Universe space environment parameter distribution reconstruction method and system based on cloud edge collaboration

Abstract

The invention discloses a global space environment parameter distribution reconstruction method and system based on cloud edge collaboration, and belongs to the technical field of space environment detection. The method comprises the following steps: completing registration of a satellite at a cloud side, and maintaining a connection state with the cloud side through an application layer heartbeat mechanism; maintaining a collaborative other satellite list, and obtaining and analyzing a face scanning calculation model to obtain an input data demand of the face scanning calculation model; generating at least one data query request based on the input data demand and the cooperative other satellite list, and sending the data query request to a cloud side and / or other satellites, so that the cloud side and / or other satellites return a data query result; and providing a data query result to the face scanning calculation model to obtain a global space environment parameter distribution reconstruction result. According to the method and the device, the timeliness of the face scanning calculation result can be improved, and the accuracy and the stability of the face scanning calculation result are further ensured or even enhanced by utilizing richer and diversified data input.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of space environment detection, and particularly relates to a global space environment parameter distribution reconstruction method and system based on cloud edge cooperation. BACKGROUND

[0002] Real-time and accurate perception of global space environment parameter distribution is of great significance for space science research, satellite on-orbit intelligent behavior decision, space weather warning, etc., and is an important part of the "intelligent satellite brain" system. Generally, obtaining global space environment parameter distribution depends on processing and inferring observation data of local space environment parameters, which is figuratively referred to as "face brushing" calculation in the field, that is, "face brushing" refers to reconstructing global space environment parameter distribution according to sparse and limited local space detection data.

[0003] At present, the "face brushing" calculation technical scheme mainly includes the following two kinds:

[0004] 1. "Face brushing" calculation based on ground system:

[0005] The traditional "face brushing" calculation process is as follows: a satellite collects observation data of space environment parameters during on-orbit operation, and transmits these data to a ground station. After receiving the data, a ground computing center uses ground computing resources to perform "face brushing" calculation to infer the distribution of global space environment parameters. For example, for "face brushing" calculation of medium and high layer atmospheric density, the typical way is to use the returned satellite observation data to process and analyze on the ground. Although this way can obtain valuable calculation results, its main process depends on data transmission from satellite to ground.

[0006] However, in this mode, satellite observation data needs to be transmitted to the ground through the downlink, which takes a long time (may be several hours or even longer, depending on satellite transit, ground station receiving capacity and data transmission bandwidth, etc.). The delay of data transmission leads to significant lag of "face brushing" calculation results relative to observation time, which is difficult to meet the requirements of real-time application scenarios such as immediate support for satellite autonomous decision-making, rapid response to space environment changes, etc., and has the disadvantage of poor timeliness.

[0007] 2. "Face brushing" calculation based on single satellite on-orbit:

[0008] In order to improve the timeliness of "face brushing" calculation results, the latest research trend is to transfer the "face brushing" calculation task to the satellite and use on-board computing resources for on-orbit real-time calculation. In this way, an intelligent satellite (or a satellite with "intelligent satellite brain") directly completes "face brushing" calculation on the satellite according to the data observed by itself. This on-board calculation mode significantly reduces the time required for data transmission to the ground, thereby improving the timeliness of the calculation results. The general process is as followsFigure 1 In this mode, to ensure the stability and accuracy of the "face brushing" calculation result, the satellite usually needs to accumulate local observation data for a period of time (for example, 60 minutes or more) as the calculation input. Therefore, the calculation mainly relies on the observation data accumulated by the single satellite itself within a period of time.

[0009] However, the characteristics of the "face brushing" calculation algorithm in this mode require the input data to have a certain spatio-temporal coverage to ensure the accuracy and stability of the calculation result. For a single satellite, in order to obtain a sufficient amount of local observation data, the satellite usually needs to be in orbit and continuously observe for a period of time (such as 60 minutes or more). This data accumulation time constitutes the main delay source of the on-board "face brushing" calculation. Therefore, although the timeliness has been improved compared to the ground calculation, for intelligent behavior support that requires higher real-time performance, this delay is still a constraint factor, and therefore its timeliness still has room for improvement.

[0010] In addition, the on-orbit calculation of the single satellite in this mode mainly relies on the observation data of the satellite itself. Although a certain amount of data can be obtained through long-term accumulation, in a short period of time, the spatial coverage range and diversity of the data are limited. This affects the rapid response capability and global optimality of the "face brushing" calculation result in a complex or dynamically changing spatial environment SUMMARY

[0011] To solve the problems of low data input timeliness and single data source faced by on-board "face brushing" calculation (global spatial environment parameter distribution calculation), the present application provides a global spatial environment parameter distribution reconstruction method and system based on cloud-edge collaboration, which not only improves the timeliness of the global spatial environment parameter distribution calculation ("face brushing" calculation) result, but also uses more rich and diverse data input to further ensure and even enhance the accuracy and stability of the "face brushing" calculation result.

[0012] To achieve the above-mentioned purpose, the technical scheme of the present application includes the following contents.

[0013] A global spatial environment parameter distribution reconstruction method based on cloud-edge collaboration applied to a satellite, the method comprising:

[0014] Registering the satellite on the cloud side and maintaining the connection state with the cloud side through an application layer heartbeat mechanism;

[0015] Maintaining a collaborative satellite list, the collaborative satellite list including: other satellite UIDs, inter-satellite link addressing information, key data types provided by other satellites, and inter-satellite reachable states obtained through inter-satellite link detection or negotiation;

[0016] Obtaining and analyzing a face brushing calculation model to obtain the input data requirements of the face brushing calculation model;

[0017] generating at least one data query request based on the input data requirement and the list of cooperative satellites, and sending the data query request to the cloud side and / or other satellites, so that the cloud side and / or other satellites return data query results; wherein the data query request is generated based on a large language model or a structured format;

[0018] providing the data query results to the face recognition computing model to obtain a global space environment parameter distribution reconstruction result.

[0019] Further, the registration of the satellite on the cloud side is completed, including:

[0020] sending a registration request to the cloud side through the satellite-ground link, so that the cloud side verifies and stores the registration information, and completes the registration of the satellite; wherein the registration information includes: a unique identifier of the satellite, network addressing information, a satellite model, a carried payload condition, a computing and storage capability summary, and a supported communication protocol version.

[0021] Further, the data query request is a data subscription data query request or a one-time data query request.

[0022] Further, generating at least one data query request based on the input data requirement and the list of cooperative satellites includes:

[0023] confirming a first other satellite that at least meets part of the input data requirement based on the list of cooperative satellites and inter-satellite determination;

[0024] discovering other satellites through inter-satellite broadcasting, and obtaining a second other satellite that at least meets part of the input data requirement by querying the edge-side data catalog of the other satellite;

[0025] computing the data provided by the first other satellite and the second other satellite, and combining the input data requirement of the face recognition computing model to obtain data requested from the cloud side;

[0026] Based on the data provided by the first other satellite, the data provided by the second other satellite, and the data requested from the cloud side, corresponding data query requests are generated respectively.

[0027] Further, the input data requirement includes an observation area, an observation time period, a data type, and a data quality, and the data type includes cloud fusion data.

[0028] The cloud side returns data query results, including:

[0029] parsing the data query request to obtain the input data requirement;

[0030] The cloud-side data catalog is queried based on the input data requirements to obtain and return the data that meets the input data requirements, as well as other satellite information that provides the data. The metadata of the cloud-side data catalog includes: satellite UID, data physical parameters, data observation area, data observation time, data quality description, data format and access method. The other satellite information that provides the available data source is used to update the list of cooperating satellites.

[0031] Furthermore, the input data requirements include the observation area, observation time period, data type, and data quality, and the data type includes: other satellite observation data of specific types;

[0032] The other satellite data return query results include:

[0033] Parse the data query request to obtain the required input data;

[0034] The system queries the side data catalog based on the input data requirements, obtains and returns the data that meets the input data requirements, as well as all data information that can be provided by the other satellites. The metadata of the side data catalog includes: data physical parameters, data observation area, data observation time, data quality description, data format and access method. All data information that can be provided by the other satellites is used to update the list of cooperating satellites.

[0035] Furthermore, before the cloud-based and / or other satellites return the data query results, the cloud-based and / or other satellites perform a quality assessment of the data query results; wherein the quality assessment process includes:

[0036] Construct a rule base consisting of several rule items r; where each rule item r includes: rule item number, physical quantity, applicable height range of the rule item r, physical model function F(h,C) for calculating theoretical reference value and tolerance threshold, where h represents height and C represents set of physical condition parameters;

[0037] Based on the height value contained in the data query results, the applicable rule item r is matched from the rule base;

[0038] Get the current set of physical condition parameters;

[0039] Call the physical model function F(h,C) and calculate the theoretical reference value based on the height value contained in the data query results and the current set of physical condition parameters;

[0040] The relative error is calculated based on the observed values ​​contained in the data query results and the theoretical reference values.

[0041] If the relative error is not less than the tolerance threshold and the sensor status shown in the data query results is normal, the data query results are marked as high-quality data.

[0042] In the case that the relative error is greater than the tolerance threshold, or the sensor state displayed in the data query result is normal, the data query result is marked as low-quality data.

[0043] Further, the method further comprises:

[0044] Periodically or triggeredly scanning the data catalog acquired from the cloud side or other satellites;

[0045] Calculating the utility score of each data summary information in the data catalog; wherein the utility score is generated based on the weight corresponding to the region where the data summary information is located, the time freshness weight, the data source weight and the data quality weight;

[0046] In the case that the utility score is greater than the utility threshold of data pre-fetching, a data acquisition request is generated to acquire the data corresponding to the data summary information from the cloud side or other satellites.

[0047] A global space environment parameter distribution reconstruction system based on cloud edge collaboration, the system comprising:

[0048] The registration of the satellite is completed on the cloud side, and the connection state with the cloud side is maintained through the application layer heartbeat mechanism;

[0049] A collaborative satellite list is maintained, which includes other satellite UIDs, inter-satellite link addressing information, key data types provided by other satellites, and inter-satellite reachable states acquired through inter-satellite link detection or negotiation;

[0050] The face recognition calculation model is acquired and parsed to obtain the input data requirements of the face recognition calculation model;

[0051] At least one data query request is generated based on the input data requirements and the collaborative satellite list, and the data query request is sent to the cloud side and / or other satellites to make the cloud side and / or other satellites return data query results; wherein the data query request is generated based on a large language model or a structured format;

[0052] The data query result is provided to the face recognition calculation model to obtain the global space environment parameter distribution reconstruction result.

[0053] An electronic device, comprising a processor and a memory storing computer program instructions; the processor executes the computer program instructions to realize the global space environment parameter distribution reconstruction method based on cloud edge collaboration.

[0054] Compared with the prior art, the present application has significant advantages and beneficial effects compared with the traditional ground "face brushing" calculation scheme and single satellite on-orbit "face brushing" calculation scheme described in the background art. These effects are mainly realized through the key technical features of the present application (such as cloud edge collaborative architecture, cloud side global data service, edge side inter-satellite collaboration, multi-channel active data acquisition, etc.), which can effectively solve the set technical problems.

[0055] (1) Significantly improve the timeliness of "face brushing" calculation results.

[0056] Compared with the prior art:

[0057] The traditional ground "face brushing" calculation is limited by the huge delay (several hours or even longer) of satellite-ground data transmission, and the calculation result is seriously lagging behind the observation, which is difficult to meet the real-time application demand.

[0058] Although the single-satellite on-orbit "face brushing" calculation has been improved, it still needs a single satellite to accumulate its own observation data for a long time (for example, 60 minutes or more), and this data accumulation time becomes the main bottleneck of timeliness.

[0059] Advantages and principles of the present application:

[0060] Through the cloud side global data service mechanism, the satellite (edge side) performing "face brushing" calculation can quickly subscribe or query from the cloud side the multi-satellite historical observation data that has been aggregated, other satellites' recent observation data, and high-value data products after cloud fusion processing. This means that the satellite does not need to accumulate data from scratch, but can directly use these "stock" and "value-added" data to significantly shorten the data preparation time.

[0061] Through the direct data collaboration and sharing method between edge sides (satellites), satellites can obtain the latest observation data from their collaborative satellites in near real time. This inter-satellite data flow avoids the transmission delay of satellite-ground-satellite, providing the latest "incremental" data supplement for "face brushing" calculation.

[0062] Through the multi-channel, multi-level active data acquisition and supply method, combined with the optimized collaborative data flow design, data can be actively and efficiently pushed to the target satellite according to the subscription demand, further reducing the data acquisition waiting time.

[0063] Theoretical expected effects:

[0064] Significantly shorten the data preparation period: theoretically, through cloud side and inter-satellite collaboration, the effective data volume that a single satellite needs to accumulate for 60 minutes can now be obtained within 10-20 minutes or even shorter time (depending on the richness of available data in the collaborative network and communication efficiency). This means that "face brushing" calculation can start earlier;

[0065] • The "face recognition" result delay is significantly reduced: if the data preparation time is shortened from 60 minutes to 15 minutes, the timeliness of the output of the "face recognition" calculation result (relative to the observed event) can be theoretically improved by about 75% ((60-15) / 60*100%). This improvement in timeliness is of decisive significance for the rapid cognition of the space environment and the immediate intelligent response of the satellite.

[0066] (2) Enhance the accuracy and stability of the "face recognition" calculation result.

[0067] • Compared with the prior art:

[0068] • Although traditional ground-based calculations can use multi-source data, the delay in data acquisition makes it difficult to respond to the rapid dynamic changes in the space environment;

[0069] • Single-satellite on-orbit calculations mainly rely on data from the satellite itself, and the data sources are single, the spatial coverage and data dimensions of the observations are limited, which may result in insufficient accuracy of the "face recognition" result in some areas or poor robustness to local disturbances.

[0070] • Advantages and principles of the present application:

[0071] • Through cloud-side global data cataloging and on-demand service mechanism, the satellite can obtain data from multiple satellites in different orbits, different sensors, and different observation periods, as well as high-quality data products processed by complex cloud models. These data far exceed the capabilities of a single satellite in terms of spatial coverage, data types, and data quality;

[0072] • Edge-side inter-satellite collaboration also enables the satellite to obtain synchronous or near-synchronous observation data from different perspectives and positions;

[0073] • Multi-channel data input enables the "face recognition" calculation model to make inferences based on more comprehensive and diverse information.

[0074] • Theoretical expected effects:

[0075] • Improve spatial coverage and resolution: by integrating multi-satellite data, the equivalent observation point density and spatial coverage for "face recognition" calculations can be increased exponentially, allowing for more detailed characterization of the distribution of global space environment parameters and improving the accuracy of the calculation results;

[0076] • Enhance the robustness of the "face recognition" model: data from different sources and perspectives can help correct the bias of single-point observations and smooth noise, making the "face recognition" model more robust to abnormal data and local space environment changes, and the output results more stable and reliable;

[0077] • Improve the ability to capture complex phenomena: for example, for some small-scale spatial environment structures or rapid evolution processes, single-star observations may miss or distort, while multi-star collaborative observation data are more likely to accurately capture

[0078] and reflect in the "face brushing" results.

[0079] (3) Provide stronger data support for satellite intelligence and autonomy.

[0080] • Compared with existing technologies:

[0081] The timeliness and data limitations of existing technologies limit the ability of satellites to make advanced autonomous decisions based on real-time spatial environment perception.

[0082] • Advantages and principles of the present invention:

[0083] The high timeliness, high quality, and wide coverage of the global spatial environment parameter distribution results provided by the present invention are the key data basis for realizing the true intelligence and autonomy of satellites.

[0084] • Theoretical expected effects:

[0085] • Support advanced autonomous behavior: for example, satellites can autonomously perform intelligent behavior (such as orbit maneuver to avoid space debris or harsh space environment), autonomously adjust payload working mode to optimize observation efficiency, or autonomously perform some complex tasks that rely on accurate judgment of the current space environment situation based on the real-time "face brushing" results provided by the present invention. This will greatly improve the on-orbit survival ability, task efficiency, and autonomous operation level of satellites. BRIEF DESCRIPTION OF DRAWINGS

[0086] Figure 1 Schematic diagram of the "face brushing" calculation summary process based on a single satellite in orbit in existing technologies.

[0087] Figure 2 Schematic diagram of the on-board "face brushing" calculation summary process using multi-orbit and multi-time length observation data (including other star observation data).

[0088] Figure 3 Cloud-edge collaborative satellite data service and face brushing system architecture. DETAILED DESCRIPTION

[0089] The present invention will be further described in detail below with reference to the accompanying drawings, which are provided only to explain the present invention and are not intended to limit the scope of the present invention.

[0090] To solve the above technical problems, the application proposes a "face brushing" system and method based on a cloud-edge collaborative satellite data service mechanism. The core idea is to use the collaborative capabilities of the cloud and the edge (edge computing sites within a constellation or constellation) to enable satellites performing "face brushing" calculations in orbit to quickly access observation data accumulated by multiple satellites (including themselves and other satellites) in different orbits and different time periods, thereby significantly shortening the preparation time of high-quality input data, and thus significantly improving the timeliness of "face brushing" calculations, and improving the accuracy and stability of the results with the help of the multidimensionality and universality of the data.

[0091] The summary process of "face brushing" calculation based on the innovative method proposed by the application is shown in Figure 2 Compared with the existing single-satellite in-orbit "face brushing" calculation, the core difference and innovation point of the application is that the cloud-edge collaborative mechanism is introduced in the "data preparation" stage, so that the satellite performing "face brushing" calculation on the edge side can not only rely on its own accumulated data, but also efficiently access and utilize the aggregated data, historical data and other related data provided by other satellites (through other "edge" satellites) and the "cloud" side (such as ground control center or cloud platform).

[0092] To support this "face brushing" calculation process, the application proposes a satellite data service based on a cloud-edge collaborative architecture, which provides timely and rich multi-source data input for on-board "face brushing" calculation through an efficient data sharing and collaborative mechanism, thereby significantly improving the timeliness and accuracy of "face brushing" calculation. The architecture of the proposed system and method is shown in Figure 3 .

[0093] This architecture mainly includes two parts: "cloud side" and "edge side".

[0094] • Edge side: mainly refers to intelligent satellites running in orbit (such as Figure 3 "Satellite 1", "Satellite 2", etc. in ). Each satellite serves as an edge computing site, with data acquisition, preliminary processing, storage, on-board "face brushing" calculation capabilities, and can perform data collaboration tasks with the cloud side and other edge side satellites.

[0095] • Cloud side: usually refers to ground control center, data center or cloud platform. It is responsible for managing all registered edge side satellites, data aggregation and fusion processing, historical data storage, global data cataloging, and providing data services to edge side satellites.

[0096] The system composition, main functional modules and operation process of the application are described in detail below.

[0097] 1. Overview of main functional modules and interaction of the system.

[0098] According to Figure 3The architecture diagram shows that the system mainly includes the following core functional domains, distributed in the cloud side and the edge side.

[0099] • Edge management (cloud side implementation, part of the information provided by the edge side):

[0100] • Edge registration: the edge satellite registers with the cloud side, establishes identity and communication connection;

[0101] • Edge cataloging: the cloud side maintains the catalog information of all registered edge satellites;

[0102] • State update / maintenance: maintain state synchronization between edge and cloud, and between edge and edge.

[0103] • Data cataloging (both cloud and edge involved):

[0104] • Data provider registration: edge satellites (as data providers) register with the cloud side for the data information they can provide; in inter-satellite collaboration, satellites can also share their data providing capabilities with other satellites;

[0105] • Metadata query: edge satellites can query the global data catalog of the cloud side, or query the data catalog of the collaborative satellites.

[0106] • Data service (both cloud and edge involved):

[0107] • Data subscription: edge satellites can subscribe to the required data from the cloud side, or subscribe to data from other collaborative satellites;

[0108] • Data push and transmission: the cloud side pushes data to edge satellites according to subscription; collaborative satellites push data to requesting satellites according to subscription;

[0109] • Data query: edge satellites can directly initiate real-time data query to the cloud side or other collaborative satellites.

[0110] • Data transmission:

[0111] • Data download (edge to cloud): edge satellites download local observation data or processing results to the cloud side;

[0112] • Data upload (cloud to edge): the cloud side uploads processed data, instructions or other satellite data to the target edge satellite;

[0113] • Inter-satellite data transmission (edge to edge): direct data transmission between edge satellites.

[0114] 2. Main functions and processes of the cloud side.

[0115] As the central coordination and data aggregation node of the system, the main functions and processes of the cloud side include the following.

[0116] (1) Satellite edge management.

[0117] • Edge registration and status maintenance process:

[0118] • After the satellite (edge) is started, it sends a registration request to the cloud side through the satellite-ground link. The registration information includes the unique identifier (UID) of the satellite, network addressing information (such as IP address or other communication protocol address), satellite model, carried payload, main computing and storage capability summary, and supported communication protocol version, etc. The cloud side verifies and stores the registration information, and completes the registration of the satellite edge site;

[0119] • After successful registration, the edge side and the cloud side maintain the connection state through the application layer heartbeat mechanism. The heartbeat information can include the basic health status, the current main task mode (for example, observation, calculation, sleep), the expected communication window, etc. For planned shutdown sleep or temporary communication interruption, the edge side can optionally report the state, and the cloud side maintains the real-time accessibility state of each edge satellite according to the actual situation, in order to facilitate subsequent data pushing and task scheduling.

[0120] • Edge catalog management and service process:

[0121] • The cloud side collects the meta information of all registered edge satellites to form a global edge satellite catalog;

[0122] • The cloud side provides query services for the edge situation to the outside (for example, to other ground application systems or to the edge satellite that makes a request). The query result can include the UID, addressing information, current accessibility state, main capability description (such as what kind of "face recognition" calculation can be performed, what kind of observation data can be provided), and its approximate orbital position or coverage area, etc.

[0123] (2) Data catalog management (global data catalog on the cloud side).

[0124] • Data provider registration: when the edge satellite has data downlink capability and is willing to share its data, it will send a data item provider registration request to the cloud side. This request describes in detail the data types and their metadata that the satellite can provide. These metadata include but are not limited to:

[0125] • Satellite UID;

[0126] • Physical parameters of observation data: such as neutral atmospheric density, electron density, magnetic field strength, etc.

[0127] • Spatio-temporal characteristics of data: such as spatial scale (large scale, medium scale, small scale) covered by data, height range, historical time range or future planned observation time window that can provide data, geographical range covered by orbit (such as specific latitude and longitude area, specific orbital plane);

[0128] • Data quality description: such as precision, resolution, confidence, etc.

[0129] • Data format and access method. After receiving this information, the cloud side integrates it into the global data catalog. This process also needs to consider the satellite's planned on-off, load working mode, etc., to ensure the accuracy of the catalog information.

[0130] • Data catalog query: the cloud side provides a query service for the global data catalog. Satellites (or other data users) that need to perform "face recognition" calculations can query the cloud-side data catalog according to their needs (such as specific physical parameters, spatiotemporal range, data quality requirements, etc.) to find available data sources and their providers (which may be historical downlink data from other satellites or cloud-side fusion processed products) that meet the conditions. The information returned by the query will include the provider UID (or cloud-side data product ID), detailed data metadata description, etc.

[0131]

[0132] (3) Data service (cloud side facing edge side).

[0133] • Data subscription function:

[0134] • Data subscription process: edge satellites that need to perform "face recognition" calculations can initiate a data subscription request to the cloud side according to their "face recognition" calculation model's requirements for input data (e.g., the need for other satellite observation data or cloud-side fusion data in a specific region, time period, and type). The subscription request contains the UID of the satellite and a detailed description of the required data (physical parameters, scale, height, time range, spatial range, data update frequency requirements, etc.);

[0135] • Data push and cloud-edge transmission: after receiving the subscription request, the cloud side will continuously monitor the data it collects (downlink data from other satellites, historical archive data, cloud-side model output, etc.). Once data that meets the subscription conditions is generated or updated, the cloud side will actively push the data to the subscribing edge satellite according to the target satellite's operating status (such as reachability, communication window, on-board storage and processing capacity, etc.). Data transmission can use the satellite-ground link (data upload).

[0136] • Data query function:

[0137] • Edge satellites can also initiate a one-time data query request to the cloud side to obtain real-time or historical data that meets specific conditions. The query conditions are similar to the data description in the subscription request. The cloud side retrieves the results from the global database according to the query conditions and returns them to the requesting satellite.

[0138] 3. Main functions and processes of the edge side (including inter-satellite collaboration). ​

[0139] Edge-side satellites are not only the original producers of data and the executors of "face recognition" computing, but also active nodes in the cloud-edge collaborative network, capable of direct collaboration with other satellites.

[0140] (1) Star-ground-cloud-edge collaboration.

[0141] ●As Figure 3 As shown in the architecture, the "star-ground-cloud-edge collaboration" module of the edge-side satellite is responsible for the above-mentioned registration, state reporting, data catalog information synchronization, data subscription, data query, and data upload and download interactions with the cloud side.

[0142] ●The "face recognition" computing module inside the edge-side satellite will request data from the cloud side through this collaboration module when it needs data. The obtained cloud-side data (such as other-satellite historical data, cloud-side fusion products) are cached and updated as needed for use by the "face recognition" computing module.

[0143] (2) Inter-satellite edge-edge collaboration.

[0144] In order to further improve the real-time performance of data acquisition, especially in the absence of immediate star-ground communication links or in the case of extremely low-latency data, the present application supports direct collaboration between edge-side satellites.

[0145] ●Edge-side other-satellite state awareness and list maintenance:

[0146] ●The "inter-satellite edge-edge collaboration" module inside each edge-side satellite (such as satellite 1) will maintain a potential collaborative other-satellite list based on the needs of "face recognition" computing and other tasks for nearby or related other-satellite data. This list can be initially obtained from the cloud side (based on the global edge-side catalog) and dynamically maintained through inter-satellite detection or cloud-side updates. The list contains the UID, addressing information (such as inter-satellite link address), key data types that can be provided, and inter-satellite reachable state (such as whether the link is established, bandwidth, delay, etc.) of the other-satellite obtained through inter-satellite link detection or negotiation.

[0147] ●Inter-satellite data catalog interaction:

[0148] ●Data provider registration (inter-satellite): When a satellite (such as satellite 2) provides data to its neighboring collaborative satellite (such as satellite 1) through an inter-satellite link, it sends the metadata information of the data items it can provide to satellite 1 through the inter-satellite communication protocol, completing the "data provider" capability update in satellite 1's collaboration list. This is similar to data provider registration on the cloud side, but occurs inter-satellite.

[0149] ●Data catalog query (inter-satellite): Satellite 1 can initiate a data catalog query request to other satellites (such as satellite 2) in its collaboration list to understand the real-time observation data details (physical parameters, spatio-temporal range, etc.) that satellite 2 can currently provide.

[0150] • Inter-satellite data service:

[0151] • Data subscription process (inter-satellite): Satellite 1 initiates a data subscription request to its co-listed satellite (e.g., Satellite 2) based on the instant demand calculated by "face recognition", requesting Satellite 2 to push data in real time through inter-satellite link when data meeting the conditions is generated.

[0152] • Data push and inter-satellite transmission: After receiving the subscription, Satellite 2 will directly push data to Satellite 1 through inter-satellite link if data meeting the conditions is generated.

[0153] • Data query (inter-satellite): Satellite 1 can also initiate an instant data query request to Satellite 2, and Satellite 2 returns the result through inter-satellite link after processing.

[0154] In summary, the cloud-edge collaborative mechanism proposed in the present application significantly improves the timeliness and data quality of on-board "face recognition" calculation in the following ways:

[0155] • Shorten data accumulation time and improve timeliness:

[0156] • The satellite performing "face recognition" calculation no longer relies solely on its own long-term (e.g., more than 60 minutes) accumulated observation data.

[0157] • By subscribing / querying to the cloud side, historical observation data of other satellites and high-quality data products processed by the cloud can be quickly obtained, which may cover different orbits and longer time spans.

[0158] • Through inter-satellite collaborative subscription / query, the latest observation data from neighboring satellites can be obtained in near real time, further shortening the data acquisition delay.

[0159] • This multi-source data input mechanism enables the satellite to collect a sufficient amount and coverage of input data in a shorter time (e.g., 10-20 minutes or less), thereby enabling earlier initiation of "face recognition" calculation and greatly improving the timeliness of the output results.

[0160] • Increase data dimensions and total amount, improve accuracy and stability:

[0161] • The data obtained from the cloud side includes multi-satellite, multi-orbit, and multi-time length observations, and may even include higher precision data processed by complex ground system correction and fusion.

[0162] • The data obtained from inter-satellite collaboration supplements the spatial and temporal blind areas of the satellite's own observations.

[0163] • More abundant, multi-dimensional (different latitudes and longitudes, different altitudes, different time periods, different sensor types) data input helps the "face recognition" calculation model to more comprehensively depict the complex characteristics of the spatial environment, thereby improving the accuracy and stability of the prediction results of the global spatial environment parameter distribution.

[0164] 4. Star 'face recognition' calculation application process based on cloud edge collaborative data service.

[0165] The cloud edge collaborative data service system provided by the application aims to provide efficient, multi-source data support for star "face recognition" calculation applications. A typical star "face recognition" calculation application using the data service process of the application can be divided into a deployment and data preparation phase and a "face recognition" calculation execution phase.

[0166] (1) Deployment and data preparation phase (data collaboration and continuous acquisition before the "face recognition" calculation model is deployed and runs).

[0167] • Data requirement analysis and subscription initiation.

[0168] • When the "face recognition" calculation model is completed on the edge satellite, the "data collaboration and management module" embedded or associated with it will analyze the specific requirements of the model for input data. These requirements usually include but are not limited to: physical parameters of the required observation data (such as neutral atmospheric density, electron density, etc.), target spatial area (such as specific latitude and longitude range, altitude layer), time window (such as historical data range, future prediction timeliness requirement corresponding to observation data freshness), orbital plane information, data accuracy and resolution requirements, etc.

[0169] • Based on the analyzed data requirements, the module will actively initiate a data subscription request.

[0170] • Cloud-side subscription: through the "star-ground-cloud edge collaboration" module of the satellite, when the satellite is reachable, a data subscription request is initiated to the cloud-side data service platform. The request contains the UID of the satellite and a detailed description of the data requirements. The cloud side will push the data to the satellite according to the subscription when it meets the conditions (such as historical downlink data from other satellites, cloud fusion data products, etc.).

[0171] • Subscription to other satellites (inter-satellite collaboration): at the same time, the "data collaboration and management module" will query (first through the cloud side to obtain global edge catalog information, then confirm inter-satellite, or directly broadcast discovery / point-to-point query) the data catalog of other collaborative satellites (edge nodes) in orbit to identify which other satellites can provide real-time or recent observation data that meet the requirements. Once a suitable data provider satellite is found and the inter-satellite is reachable, the "data collaboration and management module" will initiate a data subscription request to the target other satellite through the "inter-satellite edge collaboration" module.

[0172] • Continuous data reception and cache management.

[0173] • Data reception module of the edge satellite remains in working state to receive data pushed from the cloud side or other satellites according to subscription.

[0174] • Received data is stored in the local storage unit on the satellite and is uniformly managed by the "data coordination and management module", and key data will be sent to a "data cache area" (the "cache" module in the architecture diagram) specially designed for "face brushing" calculation.

[0175] • This "data cache area" will be dynamically updated according to the arrival of new data, for example, using a timestamp-based update strategy (retaining the latest data), or setting priorities for replacement according to data sources, quality, etc. Data in the cache usually has metadata tags (such as observation time, source satellite UID, spatial position information, etc.) for subsequent "face brushing" calculation module calls.

[0176] (2) "Face brushing" calculation execution phase (data utilization and calculation result generation).

[0177] • Calculation trigger and data acquisition:

[0178] • The "face brushing" calculation task can be started according to the preset periodic timing, or triggered by a specific event (such as receiving key data updates, ground instruction triggering, specific spatial environment conditions being met, etc.).

[0179] • After the calculation task is started, the "face brushing" calculation module first accesses the "data cache area" and selects appropriate data sets according to the requirements of this calculation (for example, the latest data combination covering a specific area is required).

[0180] • Multi-source data fusion and utilization:

[0181] • The "face brushing" calculation module will make full use of all multi-source data in the "data cache area" that meets the conditions. This includes:

[0182] • Real-time or recent observation data of the satellite;

[0183] • Historical data or global fusion data output from the cloud side push;

[0184] • Real-time or near real-time observation data from collaborative satellites pushed through inter-satellite links;

[0185] • Before computation, the module can perform preliminary filtering, registration, quality assessment and fusion of the selected data from different orbits, different altitudes, different observation times, and different sensor types to construct an optimized input data set, with particular emphasis on using the latest data to ensure the timeliness of the results.

[0186] • Computation and result output:

[0187] • Using the constructed multi-source input data set, the "face recognition" calculation model executes its core algorithm to infer the distribution of global space environment parameters.

[0188] • The "face recognition" results obtained by calculation (i.e., global space distribution prediction results) can be used to support satellite autonomous decision-making (such as orbit adjustment, task planning, etc.), stored on the satellite for subsequent analysis, or downloaded to the ground application system through the satellite-ground link.

[0189] 5. Semantic matching module.

[0190] In a preferred embodiment, the above-mentioned data catalog management and inter-satellite edge-edge collaboration also include a semantic matching module. When the semantic matching module receives a catalog matching request from the edge side or a subscription / query request from another satellite, the semantic matching module parses the physical semantic tags in the request and performs semantic-level data matching based on the physical tags and physical semantics, and then pushes the data accordingly, making the data service more accurate for space science tasks.

[0191] Specifically, the typical embodiments of the semantic matching module are large language models and semantic tag matching algorithms. Moreover, the core parameters of the large language models and the semantic tag matching algorithms can be updated based on the cloud-edge collaboration mechanism through cloud-side or other satellite transmission. The following will describe the two embodiments respectively.

[0192] (1) Embodiment based on large language model.

[0193] The core of the semantic matching module is to "compile" the user's semantic request into a deterministic and structured query object through the large language model, which is then used to distribute to data providers (cloud-side ground or other edge-side satellites) for data query processing or data push filtering. This embodiment is suitable for system environments with general-purpose deep learning accelerators.

[0194] ① Model training (cloud side, offline / non-real-time) - semantic query compiler based on large language model

[0195] • Core component: The core is a semantic query compiler based on large language model, which is implemented by a large language model fine-tuned by instructions.

[0196] ■ Instruction task of large language model: Design precise instructions for large language model to make it clear its role and task. For example, tell it that its role is a space physics and satellite data system expert, and its task is to convert natural language data requests into a structured query in JSON format for querying system data catalog, while specifying the available fields of data catalog (such as satellite_ID, timestamp_utc, altitude_km, physical_param, observed_value, quality_flag, semantic_tags_set, etc.). At the same time, specify the inference rules including: identifying key physical quantities from the request, parsing spatio-temporal constraints, mapping fuzzy semantic concepts to labels in the semantic_tags_set (for example, "geomagnetic storm" should be mapped to the label GEOMAG_STORM; "solar flare" should be mapped to SOLAR_FLARE), converting qualitative descriptions of data values (such as 'abnormal increase') to quantitative comparisons of the observed_value field, and combining all conditions with AND / OR logic, etc. The input is natural language request NL_Query, and the output is JSON format query.

[0197] ② Data subscription party (cloud side or edge side, online / real-time) semantic matching execution

[0198] • Data stream input: subscription request NL_Query expressed in natural language text form from users or edge satellite;

[0199] ■ Example of NL_Query: "Subscribe to low-orbit high-altitude high-density atmospheric data caused by super solar wind storms similar to 'Carrington event' level".

[0200] • Core execution: Based on the large model trained on the ground in the cloud, execute the compilation and execution of semantic query, that is, take NL_Query as input, generate a structured query object Structured_Query in JSON format for querying system data catalog. The parameters of the large language model can be updated through the cloud side based on the cloud-edge collaborative mechanism;

[0201] • Data stream output: output a structured JSON query object Structured_Query.

[0202] ③ Data provider (cloud side or edge side, online / real-time) execution of structured query

[0203] • A lightweight Structured_Query object is sent to the data provider (e.g., a global data catalog on the cloud side, or sent to a peer on the edge side via inter-satellite link);

[0204] • The JSON structure is parsed and converted to a query operation on its local data catalog (a regular database or file index) by a regular execution engine;

[0205] • The query is executed, i.e., either directly processed or data push filtered, to return all matching data records, upon which the data service module proceeds to return the results or push the data.

[0206] (2) An embodiment based on semantic tag matching algorithm.

[0207] This embodiment has no special requirements for the system environment, including resource-constrained edge-side satellites, and does not require deep learning accelerators, etc. ① Key structure: standardized and hierarchical tag system.

[0208] • Unified tag dictionary: all nodes (cloud-side ground and edge-side satellites) in the system use a unified, versioned "semantic tag dictionary". It is maintained by ground domain experts and can be updated through cloud-side transmission based on cloud-edge collaboration mechanism;

[0209] • Hierarchical structure of tags: tags are separated by "." for hierarchical separation to express rich and accurate meanings, which allows fuzzy or accurate matching. Examples:

[0210] ■Event class: EVENT.SOLAR.FLARE.X_CLASS, EVENT.GEOMAG.STORM.G4_SEVERE

[0211] ■Status class: STATUS.SENSOR.QUALITY.HIGH_CONFIDENCE, STATUS.DATA.AVAILABILITY.REALTIME

[0212] ■Orbit class: ORBIT.REGION.POLAR_NORTH, ORBIT.ALTITUDE.LEO_400KM

[0213] • Application of tags in data catalog:

[0214] ■In the data metadata, the semantic_tags_set field is stored, which contains a series of standardized tags describing the data;

[0215] • Example: During a G4 geomagnetic storm, high quality data acquired by a normally working sensor in the polar region, its tag set can be: {'EVENT.GEOMAG.STORM.G4_SEVERE', 'STATUS.SENSOR.QUALITY.HIGH_CONFIDENCE', 'ORBIT.REGION.POLAR_NORTH'}.

[0216] • Example: During a G4 geomagnetic storm, high quality data acquired by a normally working sensor in the polar region, its tag set can be: {'EVENT.GEOMAG.STORM.G4_SEVERE', 'STATUS.SENSOR.QUALITY.HIGH_CONFIDENCE', 'ORBIT.REGION.POLAR_NORTH'}.

[0217] ② Structured tag query request Tag_Query_Request from the data subscriber.

[0218] • The subscription request is in a structured format (e.g. JSON format). Example of a request: subscribe to data that satisfies all of the following three conditions: (1) the tag set must contain both G4_SEVERE (G4 level severe geomagnetic storm) and HIGH_CONFIDENCE (high confidence); (2) the tag set must contain either POLAR_NORTH (polar north region) or POLAR_SOUTH (polar south region); (3) the tag set must not contain DEGRADED (degraded data quality) or SENSOR.ANOMALY (sensor anomaly).

[0219] ③ Workflow of the semantic matching module.

[0220] • Input data stream: structured tag query request Tag_Query_Request;

[0221] • Core component: the semantic matching module with the deterministic rule conversion engine as the core;

[0222] • Workflow (data subscriber):

[0223] • Verification: the module first receives Tag_Query_Request and verifies whether all tags in the request are legal and defined tags according to the semantic tag dictionary. If there is an illegal tag, an error is returned;

[0224] • Conversion / compilation: the module traverses the clauses in Tag_Query_Request and converts each clause into a standard query condition that can be executed downstream;

[0225] • Output data stream: the module finally outputs a JSON object Structured_Query that is exactly the same as the output format of the above embodiment based on a large language model;

[0226] • Subsequent process (data provider): This generated Structured_Query object, as the one produced in the above described embodiment based on large language models, is sent to the data provider (cloud side or other satellite). After the data provider converts the JSON structure into a query operation on its local data catalog, it directly performs query processing or data push filtering, and the data service module accordingly performs subsequent result return or data push.

[0227] 6. Data quality assessment module.

[0228] Before the cloud side and / or other satellite returns the data query result, the cloud side and / or other satellite also need to perform quality assessment on the data query result.

[0229] The data quality module embeds simplified physical rules of physical quantities (including: ① For neutral atmospheric density physical quantity: reasonable range of atmospheric density at a certain altitude; ② For electron density physical quantity: normal mode of day-night variation of electron density). If the data is significantly abnormal (possibly caused by sensor failure), it is marked as "low confidence", and the data service will match the data subscription requirements according to the relevant assessed data quality (the data service will filter out low-confidence data that does not meet the conditions when responding to subscriptions or queries according to the requester's requirements for data quality (for example, only accept high-confidence data)), and then push or query return, thereby realizing precise service oriented to data quality.

[0230] Specifically, the kernel of the data quality assessment module can be a machine learning model, a deep learning network, or a configurable rule base. Moreover, the model, network, or rule base can be updated based on cloud-edge collaboration mechanism through cloud-side or other satellite transmission. The following will take the rule base as an example for introduction.

[0231] The rule base R is a set of rule items. Each rule item r (where r ∈ R) is defined as: r = {rule_ID, physical_param, altitude_range, F(h, C), θ tolerance}, where rule_ID represents the rule item number, physical_param represents the physical quantity, such as "neutral atmospheric density" or "electron density", altitude_range represents the altitude range applicable to this rule item, F(h, C) represents the physical model function used to calculate the theoretical reference value, h is the altitude, C is the set of variable physical condition parameters (such as solar activity index, geomagnetic index, etc., which can be periodically transmitted from the cloud side or other satellite in a cloud-edge collaborative manner), and θ tolerance represents the tolerance threshold.

[0232] For example, taking neutral atmospheric density physical quantity as an example, a physical model function F(h, C) or F density (h, C) of one example is as follows:

[0233]

[0234] Wherein, the density p ref and the altitude H are not fixed constants, but functions changing with external conditions C; p ref is the atmospheric density at the reference height h ref , H is the altitude, and η is a system parameter.

[0235] Based on the above rule base, the evaluation logic and data flow of the data quality evaluation module are as follows:

[0236] 1) The module receives a data packet Pkg_obs;

[0237] 2) According to the height h of Pkg_obs, the applicable rule r is matched from the rule base R;

[0238] 3) The latest physical condition parameter C (such as the solar activity index) is obtained;

[0239] 4) The model function F(h, C) in the rule is called to calculate the theoretical reference value p model ;

[0240] 5) The relative error is calculated, wherein p obs is the observed density value in the observation data Pkg_obs;

[0241] 6) In the case that the relative error is not less than the tolerance threshold, and the sensor state displayed in the data query result is normal, the data query result is marked as high-quality data; in the case that the relative error is greater than the tolerance threshold, or the sensor state displayed in the data query result is normal, the data query result is marked as low-quality data;

[0242] 7) A data packet Pkg_tagged = Pkg_obs o quality_flag with a quality label is generated, wherein the data packet Pkg_tagged is used for subsequent data services.

[0243] 7. Cache management module.

[0244] The cache management module is a component of the "data collaboration and management module" and is responsible for data storage, updating, and prefetching strategy. The cache management module adopts an intelligent subscription and push strategy, including: on the one hand, the cache module stores the (data-driven) data pushed passively; on the other hand, according to the internal needs of the face recognition calculation model (model-driven), the inferred data that will be needed is actively prefetched. For example, if the model shows that it is more sensitive to polar region data changes, the cache module will automatically increase the priority of obtaining polar region data, even if there is no explicit subscription request.

[0245] Specifically, when the face recognition calculation model is deployed, it will register a sensitivity profile file Profile_M with the cache management module. The sensitivity profile file can be updated based on the cloud-edge collaboration mechanism through cloud-side or exoplanet transmission. Profile_M is composed of the weight of the model for different regional data region_weights, the time freshness decay factor λ, the data source preference source_preference, the quantitative value of data quality quality_value, and the utility threshold θ of data prefetching prefetch Several parts.

[0246] The data stream and judgment logic actively prefetched by the cache management module are as follows:

[0247] 1) The cache management module periodically (or triggered) scans the data catalog obtained from the cloud side or exoplanet;

[0248] 2) For the relevant data summary information d cand in the catalog, the module precomputes its utility score S utility (d cand ) = W region · W timeliness · W source · W quality ; where W region represents the weight of the region where the data or data summary information d is located, the time freshness weight W timeliness = exp(-λ(t now -t obs )), t now represents the current time, t obs represents the data acquisition time, W source represents the data source weight, and W quality represents the data quality weight;

[0249] 3) If S utility (d cand )> θ prefetchIf, the module automatically generates a data acquisition request (subscription or query) and sends it to the "star-ground cloud edge collaboration / inter-satellite edge edge collaboration module", realizing active prefetching. Wherein, θ prefetch from Profile_M;

[0250] 4) When the cache space is insufficient, the module calculates the current utility score of all data in the cache, and preferentially eliminates the data with the lowest score.

[0251] 8. Network state awareness and prediction module.

[0252] The network state awareness and prediction module is a submodule of the "star-ground cloud edge collaboration" and "inter-satellite edge edge collaboration" modules, and provides a basis for the external communication decision of the two modules. The module dynamically senses and predicts future data communication opportunities (including when to establish a connection with the ground station, when to establish with a certain collaborative satellite, estimated bandwidth and duration). The execution of data subscription and the pushing of data will be intelligently scheduled according to this prediction. For example, a non-urgent data request can be scheduled for the next broadband data communication window, while a high-priority data may attempt to be packaged and compressed for transmission through a narrower but available inter-satellite link.

[0253] In summary, the present application is based on the cloud edge collaboration system architecture, and the functional positioning of the "cloud side" (such as the ground center) and the "edge side" (such as the on-orbit intelligent satellite) is clear. The present application proposes a cloud-side global data cataloging and on-demand service mechanism based on the cloud edge collaboration architecture: the cloud side constructs and maintains a global and dynamically updated "data catalog", which integrates data providing capability information (such as data type, spatiotemporal range, physical parameters, etc.) from each registered satellite edge side and the cloud's own processing generated fusion data output. Based on this catalog, the cloud side can provide a single satellite edge side with cross-multiple-star, multiple-orbit, and multiple-duration data query and subscription push services for "face recognition" computing needs. This enables a single satellite to obtain data far beyond its own observation range and accumulation capacity.

[0254] The present application not only designs data collaboration between the cloud and the edge, but also innovatively introduces a direct data collaboration mechanism between the edge side satellites. This includes the processes of data providing capability registration, data catalog query, data subscription, and data pushing between edge side satellites. This inter-satellite "edge-edge" collaboration can achieve lower latency data sharing, especially suitable for "face recognition" computing scenarios with extremely high real-time requirements, and is an effective supplement and enhancement to cloud-edge collaboration.

[0255] In order to meet the requirements of diversity and timeliness of input data for face recognition calculation on satellite, an innovative data acquisition and supply method is proposed. The satellite performing face recognition calculation can actively acquire data through at least two main channels: one is to subscribe or query (acquire global data, historical data, other satellite downlink data, etc.) to the cloud side through the "star-ground cloud edge collaboration" module; the other is to subscribe or query (acquire real-time or recent observation data of adjacent satellites) to the collaborative satellites through the "inter-satellite edge edge collaboration" module. This multi-channel, on-demand subscription and active push data supply mode is the key to ensure that data is quickly and accurately reached. At the same time, around the core goal of improving the timeliness and data quality of face recognition calculation, a series of collaborative data processes are designed, including but not limited to: satellite edge side registration and state maintenance process in the cloud side, data provider (satellite or cloud data product) registration process in the cloud side and inter-satellite, metadata-based data cataloging and querying process, and data subscription and conditional push process based on specific data requirements (such as physical parameters, space-time range). Through the above cloud edge / edge edge collaborative data supply capability workflow, the optimization goal of improving the timeliness and accuracy of face recognition calculation on satellite is finally achieved.

[0256] The cloud edge collaborative data service mechanism proposed in the application can ensure that the face recognition calculation application on satellite can obtain timely observation data from multiple satellites, covering multiple orbits, spanning multiple time lengths, and as recent as possible, thereby effectively solving the problems of slow data accumulation and single data source for a single satellite, and significantly improving the timeliness, accuracy and stability of face recognition calculation.

[0257] Although specific embodiments of the application are disclosed for the purpose of illustration and description, the purpose being to help understand the content of the application and to implement it, those skilled in the art can understand that various substitutions, changes and modifications are possible without departing from the spirit and scope of the application and the appended claims. Therefore, the application should not be limited to the disclosed content of the best mode, and the scope of the application claimed is defined by the scope of the claims.

Claims

1. A method for reconstructing the distribution of global spatial environment parameters based on cloud-edge collaboration, characterized in that, Applied to a satellite, the method includes: The satellite is registered on the cloud side and its connection with the cloud side is maintained through the application layer heartbeat mechanism. Maintain a list of cooperating satellites, which includes: other satellite UIDs, inter-satellite link addressing information, key data types provided by other satellites, and inter-satellite reachability obtained through inter-satellite link detection or negotiation; Obtain and parse the face recognition computing model to get the input data requirements of the face recognition computing model; Based on the input data requirements and the list of cooperating satellites, at least one data query request is generated and sent to the cloud side and / or other satellites so that the cloud side and / or other satellites can return the data query results; wherein, the data query request is generated based on a large language model or a structured format; The data query results are provided to the facial recognition computing model to obtain the reconstruction results of the distribution of spatial environmental parameters across the entire domain.

2. The method according to claim 1, characterized in that, The registration of the satellite on the cloud side includes: The satellite sends a registration request to the cloud via a satellite-to-ground link, enabling the cloud to verify and store the registration information, thus completing the satellite's registration. The registration information includes: the satellite's unique identifier, network addressing information, satellite model, payload details, an overview of computing and storage capabilities, and supported communication protocol versions.

3. The method according to claim 1, characterized in that, The data query request can be a data subscription query request or a one-time data query request.

4. The method according to claim 1, characterized in that, Based on the input data requirements and the list of collaborating satellites, at least one data query request is generated, including: Based on the list of cooperating satellites and inter-satellite determination, the first other satellite that at least meets some of the input data requirements is identified; Other satellites are discovered using inter-satellite broadcasting, and by querying the side data catalog of these other satellites, a second other satellite that at least partially meets the input data requirements is obtained; Calculate the data provided by the first and second other satellites, and combine it with the input data requirements of the facial recognition calculation model to obtain the data requested from the cloud side; Based on the data provided by the first other satellite, the data provided by the second other satellite, and the data requested from the cloud, corresponding data query requests are generated respectively.

5. The method according to claim 1, characterized in that, The input data requirements include the observation area, observation time period, data type, and data quality. The data type includes cloud-integrated data. The cloud-side returned data query results include: Parse the data query request to obtain the required input data; The cloud-side data catalog is queried based on the input data requirements to obtain and return the data that meets the input data requirements, as well as other satellite information that provides the data. The metadata of the cloud-side data catalog includes: satellite UID, data physical parameters, data observation area, data observation time, data quality description, data format and access method. The other satellite information that provides the available data source is used to update the list of cooperating satellites.

6. The method according to claim 1, characterized in that, The input data requirements include the observation area, observation time period, data type, and data quality. The data type includes: other satellite observation data of specific types. The other satellite data return query results include: Parse the data query request to obtain the required input data; Based on the input data requirement, the side data catalog is queried to obtain and return the data that meets the input data requirement, as well as all data information that can be provided by the other satellites. The metadata of the side data catalog includes: data physical parameters, data observation area, data observation time, data quality description, data format and access method. All data information that can be provided by the other satellites is used to update the list of cooperating satellites.

7. The method according to claim 1, characterized in that, Before the cloud-based and / or other satellites return the data query results, the cloud-based and / or other satellites perform a quality assessment of the data query results; wherein, the quality assessment process includes: Construct a rule base consisting of several rule items r; where each rule item r includes: rule item number, physical quantity, applicable height range of the rule item r, physical model function F(h,C) for calculating theoretical reference value and tolerance threshold, where h represents height and C represents set of physical condition parameters; Based on the height value contained in the data query results, the applicable rule item r is matched from the rule base; Get the current set of physical condition parameters; Call the physical model function F(h,C) and calculate the theoretical reference value based on the height value contained in the data query results and the current set of physical condition parameters; The relative error is calculated based on the observed values ​​contained in the data query results and the theoretical reference values. If the relative error is not less than the tolerance threshold and the sensor status shown in the data query results is normal, the data query results are marked as high-quality data. If the relative error exceeds the tolerance threshold, or if the sensor status shown in the data query results is normal, the data query results are marked as low-quality data.

8. The method according to claim 1, characterized in that, The method further includes: Periodically or triggeredly scan and catalog data acquired from the cloud or other satellites; Calculate the utility score of each data summary information in the data catalog; wherein the utility score is generated based on the weight of the region where the data summary information is located, the weight of time freshness, the weight of data source, and the weight of data quality; If the utility score is greater than the utility threshold for data prefetching, a data retrieval request is generated to retrieve the data corresponding to the data summary information from the cloud or another satellite.

9. A global spatial environment parameter distribution reconstruction system based on cloud-edge collaboration, characterized in that, The system includes: The satellite is registered on the cloud side and its connection with the cloud side is maintained through the application layer heartbeat mechanism. Maintain a list of cooperating satellites, which includes: other satellite UIDs, inter-satellite link addressing information, key data types provided by other satellites, and inter-satellite reachability obtained through inter-satellite link detection or negotiation; Obtain and parse the face recognition computing model to get the input data requirements of the face recognition computing model; Based on the input data requirements and the list of cooperating satellites, at least one data query request is generated and sent to the cloud side and / or other satellites so that the cloud side and / or other satellites can return the data query results; wherein, the data query request is generated based on a large language model or a structured format; The data query results are provided to the facial recognition computing model to obtain the reconstruction results of the distribution of spatial environmental parameters across the entire domain.

10. An electronic device, characterized in that, The electronic device includes: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements the cloud-edge collaborative global spatial environment parameter distribution reconstruction method as described in any one of claims 1-8.

Images