Remote sensing satellite downlink data instant service method and system, and electronic equipment

Abstract

The invention discloses a remote sensing satellite downlink data instant service method and system and electronic equipment, and relates to the technical field of remote sensing information engineering or other related technical fields. The method specifically establishes remote sensing information instant service integrating satellite downlink data access, satellite downlink data instant geometric positioning, target information rapid extraction and remote sensing information active service, and comprises the following steps: accessing received remote sensing satellite downlink data, and obtaining the remote sensing satellite downlink data; carrying out geometric correction on the satellite downlink data by adopting a multi-satellite image collaborative uncontrolled geometric positioning strategy to obtain instant geometric positioning data; inputting the real-time geometric positioning data into a deep learning model of predetermined ground object target identification to obtain specific target thematic information; and intelligently pushing a corresponding image content text description corpus generated based on the instant geometric positioning data and the target thematic information to the user, and optimizing an image cognition calculation model by using information fed back by the user, thereby improving the precision of a remote sensing data information service product.

Description

Technical Field

[0001] This invention relates to the field of remote sensing information engineering technology or other related fields. Specifically, it relates to a method, system, and electronic equipment for providing real-time downlink data services from remote sensing satellites. Background Technology

[0002] Currently, global Earth observation has developed strong technological capabilities and a robust system architecture, possessing the ability to acquire high-precision, high-spatial-resolution data on the atmosphere, ocean, and land, ushering in the era of big data in Earth observation. However, for a long time, satellite ground systems have operated according to a "receive-process-archive-distribute" model. Remote sensing data is mostly distributed and shared through online downloads and browsing, with thematic target information obtained through offline data processing. This remote sensing information processing is primarily conducted on standard data that has undergone system geometric and radiometric corrections, resulting in low processing efficiency and failing to meet the needs of users in the era of remote sensing big data for rapid and dynamic acquisition of remote sensing data and information.

[0003] On the one hand, in terms of service content, the shortcomings of traditional remote sensing information processing methods in terms of processing efficiency and accuracy limit the mining and application of remote sensing information. Services remain at the data level, rather than providing the information and knowledge that users truly need. This is mainly reflected in:

[0004] 1. There are still many shortcomings in the key technologies for rapid geometric positioning of remote sensing satellite downlink data, which cannot meet the needs of real-time satellite remote sensing image services. The main problems are: (1) Due to the accuracy limitations of the attitude and orbit control system of the onboard sensor, there are many uncertainties in the strict imaging model. If these errors cannot be effectively suppressed, the direct geometric positioning accuracy of the image will be limited. At the same time, it is difficult to achieve high geometric positioning accuracy when using control data for geometric fine processing; (2) Due to the limitations of direct positioning accuracy, it is usually necessary to use a certain number of ground control points to correct the positioning deviation of satellite downlink data. The main difficulties in acquiring control points are twofold: a fast and reliable automatic matching method for heterogeneous remote sensing images and the acquisition of reference images (especially high-resolution reference images). In the absence of ground control points, there is a lack of sufficient spatial coordinate references, and the absolute positioning accuracy of satellite downlink data is difficult to guarantee; (3) With the increase of remote sensing satellite data sources and the continuous improvement of image spatial resolution and temporal resolution, the processing scale of satellite downlink data orthorectification has also expanded rapidly. Traditional geometric processing methods lack the constraints of intelligent cognitive mechanisms and have a large amount of redundant information and calculations, which seriously restricts the real-time nature of satellite downlink data services.

[0005] 2. Remote sensing big data intelligent information processing and services face insurmountable limitations. Current intelligent information processing methods are primarily designed for single sensors. For multi-source, high-dimensional, heterogeneous, and complexly correlated Earth observation data, research into novel intelligent processing algorithms and models is needed. Land feature identification and information extraction are crucial intermediate steps in converting remote sensing images from data into information that serves various industries, and are also core technologies in the entire remote sensing application service chain. Existing remote sensing image processing methods are mostly based on pixel-level spectral analysis, lacking a deep understanding of the hierarchical features of land features. This makes it difficult to integrate with domain knowledge, expert experience, and geoscientific models, resulting in low accuracy in information identification and extraction, thus limiting the comprehensive mining and application of remote sensing information. Evolutionary algorithm-based remote sensing image processing methods have low time efficiency, making it difficult to meet the demands of remote sensing big data in actual production.

[0006] On the other hand, in terms of service methods, this traditional satellite operation service model is a passive, order-based service. The handover process is cumbersome, involves many steps, and is time-consuming, severely hindering the speed of remote sensing emergency response to major disasters. Major domestic satellite data service systems include the satellite data service network of the China Remote Sensing Satellite Ground Station, the domestic satellite data service system of the China Resources Satellite Application Center, the Fengyun satellite data service network of the Meteorological Bureau Satellite Center, and the data service network of the Ocean Satellite Application Center. Each system primarily focuses on satellite data querying, ordering, and downloading. Although the ability to acquire remote sensing data is continuously improving, the overall level of domestic remote sensing information product production technology and services is still not high. The production and service of information products remain at the stage of passive order processing, lacking proactive push and real-time service mechanisms. This makes it difficult to meet the requirements of industrial market commercial operations and the actual business needs of public users, and it cannot adapt to the development of remote sensing data acquisition technology and meet the rapidly growing social demand for remote sensing services. Summary of the Invention

[0007] This invention provides a method, system, and electronic device for real-time remote sensing satellite downlink data services. It changes the traditional service mode and content of satellite remote sensing data, solves the technical problem that the handover process of the existing remote sensing satellite operation service mode is cumbersome and lacks accuracy, which affects the real-time service capability of remote sensing data, meets users' needs for rapid and dynamic acquisition of remote sensing data and information, and greatly improves the service efficiency of remote sensing information.

[0008] To achieve the above objectives, according to one aspect of this application, a method for real-time remote sensing satellite downlink data service is provided. Specifically, this method integrates satellite downlink data access, real-time geometric positioning of satellite downlink data, rapid extraction of target information, and proactive remote sensing information service into a unified real-time remote sensing information service. The method includes: accessing received remote sensing satellite downlink data; performing geometric correction on the remote sensing satellite downlink data using a multi-satellite image collaborative uncontrolled geometric positioning strategy to obtain real-time geometric positioning data; inputting the real-time geometric positioning data into a deep learning model for identifying predetermined ground features to obtain thematic information about specific targets; intelligently pushing corresponding image content text description corpora generated based on the real-time geometric positioning data and the thematic information to the user terminal, and optimizing the image cognitive computing model using user feedback to adjust the accuracy of the remote sensing data information service product.

[0009] Optionally, the step of performing geometric correction on the remote sensing satellite downlink data using a multi-satellite image collaborative uncontrolled geometric positioning strategy includes: analyzing the original geometric positioning error of remote sensing images in the historical archive database of the sensors of the remote sensing satellite downlink data; modeling the original geometric positioning error using a harmonic model to obtain an error harmonic model; correcting the original imaging geometric model using the error harmonic model to compensate for the periodic error of the original imaging geometric model, obtaining error-compensated remote sensing satellite downlink data; acquiring low-precision control points from publicly available low-resolution reference images based on the error-compensated remote sensing satellite downlink data; converting the low-precision control points into relaxed constraint conditions and incorporating them into a regional network adjustment model to obtain a regional network adjustment model with relaxed constraint conditions, wherein the regional network adjustment model is calculated based on multiple connection points between multiple views of the remote sensing satellite downlink data area; and performing uncontrolled geometric correction on the error-compensated remote sensing satellite data using the optimized regional network adjustment model to obtain high-precision geometric positioning data.

[0010] Optionally, the process of performing control point-free geometric positioning also includes: transforming the nonlinear optimization problem with inequality constraints into an optimization problem with equality constraints based on a regional network model with relaxation constraints.

[0011] Optionally, the step of inputting real-time geometric positioning data into a deep learning model for identifying predetermined ground features to obtain thematic information about a specific target includes: analyzing the characteristics of predetermined ground features in remote sensing images based on multi-source remote sensing satellite data, selecting typical image patches for annotation, and standardizing the annotation data to obtain an annotated sample dataset of a specific target in the multi-source satellite remote sensing data; constructing an initial deep learning model suitable for the predetermined ground features based on the characteristics of the predetermined ground features in the remote sensing images; training the initial deep learning model using the annotated sample dataset to obtain a deep learning model for identifying predetermined ground features, wherein the training process includes iterative calculations of model forward inference, target loss calculation, model weight update, and model evaluation and selection, ultimately obtaining a deep learning model that meets the requirements of the specific target identification task; and inputting real-time geometric positioning data into the deep learning model for identifying predetermined ground features to obtain thematic information about the specific target.

[0012] Optionally, the step of intelligently pushing the corresponding image content text description corpus generated based on real-time geometric positioning data and the thematic information to the user terminal includes: inputting the real-time geometric positioning data into an image understanding model to obtain image semantic information; extracting keywords based on the image semantic information to obtain an image text description, wherein the image understanding model adopts an encoder-decoder structure; obtaining target location, attribute, and quantity information based on thematic information of a specific target; integrating the image text description, the target location, attribute, and quantity information to obtain the image content description text corpus; analyzing the target user's historical query records and preferences, and using a content recommendation algorithm to push the real-time geometric positioning data and the corresponding image content description text corpus to the user terminal through a communication method set by the user.

[0013] Optionally, the step of optimizing the image cognitive computing model using user feedback information includes: receiving user feedback information after pushing the real-time geometric positioning data and the corresponding image content description text corpus to the user terminal; verifying the user feedback information and incorporating the correct target information into the sample dataset to update the constructed sample library; and using the updated sample library to optimize the image cognitive computing model using a transfer learning strategy to adjust the accuracy of the remote sensing data information service product.

[0014] Optionally, after pushing the real-time geometric positioning data and the corresponding image content description text corpus to the user terminal, the method further includes: through the user terminal's operation page, the user clicks the received feedback link to jump to the optimization module and provides feedback on the error information pushed, wherein the error information includes: false alarms, location, content, and area deviation.

[0015] According to another aspect of the present invention, a real-time remote sensing satellite downlink data service system is also provided. Specifically, this system integrates satellite downlink data access, real-time geometric positioning of satellite downlink data, rapid target information extraction, and proactive remote sensing information services. The system includes: a data access unit for accessing received remote sensing satellite downlink data and acquiring the data; a real-time geometric positioning unit for performing geometric correction on the remote sensing satellite downlink data using a multi-satellite image collaborative uncontrolled geometric positioning strategy to obtain real-time geometric positioning data; a target identification unit for inputting the real-time geometric positioning data into a deep learning model for identifying predetermined ground features and targets to obtain thematic information about specific targets; and a proactive remote sensing information service unit for intelligently pushing corresponding image content text description corpora generated based on the real-time geometric positioning data and the thematic information to the user terminal, and optimizing the image cognitive computing model using user feedback information to adjust the accuracy of the remote sensing data information service product.

[0016] Optionally, the instantaneous geometric positioning unit includes: a raw geometric positioning error analysis module, used to analyze the raw geometric positioning error of remote sensing images in the historical archive database of the remote sensing satellite downlink data based on the sensor; an error modeling module, used to model the raw geometric positioning error using a harmonic model to obtain an error harmonic model; an error compensation module, used to correct the original imaging geometric model using the error harmonic model to compensate for the periodic error of the original imaging geometric model, and obtain error-compensated remote sensing satellite downlink data; a first acquisition module, used to acquire low-precision control points from publicly available low-resolution reference images based on the error-compensated remote sensing satellite downlink data; a regional network adjustment model optimization module, used to transform the low-precision control points into relaxed constraint conditions and incorporate them into the regional network adjustment model to obtain a regional network adjustment model with relaxed constraint conditions, wherein the regional network adjustment model is calculated based on multiple connection points between multiple views of the remote sensing satellite downlink data area; and a geometric correction module, used to perform control point-free geometric correction on the error-compensated remote sensing satellite data using the optimized regional network adjustment model to obtain high-precision geometric positioning data.

[0017] Optionally, the process of performing control point-free geometric positioning also includes: a nonlinear optimization problem transformation module, which is used to transform the nonlinear optimization problem with inequality constraints into an optimization problem with equality constraints based on the regional network model with relaxation constraints.

[0018] Optionally, the target recognition unit includes: a sample annotation module, used to analyze the characteristics of predetermined ground features in remote sensing images based on multi-source remote sensing satellite data, select typical image patches for annotation, and standardize the annotation data to obtain an annotated sample dataset of specific targets in multi-source satellite remote sensing data; an initial deep learning model construction module, used to construct an initial deep learning model suitable for the predetermined ground features based on the characteristics of the predetermined ground features in remote sensing images; a model training module, used to train the initial deep learning model using the annotated sample dataset to obtain a deep learning model for the recognition of predetermined ground features, wherein the training process includes iterative calculations of model forward inference, target loss calculation, model weight update, and model evaluation and selection, ultimately obtaining a deep learning model that meets the requirements of the specific target recognition task; and a target recognition module, used to input real-time geometric positioning data into the deep learning model for the recognition of predetermined ground features to obtain thematic information of the specific target.

[0019] Optionally, the remote sensing information active service unit includes: an image text description generation module, used to input the real-time geometric positioning data into an image understanding model to obtain image semantic information, extract keywords based on the image semantic information, and obtain an image text description, wherein the image understanding model adopts an encoder-decoder structure; a thematic information acquisition module, used to acquire target location, attribute, and quantity information based on thematic information of a specific target; an image content integration module, used to integrate the image text description, the target location, attribute, and quantity information to obtain the image content description text corpus; and an information push module, used to analyze the target user's historical query records and preferences, and use a content recommendation algorithm to push the real-time geometric positioning data and the corresponding image content description text corpus to the user terminal through a user-defined communication method.

[0020] Optionally, the remote sensing information active service unit further includes: a user feedback information receiving module, used to receive user feedback information after the real-time geometric positioning data and the corresponding image content description text corpus are pushed to the user terminal; a sample library update module, used to verify the user feedback information and include the correct target information into the sample dataset to update the constructed sample library; and a cognitive computing model optimization module, used to optimize the image cognitive computing model using a transfer learning strategy with the updated sample library to adjust the accuracy of the remote sensing data information service product.

[0021] Optionally, the remote sensing information active service unit further includes an error information feedback unit, which is used to, after the real-time geometric positioning data and the corresponding image content description text corpus are pushed to the user terminal, allow the user to click on the received feedback link through the user terminal's operation page to jump to the optimization module and provide feedback on the error information pushed, wherein the error information includes: target false alarm, location, content and area deviation.

[0022] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the remote sensing satellite downlink data real-time service method described above.

[0023] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps of the remote sensing satellite downlink data real-time service method described in any one of the above embodiments.

[0024] Based on the above disclosure, this invention employs a multi-satellite image collaborative uncontrolled geometric positioning strategy to perform geometric correction on remote sensing satellite downlink data, obtaining real-time geometric positioning data. This real-time geometric positioning data is then input into a deep learning model for identifying predetermined ground objects, yielding thematic information on specific targets. The corresponding image content text description corpus generated based on the real-time geometric positioning data and thematic information is intelligently pushed to the user terminal, and user feedback is used to optimize the image cognitive computing model and adjust the accuracy of the remote sensing data information service product.

[0025] Compared with existing technologies, the beneficial effects of this invention are as follows: it realizes an integrated real-time remote sensing information service that combines remote sensing satellite downlink data access, real-time geometric positioning of satellite downlink data, rapid extraction of target information, and proactive remote sensing information services, forming a complete chain of "data-information-push and feedback." This changes the traditional "reception-processing-archiving-distribution" service model of satellite remote sensing data and the service content that remains at the remote sensing data level, reducing the processing cost and application difficulty for users of remote sensing images. It solves the problems of cumbersome handover links, numerous steps, long duration, and insufficient accuracy in the traditional satellite operation service model, realizing the transformation from the traditional passive remote sensing data service model to proactive intelligent service, meeting the needs of a wide range of users in the era of remote sensing big data for rapid and dynamic acquisition of remote sensing data and information, and significantly improving the service efficiency of remote sensing information.

[0026] For remote sensing images without control points, geometric positioning without control points can be achieved using a regional network adjustment model with relaxed constraints. This enables high-precision geometric positioning even in the absence of high-precision control points, providing accurate positioning information for disaster emergency responses such as forest fire warnings. It overcomes the limitations of low control point accuracy, ensuring that the corrected images have both good absolute positioning accuracy and image edge-matching accuracy. Furthermore, a target-specific deep learning model is used to quickly extract thematic information about the target, and the corresponding image content text description corpus generated based on real-time geometric positioning data and thematic information is intelligently pushed to the user terminal. This allows for the rapid and proactive provision of accurate and content-rich remote sensing information services to users. Attached Figure Description

[0027] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0028] Figure 1 A hardware structure block diagram of a computer terminal (or mobile device) for implementing a real-time service method for downlink data from remote sensing satellites is shown.

[0029] Figure 2 This is a flowchart of the real-time service method for remote sensing satellite downlink data in an embodiment of the present invention;

[0030] Figure 3 This is a schematic diagram of the real-time service system for remote sensing satellite downlink data in an embodiment of the present invention;

[0031] Figure 4 This is a structural block diagram of an electronic device that performs a real-time service method for remote sensing satellite downlink data in an embodiment of the present invention. Detailed Implementation

[0032] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0033] It should be noted that the terms "first," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0034] It should be noted that the remote sensing satellite downlink data real-time service method, system, and electronic equipment disclosed herein can be used in the field of remote sensing information engineering technology to realize real-time service of remote sensing satellite downlink data based on remote sensing information engineering, and can also be used in any field other than the field of remote sensing information engineering technology to realize real-time service of remote sensing satellite downlink data based on remote sensing information engineering. This disclosure does not limit the application field of the remote sensing satellite downlink data real-time service method, system, and electronic equipment.

[0035] It should be noted that the information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, and displayed data) collected in this public disclosure are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with the relevant laws, regulations, and standards of the relevant regions, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding operation entry points are provided for users to choose to authorize or refuse. For example, this system has interfaces with relevant users or organizations. Before obtaining relevant information, a request to obtain the information needs to be sent to the aforementioned user or organization through the interface, and the relevant information is obtained only after receiving consent from the aforementioned user or organization.

[0036] It should be noted that in this disclosure, customer information is collected and analyzed, and users are provided with corresponding operation entry points to choose whether to agree to or reject the automated decision results; if the user chooses to reject, the process will proceed to the expert decision-making process.

[0037] The following embodiments of the present invention realize an integrated real-time remote sensing information service that combines remote sensing satellite downlink data access, real-time geometric positioning of satellite downlink data, rapid extraction of target information, and proactive remote sensing information services. This significantly shortens the time from data to information to service, meets the needs of a wide range of users in the era of remote sensing big data for rapid and dynamic acquisition of remote sensing data and information, and greatly improves the service efficiency of remote sensing information.

[0038] The following embodiments of the present invention can be applied to various systems / applications / devices for processing remote sensing satellite downlink data. The present invention can utilize historical archived data to compensate for periodic systematic errors, and combine low-precision control points and repeated observation information from multi-source satellites to achieve rapid, automatic, and high-precision geometric positioning of satellite imagery in the absence of high-precision control points. Simultaneously, by utilizing cognitive computing technologies based on visual attention mechanisms and deep learning, combined with geoscientific knowledge constraints, the invention significantly improves the identification efficiency and accuracy of specific targets (such as fire points, urban buildings, water bodies, mining areas, forest cover, and burned areas) in remote sensing imagery. Furthermore, by utilizing remote sensing image recommendation algorithms and a crowdsourcing-based intelligent information feedback mechanism, the invention enables proactive intelligent recommendation of remote sensing data information to users, completing the feedback of user prior knowledge to proactive intelligent services and systems, realizing the verification and optimization of the remote sensing image cognitive computing model, and improving the accuracy of real-time service products.

[0039] It should be noted that the implementation scenarios of this invention focus on the real-time processing and service of remote sensing satellite downlink data, and are particularly suitable for fields such as emergency response, environmental monitoring, and resource development. Specific scenarios include, but are not limited to: Disaster emergency response: In disaster emergencies such as fire prevention and flood monitoring, this invention can quickly push relevant ground target information such as fire points and flood ranges from remote sensing satellite downlink data to monitor and warn of potential disasters; Environmental monitoring and management: For environmental issues such as forest cover changes, water pollution, and illegal mining, this invention can quickly extract relevant ground target information from remote sensing satellite downlink data to monitor environmental changes, warn of potential ecological and environmental problems, and promote precise environmental management and protection; Remote sensing applications: In commercial application fields such as agriculture, forestry, and marine resource monitoring, this invention can provide customized remote sensing information products, such as crop growth monitoring, forest resource change analysis, and marine ecological environment assessment, to meet the personalized needs of users in different industries.

[0040] The present invention will now be described in detail with reference to various embodiments.

[0041] Example 1

[0042] According to an embodiment of the present invention, an embodiment of a method for real-time service of downlink data from remote sensing satellites is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0043] The real-time remote sensing satellite downlink data service method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal or similar computing system. Figure 1A hardware block diagram of a computer terminal (or mobile device) for implementing a real-time service method for remote sensing satellite downlink data is shown. Figure 1 As shown, computer terminal 10 (or mobile device) may include one or more ( Figure 1 The processor 102 (which may include, but is not limited to, a microprocessor MCU (Microcontroller Unit) or a programmable gate array (FPGA)) is shown as 102a, 102b, ..., 102n. It also includes a memory 104 for storing data and a transmission system 106 for communication functions. In addition, it may include: a display, an input / output interface (I / O interface), a Universal Serial Bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the electronic system described above. For example, the computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0044] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0045] The memory 104 can be used to store software programs and modules of application software, such as the program instruction / data storage system corresponding to the real-time remote sensing satellite downlink data service method in this embodiment of the application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned real-time remote sensing satellite downlink data service method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage systems, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0046] The transmission system 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission system 106 includes a network interface controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission system 106 may be a radio frequency (RF) module used for wireless communication with the Internet.

[0047] The display can be, for example, a touchscreen liquid crystal display (LCD), which allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0048] Under the aforementioned operating environment, this application provides a method for real-time remote sensing satellite downlink data service, specifically establishing an integrated real-time remote sensing information service that combines satellite downlink data access, real-time geometric positioning of satellite downlink data, rapid extraction of target information, and proactive remote sensing information service. Figure 2 This is a flowchart of a real-time remote sensing satellite downlink data service method according to an embodiment of the present invention, such as... Figure 2 As shown, the method includes the following steps S201 to S204, and each step is described in detail below.

[0049] Step S201: Access the received remote sensing satellite downlink data and acquire the remote sensing satellite downlink data.

[0050] Step S202: The acquired remote sensing satellite downlink data is geometrically corrected using a multi-satellite image collaborative uncontrolled geometric positioning strategy to obtain real-time geometric positioning data.

[0051] Step S203: Input the real-time geometric positioning data into the deep learning model for identifying the predetermined ground object to obtain thematic information of the specific target. The deep learning model for identifying the predetermined ground object is pre-trained based on the labeled sample dataset of the specific target from multi-source satellite remote sensing data.

[0052] Step S204 involves intelligently pushing the corresponding image content text description corpus generated based on real-time geometric positioning data and target thematic information to the user terminal, and using user feedback information to optimize the image cognitive computing model and adjust the accuracy of remote sensing data information service products.

[0053] In this embodiment, step S202 specifically includes: analyzing the original geometric positioning error of remote sensing images in the historical archive database of the sensor based on the sensor of the remote sensing satellite downlink data; modeling the original geometric positioning error using a harmonic model to obtain an error harmonic model; correcting the original imaging geometric model using the error harmonic model to compensate for the periodic error of the original imaging geometric model, thereby obtaining error-compensated remote sensing satellite downlink data; collecting low-precision control points from publicly available low-resolution reference images based on the error-compensated remote sensing satellite downlink data; converting the low-precision control points into relaxed constraint conditions and incorporating them into the regional network adjustment model to obtain a regional network adjustment model with relaxed constraint conditions, wherein the regional network adjustment model is calculated based on the multiple connection points between multiple views of the remote sensing satellite downlink data area; and performing control point-free geometric correction on the error-compensated remote sensing satellite data using the optimized regional network adjustment model to obtain high-precision geometric positioning data.

[0054] Taking the domestically produced Gaofen-1 B / C / D (GF1-B / C / D) satellite as an example, this study statistically analyzes the geometric positioning errors of long-term GF1-B / C / D satellite multispectral downlink data (15,822 scenes from 2019 to 2022) from the Scientific Data Center of the Remote Sensing Satellite Ground Station. The analysis reveals that the magnitude and direction of positioning errors within a single orbit are relatively consistent, while positioning errors in images from different orbits vary significantly. However, positioning errors from images from the same sensor exhibit certain underlying patterns over time. This suggests that existing imaging geometry theories lack time-dependent factors, and targeted improvements can reduce the uncertainty of imaging geometric models.

[0055] The harmonic model is used to model the imaging geometric model error of the sensor, as shown in formula (1):

[0056]

[0057] ... (1)

[0058] In equation (1), and These represent the errors of the imaging geometry model in the X and Y directions, respectively. For the imaging date, The harmonic model period is 365.24 days. , , For the harmonic model parameters in the X direction, , , These are the harmonic model parameters in the Y direction.

[0059] Subsequently, the original imaging geometric model is corrected based on the obtained error harmonic model to reduce the positioning error. In this embodiment, the average positioning error in the X and Y directions can be reduced from 8 pixels (about 64 meters) and 5 pixels (about 40 meters) to 3 pixels (about 24 meters) and 2 pixels (about 16 meters) respectively, thereby significantly improving the geometric positioning accuracy of satellite downlink data.

[0060] In response to the insufficient accuracy of autonomous geometric positioning of downlink data from domestic satellites and the difficulty in obtaining high-precision, high-resolution reference images, low-precision control points extracted from publicly available low-resolution reference images (such as Landsat / Sentinel-2) are incorporated into the regional network model to increase ground constraints, thereby achieving high-precision geometric positioning of domestic meter-level resolution optical satellites under the condition of lacking high-precision control data.

[0061] Based on the error-compensated remote sensing satellite downlink data, corresponding publicly available lower-resolution satellite remote sensing images are acquired. Low-precision control points are collected from the lower-resolution reference images. On the basis of the traditional regional network adjustment model, the control points collected from the low-resolution reference images are used as relaxation constraints. The imaging model of the satellite images is optimized by utilizing multiple connection points between multi-view images. This can overcome the limitation of low control point accuracy, so that the corrected images have both good absolute positioning accuracy and image edge matching accuracy. It has strong practicality and realizes efficient and reliable multi-view matching of multi-source satellite image sequences.

[0062] The regional network adjustment model with relaxed constraints is shown in equation (2):

[0063] ...(2)

[0064] in,

[0065] ;

[0066] ;

[0067] Represents the object space coordinates of the connection point. Indicates the image plane coordinates of the connection points; Represents the object space coordinates of the control points. Represents the image plane coordinates of the control points; The geometric model optimization parameters for image n are represented. This indicates the precision of the k-th control point; Indicates the first The ground resolution of the landscape image.

[0068] In the process of optimizing the imaging model of satellite images by utilizing multiple connection points between multiple views in the regional network adjustment model with relaxed constraints, this implementation case transforms the nonlinear optimization problem with inequality constraints into an optimization problem with lighting constraints based on the Karush-Kuhn-Tucker (KKT) conditions, as expressed in formula (3):

[0069] ...(3)

[0070] in, and , .

[0071] The KKT conditions are a set of necessary conditions for solving constrained optimization problems. By introducing Lagrange multipliers, the original problem is transformed into a more easily solvable form. In this embodiment, inequality constraints are transformed into equality constraints, allowing for more flexible use of prior accuracy information of control points during the solution process. The goal is to minimize the overall positioning error while ensuring the effectiveness of the control point constraints. This transformation strategy not only improves positioning accuracy but also maintains the robustness and computational efficiency of the model, providing strong technical support for real-time geometric positioning of remote sensing images.

[0072] In this embodiment, step S203 specifically includes:

[0073] Based on multi-source remote sensing satellite data, considering the characteristics of the predetermined ground features in the remote sensing images, representative and typical image patches are selected for annotation, and the annotated data is standardized to obtain an annotated sample dataset of specific targets in multi-source satellite remote sensing data, which is then divided into training dataset and test dataset.

[0074] Based on the characteristics of the predetermined ground features in remote sensing images, a deep learning model suitable for the specific ground features is constructed, including network architecture, activation function, loss function and optimizer. The constructed deep learning model can be a single neural network model or an integration of multiple neural network models, or it can be combined with visual attention mechanism and geoscience knowledge according to the actual situation.

[0075] The deep learning model is trained using the above-mentioned labeled sample dataset, including the iterative calculation process of model forward inference, target loss calculation, model weight update, and model evaluation and selection. Finally, a deep learning model with strong generalization ability, high recognition accuracy, and meeting the requirements of specific target recognition tasks is obtained. In the iterative calculation process, data augmentation, dataset filtering, dynamic network architecture adjustment, multiple optimization objectives, and multi-stage training techniques can also be used to further improve the model's generalization ability and recognition accuracy.

[0076] Real-time geometric positioning data is input into a deep learning model for identifying predetermined ground features to obtain thematic information on specific targets.

[0077] Taking the identification of specific target information of forest fireworks as an example, this embodiment provides a method for identifying forest fireworks information based on a context-aware YOLO optimization model. The specific implementation steps are as follows:

[0078] First, a typical forest fire annotation sample dataset is constructed. Positive samples need to consider cases with only smoke and cases with both smoke and fire points. Negative samples need to collect a large number of hard negative samples that are easily confused with fire, including clouds, highly reflective surfaces (such as white building roofs, beaches and salt fields), water vapor from industrial emissions, mountain shadows, and hot surfaces (such as deserts). After annotation, image standardization, as well as Mosaic augmentation, random rotation, scaling, and spectral enhancement, are required to increase data diversity and prevent overfitting. A "hard sample library" is also constructed, mainly consisting of hard negative samples, which are then fed to the model in subsequent training iterations.

[0079] A context-aware YOLO optimization model is constructed based on the classic YOLO architecture, introducing context awareness and attention mechanisms to address the challenges of detecting forest fires in satellite remote sensing imagery, such as small targets, varied shapes, complex backgrounds, and susceptibility to confusion with clouds, fog, and highly reflective ground features. A Convolutional Block Attention Module (CBAM) is inserted into the key stages of the backbone network. Its channel attention strengthens the weights of channels more sensitive to fire, while spatial attention learns the diffusion shape of smoke and the location features of open flames. In the neck network, in addition to fusing its own multi-scale features, a multi-branch context extraction module is employed. This allows the model to acquire features from different receptive fields without sacrificing resolution, effectively enabling the model to simultaneously see the surrounding contextual information of the target (vegetation and bare ground around the fire). Due to the scarcity of fire targets and the extreme imbalance between positive and negative samples, the Focal Loss loss function is used instead of the standard binary cross-entropy loss. By reducing the weight of simple negative samples, the model focuses more on learning hard negative and positive samples.

[0080] The forest fireworks labeled sample dataset was input into the constructed YOLO optimization model for training and tuning. This included: initializing with pre-trained backbone weights on a large dataset, which can accelerate convergence and improve performance; adopting the OneCycleLR adaptive scheduling strategy; using a hyperparameter optimization library for automated search and enabling the model exponentially weighted average (EMA). EMA maintains the historical average of the model weights, which can make the training process smoother and usually improve the robustness and accuracy of the final model; at the same time, the false positive rate of the model on these samples was closely monitored during training, and the constructed "hard sample library" was input into the model during training iterations to improve the model's accuracy.

[0081] By inputting real-time geometric positioning data into the above training, a forest fire detection model is obtained, resulting in a thematic information map of forest fires.

[0082] In this embodiment, step S204 specifically includes:

[0083] Real-time geometric positioning data is input into an image understanding model to obtain image semantic information. Keywords are extracted based on the image semantic information to obtain textual descriptions of the images. The image understanding model adopts an encoder-decoder structure. Location, attribute, and quantity information of targets are obtained based on specific target-themed information products. The obtained textual descriptions and target location, attribute, and quantity information are integrated to obtain image content description text corpus. Historical query records and preferences of target users are analyzed, and a content recommendation algorithm is used to push real-time geometric positioning data and corresponding image content description text corpus to users who may be interested through user-defined communication methods. After receiving the recommended remote sensing data and corresponding image content description text corpus, users provide feedback in the system's optimization module based on the actual situation. The specific steps of the feedback include: the user clicks the received feedback link to jump to the optimization module and provides feedback on the error information pushed, including false alarms, location, content, and area deviations; the information provided by the user is verified, and the correct target information is included in the sample dataset, updating the constructed sample library.

[0084] By using the updated samples, the specific target information extraction model and image understanding model are fine-tuned and optimized using transfer learning methods, thereby improving the accuracy of the corresponding cognitive computing models and ultimately enhancing the accuracy of remote sensing data information service products.

[0085] Taking the forest fire information identified above as an example, this embodiment uses a remote sensing image recommendation algorithm based on content information extraction to push the information to users who may be interested. Then, it uses a crowdsourcing-based information feedback method to optimize the corresponding cognitive computing model, thereby realizing the proactive service of remote sensing information.

[0086] The remote sensing image recommendation algorithm based on content information extraction is used to push information to users who may be interested. The specific steps are as follows:

[0087] First, geometric positioning data is input into a pre-trained high-level image description generation model to infer and generate a text description of the image content, such as "This is a satellite remote sensing image of Baizi Village, Yajiang County, Sichuan Province, taken by Gaofen-2 on March 16, 2024. It is mainly forested, with some scattered farmland and residential areas around it." The pre-trained high-level image description generation model is trained and optimized using a large dataset of "image-description" pairs from multiple source remote sensing satellites. In this embodiment, the encoder of the high-level image description generation model uses Vision Transformer as the backbone network, which can better capture the global contextual information of the image and is responsible for encoding the input image into a feature representation rich in semantic information. Its decoder uses Transformer Decoder, which uses a self-attention mechanism and an encoder-decoder attention mechanism. When generating each word, it can dynamically "look back" at the image features extracted by the encoder to determine which parts of the image are most relevant in the current generation step.

[0088] Based on the aforementioned forest fireworks information products, information on the location, attributes, and number of forest fireworks is obtained, such as "the fireworks point is located at 30.1918°N, 101.1368°E, and there is one fireworks point."

[0089] By integrating the obtained image content text descriptions and information on the location and number of forest fires, an image content description text corpus is obtained, such as "This is a satellite remote sensing image of Baizi Village, Yajiang County, Sichuan Province, taken by Gaofen-2 on March 16, 2024. It is mainly forested, with some scattered farmland and residential areas around it. A suspected fire spot was found at 30.1918°N, 101.1368°E."

[0090] Analyze the target users' historical query records and preferences, as well as relevant information about their organizations, to identify potential target users who may be interested in this topic;

[0091] Then, the obtained image content description text corpus and geometrically positioned remote sensing images are pushed together to potentially interested users (such as relevant users in forestry management departments) through the communication method set by the user.

[0092] Subsequently, the corresponding cognitive computing model is optimized using a crowdsourcing-based information feedback method. Crowdsourcing, as a new distributed production service model, can divide massive data tasks into sub-tasks and integrate dispersed human resources through the internet to achieve efficient and accurate data mining and analysis. On the one hand, it fully utilizes remote sensing data and annotation user resources to achieve rapid task decomposition, thereby quickly achieving the annotation of large batches of samples and the feedback verification of thematic information. On the other hand, the sample verification library and information update library built based on the crowdsourcing feedback mechanism can be fed back into the established cognitive model, thereby optimizing the model parameters and further improving the accuracy of the computing model. The specific steps are as follows:

[0093] After receiving the recommended remote sensing data and corresponding image content description text, the user clicks the received feedback link to jump to the optimization module and provides feedback on the error information in the push (such as the user reporting that a fire point information was missed).

[0094] Then, the information provided by the users is verified, and the correct target information is included in the sample dataset to update the constructed sample library.

[0095] The forest fire detection model was fine-tuned and optimized using transfer learning with the updated samples, which further improved the accuracy of the forest fire detection model and thus enhanced the accuracy of remote sensing data information service products.

[0096] The aforementioned proactive push service using remote sensing satellite downlink data to provide forest fire information can provide data and information support for forest fire prevention early warning and emergency response. Through the above steps, a harmonic model is used to compensate for the periodic systematic errors of remote sensing images, and a regional network adjustment model with relaxed constraints is used for control point-free geometric positioning. This enables high-precision geometric positioning even in the absence of high-precision control points, providing accurate positioning information for disaster emergency responses such as forest fire warnings. By using a pre-trained deep learning model for specific target features to identify specific target features, the efficiency and accuracy of data conversion into information can be improved. Finally, a recommendation algorithm is used to proactively push data and text description corpora to users, enhancing the real-time service capability of remote sensing data. User feedback information can be used to continuously improve the accuracy and precision of the model and services. Unlike the traditional data reception-processing-archiving-distribution technology process, this implementation case establishes an integrated real-time remote sensing information service that combines satellite downlink data access, real-time geometric positioning of downlink data, rapid extraction of target information, and proactive remote sensing information services. This changes the traditional service model and content of satellite remote sensing data, solves the technical problems of cumbersome handover links and insufficient accuracy in the existing remote sensing satellite operation service model, meets users' needs for rapid and dynamic acquisition of remote sensing data and information, and significantly improves the service efficiency of remote sensing information.

[0097] The following is a detailed description with reference to another embodiment.

[0098] Example 2

[0099] The remote sensing satellite downlink data real-time service system provided in this embodiment includes multiple implementation units, each of which corresponds to a specific implementation step in the above embodiment one. The specific implementation method and beneficial effects can be referred to the aforementioned method embodiment, and will not be repeated here.

[0100] This embodiment provides a real-time remote sensing satellite downlink data service system, specifically establishing an integrated real-time remote sensing information service that combines satellite downlink data access, real-time geometric positioning of satellite downlink data, rapid extraction of target information, and proactive remote sensing information services.

[0101] Figure 3 This is a schematic diagram of an optional real-time remote sensing satellite downlink data service system according to an embodiment of the present invention, such as... Figure 3 As shown, the remote sensing satellite downlink data real-time service system may include: a data access unit 31, a real-time geometric positioning unit 32, a target identification unit 33, and a remote sensing information active service unit 34.

[0102] Among them, the data access unit 31 is used to access the received remote sensing satellite downlink data and acquire remote sensing satellite downlink data.

[0103] The instantaneous geometric positioning unit 32 is used to perform geometric correction on remote sensing satellite downlink data using a multi-satellite image collaborative uncontrolled geometric positioning strategy to obtain instantaneous geometric positioning data.

[0104] The target recognition unit 33 is used to input real-time geometric positioning data into a deep learning model for the recognition of predetermined ground features and targets to obtain thematic information on specific targets.

[0105] The remote sensing information active service unit 34 is used to intelligently push the corresponding image content text description corpus generated based on real-time geometric positioning data and thematic information to the user terminal, and to optimize the image cognitive computing model and adjust the accuracy of remote sensing data information service products by using user feedback information.

[0106] The aforementioned real-time remote sensing satellite downlink data service system can access the received remote sensing satellite downlink data through the data access unit 31 to obtain the downlink data. The real-time geometric positioning unit 32 uses a multi-satellite image collaborative uncontrolled geometric positioning strategy to perform geometric correction on the downlink data to obtain real-time geometric positioning data. The target recognition unit 33 fuses the geometric positioning data from multiple remote sensing satellites and uses a specific target deep learning model to identify predetermined ground objects to obtain target information. Finally, the remote sensing information active service unit 34 intelligently pushes the corresponding image content text description corpus generated based on the real-time geometric positioning data and thematic information to the user terminal, and uses user feedback information to optimize the image cognitive calculation model and adjust the accuracy of the remote sensing data information service product.

[0107] In this embodiment, for remote sensing images without control points, a regional network model with relaxed constraints can be used for geometric positioning without control points. This overcomes the limitation of low accuracy of control points, enabling the corrected images to have both good absolute positioning accuracy and image edge-matching accuracy. Then, a target-specific deep learning model based on visual attention mechanisms and deep learning strategies is used to quickly extract target thematic information. Finally, a remote sensing image content recommendation algorithm is used to intelligently push the corresponding image content text description corpus generated based on real-time geometric positioning data and thematic information to the user terminal. This can quickly and proactively provide users with accurate and content-rich remote sensing information services, improve the real-time service capability of remote sensing data, and solve the problems of cumbersome handover links, multiple steps, long duration, and insufficient accuracy in traditional satellite operation service modes. This meets users' needs for rapid and dynamic acquisition of remote sensing data and information, and significantly improves the service efficiency of remote sensing information.

[0108] Optionally, the instantaneous geometric positioning unit includes: a raw geometric positioning error analysis module, used to analyze the raw geometric positioning error of remote sensing images in the historical archive database of the sensor based on the sensor of the remote sensing satellite downlink data; an error modeling module, used to model the raw geometric positioning error using a harmonic model to obtain an error harmonic model; an error compensation module, used to correct the original imaging geometric model using the error harmonic model to compensate for the periodic error of the original imaging geometric model, and obtain error-compensated remote sensing satellite downlink data; a first acquisition module, used to acquire low-precision control points from publicly available low-resolution reference images based on the error-compensated remote sensing satellite downlink data; a regional network adjustment model optimization module, used to transform the low-precision control points into relaxed constraint conditions and incorporate them into the regional network adjustment model to obtain a regional network adjustment model with relaxed constraint conditions, wherein the regional network adjustment model is calculated based on the multiple connection points between multiple views of the remote sensing satellite downlink data area; and a geometric correction module, used to perform control point-free geometric correction on the error-compensated remote sensing satellite data using the optimized regional network adjustment model to obtain high-precision geometric positioning data.

[0109] Optionally, the process of performing control point-free geometric positioning also includes: a nonlinear optimization problem transformation module, which is used to transform the nonlinear optimization problem with inequality constraints into an optimization problem with equality constraints based on the regional network model with relaxation constraints.

[0110] Optionally, the target recognition unit includes: a sample annotation module, used to analyze the characteristics of predetermined ground features in remote sensing images based on multi-source remote sensing satellite data, select typical image patches for annotation, and standardize the annotation data to obtain an annotated sample dataset of specific targets in multi-source satellite remote sensing data; an initial deep learning model construction module, used to construct an initial deep learning model suitable for the predetermined ground features based on the characteristics of the predetermined ground features in remote sensing images; a model training module, used to train the initial deep learning model using the annotated sample dataset to obtain a deep learning model for the recognition of predetermined ground features, wherein the training process includes iterative calculations of model forward inference, target loss calculation, model weight update, and model evaluation and selection, ultimately obtaining a deep learning model that meets the requirements of the specific target recognition task; and a target recognition module, used to input real-time geometric positioning data into the deep learning model for the recognition of predetermined ground features to obtain thematic information of the specific target.

[0111] Optionally, the remote sensing information active service unit includes: an image text description generation module, used to input real-time geometric positioning data into an image understanding model, obtain image semantic information, extract keywords based on the image semantic information, and obtain image text descriptions, wherein the image understanding model adopts an encoder-decoder structure; a thematic information acquisition module, used to obtain target location, attribute, and quantity information based on thematic information of a specific target; an image content integration module, used to integrate image text descriptions, target location, attribute, and quantity information to obtain image content description text corpus; and an information push module, used to analyze the target user's historical query records and preferences, and use a content recommendation algorithm to push real-time geometric positioning data and corresponding image content description text corpus to the user terminal through a user-defined communication method.

[0112] Optionally, the remote sensing information active service unit further includes: a user feedback information receiving module, used to receive user feedback information after pushing real-time geometric positioning data and corresponding image content description text corpus to the user terminal; a sample library update module, used to verify the user feedback information and include the correct target information into the sample dataset to update the constructed sample library; and a cognitive computing model optimization module, used to optimize the image cognitive computing model using a transfer learning strategy with the updated sample library to adjust the accuracy of the remote sensing data information service product.

[0113] Optionally, the remote sensing information active service unit further includes an error information feedback unit, which is used to allow users to click on the received feedback link to jump to the optimization module through the user terminal's operation page after pushing real-time geometric positioning data and corresponding image content description text corpus to provide feedback on the error information pushed in the push. The error information includes: false alarms of targets, location, content and area deviation.

[0114] The aforementioned remote sensing satellite downlink data real-time service system may also include a processor and a memory. The aforementioned data access unit 31, real-time geometric positioning unit 32, target identification unit 33, remote sensing information active service unit 34, etc., are all stored in the memory as program units, and the processor executes the aforementioned program units stored in the memory to realize the corresponding functions.

[0115] The aforementioned processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured, and the processing of remote sensing satellite downlink data can be achieved by adjusting kernel parameters.

[0116] The aforementioned memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0117] Example 3

[0118] Embodiments of this application may provide an electronic device. Figure 4 This is a structural block diagram of an electronic device for performing a real-time service method for remote sensing satellite downlink data according to an embodiment of this application. Figure 4 As shown, the electronic device may include: one or more ( Figure 4 (Only one is shown) Processor 402, memory 404, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0119] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the remote sensing satellite downlink data real-time service method and system in this application embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the aforementioned remote sensing satellite downlink data real-time service method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage systems, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0120] The processor can access the information and application programs stored in the memory through the transmission system to perform the following steps: access the received remote sensing satellite downlink data to acquire the remote sensing satellite downlink data; perform geometric correction on the remote sensing satellite downlink data using a multi-satellite image collaborative uncontrolled geometric positioning strategy to obtain real-time geometric positioning data; input the real-time geometric positioning data into a deep learning model for identifying predetermined ground objects to obtain thematic information on specific targets; intelligently push the corresponding image content text description corpus generated based on the real-time geometric positioning data and thematic information to the user terminal, and optimize the image cognitive computing model and adjust the accuracy of the remote sensing data information service product using user feedback information.

[0121] Those skilled in the art will understand that Figure 4 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, handheld computers, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 4 This does not limit the structure of the aforementioned electronic system. For example, electronic devices may also include components that are more... Figure 4 The more or fewer components shown (such as network interfaces, display systems, etc.), or having the same Figure 4 The different configurations shown.

[0122] Those skilled in the art will understand that all or part of the steps in the various remote sensing satellite downlink data real-time service methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0123] Example 4

[0124] Embodiments of this application also provide a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the real-time remote sensing satellite downlink data service method provided in Embodiment 1.

[0125] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute the remote sensing satellite downlink data real-time service method of any one of the above embodiments.

[0126] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0127] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the real-time remote sensing satellite downlink data service method described in various embodiments of this application.

[0128] This application also provides a computer program product, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the real-time remote sensing satellite downlink data service method described in various embodiments of this application.

[0129] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0130] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0131] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The system embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.

[0132] 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 units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0133] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0134] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a 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, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0135] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for real-time service of remote sensing satellite downlink data, characterized in that, Specifically, it establishes an integrated real-time remote sensing information service encompassing satellite downlink data access, real-time geometric positioning of satellite downlink data, rapid target information extraction, and proactive remote sensing information services, including: Access the received remote sensing satellite downlink data to obtain remote sensing satellite downlink data; A multi-satellite image collaborative uncontrolled geometric positioning strategy is used to perform geometric correction on the downlink data of the remote sensing satellites to obtain real-time geometric positioning data; The real-time geometric positioning data is input into a deep learning model for identifying predetermined ground features to obtain thematic information about specific targets. The corresponding image content text description corpus generated based on the real-time geometric positioning data and the thematic information is intelligently pushed to the user terminal, and the image cognitive computing model is optimized by using user feedback information to adjust the accuracy of remote sensing data information service products.

2. The method according to claim 1, characterized in that, The steps for geometrically correcting the downlink data from the remote sensing satellites using a multi-satellite image collaborative uncontrolled geometric positioning strategy include: Based on the sensor of the downlink data from the remote sensing satellite, analyze the original geometric positioning error of the remote sensing images in the historical archive database of the sensor; The original geometric positioning error is modeled using a harmonic model to obtain the error harmonic model; The original imaging geometry model is corrected using the aforementioned error harmonic model to compensate for the periodic error of the original imaging geometry model, thereby obtaining error-compensated remote sensing satellite downlink data. Low-precision control points are collected from publicly available low-resolution reference images based on error-compensated remote sensing satellite downlink data. The low-precision control points are transformed into relaxed constraints and incorporated into the regional network adjustment model to obtain a regional network adjustment model with relaxed constraints. The regional network adjustment model is calculated based on the multiple connection points between multiple views of the remote sensing satellite downlink data area. The optimized regional network adjustment model is used to perform control point-free geometric correction on the error-compensated remote sensing satellite data to obtain high-precision geometric positioning data.

3. The method according to claim 2, characterized in that, The process of performing control-point-free geometric positioning also includes: Based on the regional network model with relaxed constraints, the nonlinear optimization problem with inequality constraints is transformed into an optimization problem with equality constraints.

4. The method according to claim 1, characterized in that, The steps of inputting real-time geometric positioning data into a deep learning model for identifying predetermined ground features to obtain thematic information about specific targets include: Based on multi-source remote sensing satellite data, the characteristics of predetermined ground features in remote sensing images are analyzed, typical image patches are selected for annotation, and the annotation data is standardized to obtain an annotation sample dataset of specific targets in multi-source satellite remote sensing data. Based on the characteristics of the predetermined ground features in the remote sensing image, an initial deep learning model suitable for the predetermined ground features is constructed. The labeled sample dataset is used to train an initial deep learning model to obtain a deep learning model for identifying a predetermined ground object. The training process includes iterative calculations of model forward inference, target loss calculation, model weight update, and model evaluation and selection, ultimately obtaining a deep learning model that meets the requirements of a specific target recognition task. Real-time geometric positioning data is input into a deep learning model for identifying predetermined ground features to obtain thematic information about specific targets.

5. The method according to claim 1, characterized in that, The steps of intelligently pushing corresponding image content text description corpus generated based on real-time geometric positioning data and the thematic information to the user terminal include: The instantaneous geometric positioning data is input into the image understanding model to obtain image semantic information. Keywords are extracted based on the image semantic information to obtain image text description. The image understanding model adopts an encoder-decoder structure. Thematic information on specific objectives is used to obtain target location, attributes, and quantity information; By integrating the image text description, the target location, attribute, and number information, the image content description text corpus is obtained. By analyzing the target user's historical query records and preferences, a content recommendation algorithm is used to push the real-time geometric positioning data and corresponding image content description text corpus to the user's terminal through the communication method set by the user.

6. The method according to claim 1, characterized in that, The steps for optimizing the image cognitive computing model using user feedback include: After pushing the real-time geometric positioning data and the corresponding image content description text corpus to the user terminal, user feedback information is received. Verify user feedback information, incorporate correct target information into the sample dataset, and update the constructed sample library; The image cognitive computing model is optimized using a transfer learning strategy with the updated sample library to adjust the accuracy of remote sensing data information service products.

7. The method according to claim 6, characterized in that, After pushing the real-time geometric positioning data and the corresponding image content description text corpus to the user terminal, the method further includes: Through the user interface, users can click on the received feedback link to jump to the optimization module and provide feedback on the error information in the push notification. The error information includes: false alarms, location, content, and area deviation.

8. A real-time remote sensing satellite downlink data service system, characterized in that, Specifically, it establishes an integrated real-time remote sensing information service encompassing satellite downlink data access, real-time geometric positioning of satellite downlink data, rapid target information extraction, and proactive remote sensing information services, including: The data access unit is used to access the received remote sensing satellite downlink data and acquire remote sensing satellite downlink data; The instantaneous geometric positioning unit is used to perform geometric correction on the remote sensing satellite downlink data using a multi-satellite image collaborative uncontrolled geometric positioning strategy to obtain instantaneous geometric positioning data. The target recognition unit is used to input the real-time geometric positioning data into a deep learning model for identifying predetermined ground features and targets to obtain thematic information about specific targets. The remote sensing information active service unit is used to intelligently push the corresponding image content text description corpus generated based on the real-time geometric positioning data and the thematic information to the user terminal, and to optimize the image cognitive computing model and adjust the accuracy of the remote sensing data information service product by using user feedback information.

9. An electronic device, characterized in that, It includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the real-time remote sensing satellite downlink data service method according to any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the real-time remote sensing satellite downlink data service method according to any one of claims 1 to 7.

Images