems to generate warnings or evacuation instructions, which are then transmitted back to the satellite. The satellite then sends the instructions to the user terminal via a downlink. The entire process takes the sum of the satellite transit period and the ground analysis delay, typically taking tens of minutes to several hours. However, as... Figure 3As shown in the lower part, the spaceborne AI disaster emergency response method provided by this invention enables a real-time spaceborne AI response mode. After the satellite receives disaster monitoring data, the spaceborne AI disaster emergency response module instantly completes multi-source data fusion, disaster type identification, risk level assessment, and emergency guidance text generation on-board. Within the same transit window, the response short message is directly sent to user terminals within the coverage area via downlink, without needing to go through ground stations, data centers, or manual analysis. The response time is shortened to within the satellite transit cycle, typically 10-30 minutes, and can be reduced to minutes for low-Earth orbit constellations. This time-series comparison clearly demonstrates the significant improvement in response speed of this invention compared to existing technologies, as well as the onboard closed-loop emergency response capability that completely eliminates reliance on ground systems.

[0066] In some embodiments, a satellite may receive disaster monitoring data from multiple different regions and of different types within the same transit cycle, while onboard computing resources, such as processor power and lightweight disaster emergency AI model inference queues, are limited. Therefore, the onboard AI disaster emergency response method provided by this invention also includes a priority scheduling mechanism.

[0067] Specifically, the onboard AI disaster emergency response module generates processing priority labels for each disaster monitoring data based on the urgency level indicators carried in the received disaster monitoring data, such as those actively marked as urgency by user terminals, or based on pre-set regional priorities on the satellite, such as nuclear power plants and densely populated areas being preset as high priority. When onboard computing resources are limited, for example, when onboard computing resources exceed a preset threshold, the disaster monitoring data is sorted according to priority labels, and the lightweight disaster emergency AI model is prioritized to perform fusion analysis and emergency guidance text generation on high-priority disaster monitoring data.

[0068] For example, in the event of a sudden earthquake in a mountainous area, multiple user terminals simultaneously transmit disaster monitoring data to the satellite. User terminals located in the epicenter region proactively add an emergency identifier to their short messages, while user terminals in surrounding areas do not. Upon receiving the disaster monitoring data, onboard computing resources are limited. At this point, the onboard AI disaster emergency response module generates high-priority tags for data with emergency identifiers based on the level of urgency in the disaster monitoring data. It then prioritizes the use of a lightweight disaster emergency AI model to perform fusion analysis and generate emergency guidance text for this data, ensuring that users in the epicenter region receive evacuation instructions first.

[0069] For example, a priority list of regions can be pre-configured on the satellite. For instance, a 10-kilometer radius around a nuclear power plant could be designated as high priority, the downstream area of ​​a large reservoir as high priority, and other areas as ordinary priority. During a satellite pass, it might simultaneously receive water level monitoring data from the vicinity of the nuclear power plant and fire reports from ordinary mountainous areas. Since the nuclear power plant area is pre-defined as high priority, a high-priority label is generated for the nuclear power plant data. When on-board computing resources are limited, a lightweight disaster emergency AI model is prioritized to process the nuclear power plant water level data, generate emergency guidance text, and then process the fire report, thereby ensuring the fastest possible response to disasters in critical infrastructure areas.

[0070] By introducing urgency level indicators or regional priorities, disaster monitoring data from high-priority areas can be processed first when on-board computing resources are limited. This ensures that users in high-risk and urgent areas can receive emergency guidance information as soon as possible, avoids serious disaster response delays caused by competition for computing resources, and improves the real-time response capability of onboard AI models in concurrent multi-disaster scenarios.

[0071] In some embodiments, in order to support ground disaster assessment and subsequent rescue scheduling, so that the ground command system can obtain complete disaster event information, the spaceborne AI disaster emergency response method provided by the present invention also includes an on-board data storage and downlink mechanism.

[0072] Specifically, the onboard AI disaster emergency response module can package received disaster monitoring data, identified disaster types and assessed risk levels, and generated disaster emergency guidance text into a complete disaster response data package. To ensure the security and reliability of the data during long-term storage on the satellite, this data package can be encrypted and stored in the satellite's onboard non-volatile memory. When the satellite reaches the ground station's communication coverage area according to its predetermined orbit, it automatically establishes a data transmission link with the ground station and downloads the stored disaster response data package in batches to the ground disaster assessment system. The ground disaster assessment system can then perform disaster assessment and rescue dispatching work based on the received data, including disaster verification, loss assessment, dispatch of rescue forces, and subsequent disaster prevention decisions.

[0073] This invention, by setting up an on-board data storage and downlink mechanism, can ensure that disaster-stricken users in areas without terrestrial network coverage can obtain real-time emergency guidance from the satellite, and can also ensure that the ground command system can obtain complete and accurate disaster data. It realizes the coordination between on-board intelligent response and ground global scheduling, and provides reliable data support for post-disaster relief and disaster assessment. This invention provides real-time on-board response without losing any key data, and can balance the timeliness of emergency response and the integrity of disaster management.

[0074] The foregoing has described the processing flow and data flow of the spaceborne AI disaster emergency response method provided by this invention. To more clearly illustrate the interaction flow of this invention in a real-world scenario, three specific application embodiments are described below. It should be understood that the following embodiments are only used to further illustrate the technical implementation path and preferred application mode of this invention, and do not constitute a limitation on the scope of protection of the claims. Those skilled in the art can make adaptive adjustments to the embodiments according to actual spaceborne computing power constraints, communication systems, or terminal forms. All equivalent substitutions or improvements made based on the core concept of this invention fall within the protection scope of this invention.

[0075] Example 1 illustrates the use of a flash flood warning in a mountainous area as an example: A resident in a mountainous area received a heavy rain warning from the meteorological department via a handheld terminal, but there was no mobile network in the area. Flash flood monitoring sensors detected a rapid rise in water levels and transmitted a short message to the satellite: "Water level has surged, exceeding the warning line by 2 meters, location: 112°E, 38°N." After the satellite passed overhead and received the message, the onboard AI disaster emergency response module invoked a lightweight disaster emergency AI model. Combining this with the area's digital elevation model and the distribution of residential areas, it performed multi-source data fusion analysis, identifying the disaster type as flash flood and the risk level as severe. Subsequently, it invoked the satellite's pre-set evacuation road network to plan the optimal evacuation route and target shelter locations, generating an emergency guidance text such as: "Flash Flood Warning! Please immediately move to higher ground in the northwest direction. Do not walk along the river. Suggested evacuation route: along the path behind the village to the ridge. Estimated evacuation time: 15 minutes. The township emergency command center has been notified." The satellite then transmitted a response short message to user terminals within the coverage area via downlink. Users received the information 20 minutes later and organized villagers to evacuate safely according to the route. There were no casualties.

[0076] Example 2 illustrates the initial firefighting guidance for forest fires: Forest rangers discovered an initial fire during their patrol and transmitted a message via satellite terminal: "Forest fire detected. Location: Miaozishan. Fire area approximately 20 square meters. No water source nearby." Upon satellite reception, the onboard AI disaster emergency response module invoked a lightweight disaster emergency AI model. Combining this with pre-set geographical information such as vegetation type, wind direction, and firebreak distribution, along with the user-reported text description, the system identified the disaster type as a forest fire and the risk level as initial. It then matched corresponding evacuation strategies from a historical disaster case database, extracted safety precautions, and generated emergency guidance text, such as: "Fire level: Initial fire. Immediately organize personnel to dig a firebreak behind the fire front, at least 5 meters wide. Do not fight the fire with the wind at your back. If the fire spreads, retreat southeast to a firebreak. Reported to the forest fire monitoring center." The satellite then transmitted a response short message via downlink. Following the guidance, the forest rangers controlled the fire and, with the assistance of subsequent reinforcements, extinguished it in its initial stage.

[0077] Example 3 illustrates the large-scale evacuation of people after an earthquake: A 6.5-magnitude earthquake struck an area, destroying all ground base stations. Numerous users in the affected area reported being trapped or injured via satellite. Upon receiving the satellite's data, the onboard AI disaster emergency response module invoked a lightweight disaster emergency AI model. This model integrated text descriptions (including location information) from multiple users with pre-set data on building density and road networks, identifying the disaster type as an earthquake and the risk level as severe. Based on the epicenter location and predicted road damage, optimal evacuation routes and target shelter locations were planned for different areas. Safety precautions from a historical disaster case database were then matched to generate evacuation instructions for each area, such as: Residents of Area A, please evacuate to the stadium along XX Road; Residents of Area B, please avoid landslide-affected areas and detour via Yucai Road to the school playground; please prioritize assistance for the elderly and children, and stay away from dangerous walls. Because the risk level reached a preset broadcast threshold, the satellite automatically switched to regional broadcast mode, simultaneously distributing emergency guidance text to all user terminals within the coverage area. Instructions were continuously updated during the satellite's transit, guiding tens of thousands of people to evacuate in an orderly manner and significantly reducing secondary casualties.

[0078] The following describes the spaceborne AI disaster emergency response device provided by the present invention. The spaceborne AI disaster emergency response device described below and the spaceborne AI disaster emergency response method described above can be referred to in correspondence.

[0079] Figure 4 This is one of the structural schematic diagrams of the spaceborne AI disaster emergency response device provided by the present invention, applied to a spaceborne AI disaster emergency response module mounted on a satellite. Figure 4 As shown, the spaceborne AI disaster emergency response device 400 provided by the present invention includes: The data receiving unit 401 is used to receive disaster monitoring data from at least two sources, including ground sensors, user terminals, or spaceborne remote sensing equipment. The data processing unit 402 is used to call the lightweight disaster emergency AI model deployed in the emergency response module to perform fusion analysis on disaster monitoring data, identify disaster types and assess risk levels, and generate disaster emergency guidance text as a guide for disaster avoidance actions. The instruction issuing unit 403 is used to encode the disaster emergency guidance text into a response short message and send it to all user terminals within the coverage area via the downlink, so that the user terminals can receive and output the emergency guidance text.

[0080] In one possible design, the data processing unit 402 is used for: Spatiotemporal alignment and feature extraction are performed on the received disaster monitoring data to construct a unified multi-source disaster feature vector; The feature vectors of multi-source disasters are input into a lightweight disaster emergency AI model for analysis, and the disaster type and risk level are output.

[0081] In one possible design, the data processing unit 402 is also used for: Invoke the onboard geographic information system and evacuation road network; Based on the disaster type, risk level, and disaster location information extracted from disaster monitoring data, and combined with geographic information systems and evacuation road networks, the optimal evacuation routes and target shelter locations are planned. By combining disaster type, risk level, optimal evacuation route, and target shelter location, a disaster evacuation action guide is generated, and this guide is output as a disaster emergency guidance text.

[0082] In one possible design, the data processing unit 402 is also used for: The system calls upon the onboard historical disaster case database and matches historical risk avoidance strategies corresponding to the disaster type and risk level from the database. Historical risk avoidance strategies are analyzed to extract safety precautions, which are then added to the disaster emergency guidance text to generate an updated disaster emergency guidance text.

[0083] exist Figure 4 On this basis, Figure 5 This is a second structural schematic diagram of the spaceborne AI disaster emergency response device provided by the present invention, as shown below. Figure 5 As shown, the spaceborne AI disaster emergency response device 400 provided by the present invention further includes: a priority scheduling unit 404; the priority scheduling unit 404 is used for: Based on the urgency level indicators carried in the disaster monitoring data, or based on the pre-set regional priorities on the satellite, a processing priority label is generated; When on-board computing resources are limited, the lightweight disaster emergency AI model is prioritized to perform fusion analysis and emergency guidance text generation on high-priority disaster monitoring data, according to processing priority labels.

[0084] In one possible design, if there are multiple short response messages to be sent within the coverage area, the instruction issuing unit 403 is used to: Generate corresponding distribution priority tags based on the risk level of each response short message, and bind the distribution priority tags to the corresponding response short messages; If the on-board communication bandwidth is limited or the risk level is lower than the preset broadcast threshold, multiple response short messages are queued and sorted according to the distribution priority label, and the corresponding response short messages are sent to user terminals in high-risk areas via the downlink. If any risk level reaches or exceeds the preset broadcast threshold, the system switches to regional broadcast mode and simultaneously sends the corresponding response short message to all user terminals within the coverage area via the downlink broadcast channel.

[0085] In one possible design, the instruction issuing unit 403 is also used for: Package disaster monitoring data, disaster types, risk levels, and disaster emergency guidance texts into a disaster response data package; Disaster response data packets are stored in the satellite's onboard non-volatile memory; When the satellite flies over the ground station's communication coverage area, it will transmit disaster response data packets in batches to the ground disaster assessment system for disaster assessment and rescue dispatch.

[0086] Figure 6 This is a schematic diagram of the structure of the IoT satellite provided by the present invention, such as... Figure 6 As shown, the IoT satellite may include: an on-board processor 610, a space-ground communication interface 620, an on-board memory 630, and an internal data bus 640. The on-board processor 610, the space-ground communication interface 620, and the on-board memory 630 communicate and transmit commands to each other via the internal data bus 640. The on-board processor 610 can call logical instructions in the on-board memory 630 to execute an on-board AI disaster emergency response method. This method is applied to the on-board AI disaster emergency response module mounted on the satellite, including: receiving disaster monitoring data from at least two sources, including ground sensors, user terminals, or on-board remote sensing equipment; calling a lightweight disaster emergency AI model deployed in the emergency response module to perform fusion analysis on the disaster monitoring data, identify the disaster type and assess the risk level, and generate disaster emergency guidance text as disaster emergency guidance text; encoding the disaster emergency guidance text into a response short message and sending it to user terminals within the coverage area via downlink, enabling user terminals to receive and output the emergency guidance text.

[0087] Furthermore, the logical instructions in the aforementioned onboard memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a portion 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 to cause an onboard computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks or optical disks, and onboard non-volatile memory.

[0088] The present invention also provides a spaceborne AI disaster emergency response system, including a user terminal and the aforementioned IoT satellite; the user distress terminal is connected to the IoT satellite via a store-and-forward communication link, and is used to send disaster monitoring data to the IoT satellite and receive disaster emergency guidance texts issued by the IoT satellite.

[0089] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the spaceborne AI disaster emergency response method provided by the above methods. This method is applied to a spaceborne AI disaster emergency response module mounted on a satellite, including: receiving disaster monitoring data from at least two sources, such as ground sensors, user terminals, or spaceborne remote sensing equipment; calling a lightweight disaster emergency AI model deployed in the emergency response module to perform fusion analysis on the disaster monitoring data, identify the disaster type and assess the risk level, and generate disaster emergency guidance text as disaster emergency guidance text; encoding the disaster emergency guidance text into a response short message and sending it to the user terminal in the coverage area via the downlink, so that the user terminal receives and outputs the emergency guidance text.

[0090] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program is implemented to perform the spaceborne AI disaster emergency response method provided by the above methods. This method is applied to a spaceborne AI disaster emergency response module mounted on a satellite, and includes: receiving disaster monitoring data from at least two sources, including ground sensors, user terminals, or spaceborne remote sensing equipment; invoking a lightweight disaster emergency AI model deployed in the emergency response module to perform fusion analysis on the disaster monitoring data, identify the disaster type and assess the risk level, and generate disaster emergency guidance text as disaster emergency guidance text; encoding the disaster emergency guidance text into a response short message and sending it to user terminals within the coverage area via a downlink, so that user terminals receive and output the emergency guidance text.

[0091] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The 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 the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0092] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A spaceborne AI-based disaster emergency response method, characterized in that, The method, applied to a satellite-borne AI disaster emergency response module, includes: Receive disaster monitoring data from at least two sources: ground sensors, user terminals, or spaceborne remote sensing equipment; The lightweight disaster emergency AI model deployed in the emergency response module is invoked to perform fusion analysis on the disaster monitoring data, identify the disaster type and assess the risk level, and generate disaster emergency guidance text as a guide for evacuation actions for the disaster type. The disaster emergency guidance text is encoded into a response short message and sent to user terminals within the coverage area via the downlink, so that the user terminals receive and output the emergency guidance text.

2. The method according to claim 1, characterized in that, The fusion analysis of the disaster monitoring data includes: The received disaster monitoring data is spatiotemporally aligned and features are extracted to construct a unified multi-source disaster feature vector; The multi-source disaster feature vector is input into the lightweight disaster emergency AI model for analysis, and the disaster type and risk level are output.

3. The method according to claim 1, characterized in that, The lightweight disaster emergency AI model is invoked to generate evacuation action guidelines for the disaster type, which serve as the disaster emergency guidance text, including: Invoke the onboard geographic information system and evacuation road network; Based on the disaster type, the risk level, and the disaster location information extracted from the disaster monitoring data, and in conjunction with the geographic information system and the evacuation road network, the optimal evacuation route and the location of the target shelter are planned; The disaster type, risk level, optimal evacuation route, and target shelter location are combined to generate the evacuation action guide, which is then output as the disaster emergency guidance text.

4. The method according to claim 3, characterized in that, Also includes: The system invokes the onboard historical disaster case database and matches historical risk avoidance strategies corresponding to the disaster type and risk level from the database. The historical risk avoidance strategies are analyzed to extract safety precautions, which are then added to the disaster emergency guidance text to generate an updated disaster emergency guidance text.

5. The method according to claim 1, characterized in that, Also includes: Based on the urgency level identifier carried in the disaster monitoring data, or based on the pre-set regional priority on the satellite, a processing priority label is generated; When on-board computing resources are limited, the lightweight disaster emergency AI model is prioritized to perform fusion analysis on high-priority disaster monitoring data and generate emergency guidance text, according to the processing priority tags.

6. The method according to claim 1, characterized in that, If there are multiple response short messages to be sent within the coverage area, the step of encoding the disaster emergency guidance text into response short messages and sending them to all user terminals within the coverage area via the downlink includes: Generate a corresponding distribution priority label based on the risk level corresponding to each response short message, and bind the distribution priority label to the corresponding response short message; If the on-board communication bandwidth is limited or the risk levels are all below the preset broadcast threshold, the multiple response short messages are queued according to the distribution priority label, and the corresponding response short messages are sent to the user terminals in the high-risk areas via the downlink. If any of the aforementioned risk levels reaches or exceeds the preset broadcast threshold, the system switches to regional broadcast mode, and the corresponding response short message is simultaneously sent to all user terminals within the coverage area via the downlink broadcast channel.

7. The method according to any one of claims 1-6, characterized in that, Also includes: The disaster monitoring data, the disaster type, the risk level, and the disaster emergency guidance text are packaged into a disaster response data package; The disaster response data packet is stored in the satellite's onboard non-volatile memory; When the satellite flies over the communication coverage area of ​​the ground station, it will transmit the disaster response data packets in batches to the ground disaster assessment system for disaster assessment and rescue dispatch.

8. A spaceborne AI disaster emergency response device, characterized in that, The device, used in a satellite-borne AI disaster emergency response module, comprises: The data receiving unit is used to receive disaster monitoring data from at least two sources, including ground sensors, user terminals, or spaceborne remote sensing equipment. The data processing unit is used to call the lightweight disaster emergency AI model deployed in the emergency response module to perform fusion analysis on the disaster monitoring data, identify the disaster type and assess the risk level, and generate disaster emergency guidance text as a guide for evacuation actions for the disaster type. The instruction issuing unit is used to encode the disaster emergency guidance text into a response short message and send it to all user terminals within the coverage area via the downlink, so that the user terminals receive and output the emergency guidance text.

9. An Internet of Things (IoT) satellite, characterized in that, include: The spaceborne memory, the spaceborne processor, and the computer program stored on the spaceborne memory and running on the spaceborne processor, wherein the spaceborne processor, when executing the computer program, implements the spaceborne AI disaster emergency response method as described in any one of claims 1 to 7.

10. A spaceborne AI disaster emergency response system, characterized in that, include: User terminal and the IoT satellite as described in claim 9; The user terminal is connected to the IoT satellite via a store-and-forward communication link and is used to receive disaster emergency guidance texts issued by the IoT satellite.

Images