A remote sensing image processing method, device and storage medium

Abstract

The application discloses a remote sensing image processing method and device and a storage medium, relates to the field of remote sensing data processing, and aims at solving at least one problem of high bandwidth requirement of a satellite-ground link, high satellite computing resource occupancy rate and poor data processing real-time performance. The application receives a remote sensing data processing task at a ground station, schedules a satellite for the remote sensing data processing task, cooperatively completes the remote sensing data processing task by using the satellite and the ground station, optimizes a first processing model used by the satellite for processing the remote sensing data processing task at the ground station, and uploads at least the optimized model parameters to the satellite. The application reduces the requirement of the satellite-ground link bandwidth, improves the data processing real-time performance, and improves the satellite computing efficiency and service life.

Description

Technical Field

[0001] This invention relates to the field of remote sensing data processing, and in particular to a remote sensing image processing method, apparatus, and storage medium. Background Technology

[0002] Remote sensing data is information about the Earth's surface or atmosphere collected by long-range remote sensors. It reflects various physical and chemical properties of the Earth's surface and has wide applications in many fields such as environmental monitoring, disaster emergency response, agriculture, and urban planning.

[0003] Compared to traditional manual analysis methods, using technologies such as artificial intelligence can significantly improve the efficiency of remote sensing data processing (such as target identification and classification).

[0004] Currently, there are roughly two methods for using artificial intelligence to process remote sensing data: one is to collect remote sensing data by satellite and then transmit it to ground stations for centralized processing; the other is to utilize the data processing resources provided by the satellite to process the remote sensing data at the satellite end and then transmit the processing results to the ground station. For the former, due to the massive daily increase in remote sensing data, the downlink of remote sensing data places extremely high demands on bandwidth. Limited by current channel conditions, it is difficult to transmit all remote sensing data in real time, resulting in severe data backlog, increased processing delays at ground stations, and low real-time performance. For the latter, limited by the size and heat dissipation conditions of the satellite, the computing power of the processors on board is limited. When faced with complex remote sensing data processing tasks, they require a long time, resulting in low processing efficiency, poor real-time performance, and limited processing accuracy. Furthermore, prolonged data processing exacerbates the consumption of limited energy and causes rapid temperature rise, affecting the satellite's energy storage and lifespan. Summary of the Invention

[0005] The purpose of this invention is to provide a ground station, satellite, and satellite-ground coordination method and system to address all or part of the problems mentioned above, thereby solving at least one of the following problems: high requirements for satellite-ground link bandwidth, high satellite computing resource utilization, and poor real-time performance of data processing.

[0006] The technical solution adopted in this invention is as follows:

[0007] A remote sensing image processing method applied to a ground station, the method comprising:

[0008] Receiving and processing remote sensing data;

[0009] The remote sensing data processing task schedules satellites to cooperate in processing the remote sensing image data acquired by the satellites, including: receiving first remote sensing image data transmitted down from the satellites, and processing the processing results retained by the remote sensing data processing task, wherein the first remote sensing image data only includes remote sensing image data for which the satellites have not transmitted processing results; and processing the first remote sensing image data according to the remote sensing data processing task.

[0010] Optimize the first processing model used by the satellite when processing the remote sensing data processing task;

[0011] The optimized model parameters should be uploaded to at least the satellite.

[0012] Secondly, this application also provides another remote sensing image processing method applied to satellites, the method comprising:

[0013] Receive scheduling instructions;

[0014] According to the remote sensing data processing task indicated by the scheduling instruction, the first processing model is used in cooperation with the ground station to complete the processing of the acquired remote sensing image data, including: transmitting the first remote sensing image data and the processing results retained by the remote sensing data processing task to the ground station. The first remote sensing image data only includes remote sensing image data for which the processing results have not been transmitted.

[0015] Receive the model parameters after the ground station optimizes the first processing model.

[0016] In a third aspect, this application also provides another remote sensing image processing method, which includes:

[0017] Receiving and processing remote sensing data;

[0018] Based on the remote sensing data processing task, the ground station is controlled to coordinate with satellites to process the remote sensing image data acquired by the satellites. This includes: controlling the ground station to receive the first remote sensing image data transmitted by the satellites, and processing the processing results retained by the remote sensing data processing task. The first remote sensing image data only includes remote sensing image data for which the satellites have not transmitted processing results. The ground station is then controlled to process the first remote sensing image data according to the remote sensing data processing task.

[0019] The ground station is controlled to optimize the first processing model used by the satellite when processing the remote sensing data;

[0020] The optimized model parameters should be uploaded to at least the satellite.

[0021] Based on the concept of this application, this application also provides a computer-readable storage medium storing computer instructions, which, when executed, can perform the aforementioned remote sensing image processing method.

[0022] In addition, this application also provides a remote sensing image processing device for use at a ground station, the device comprising:

[0023] The information management module is used to receive remote sensing data processing tasks;

[0024] The task scheduling module is used to coordinate with satellites to process remote sensing image data acquired by the satellites based on the remote sensing data processing task. This includes: instructing the scheduled satellites to transmit first remote sensing image data, and processing the processing results retained by the remote sensing data processing task. The first remote sensing image data only includes remote sensing image data for which the satellites have not transmitted processing results.

[0025] The ground-based algorithm module is used to process the first remote sensing image data according to the remote sensing data processing task, and to optimize the first processing model used by the satellite when processing the remote sensing data processing task.

[0026] The ground communication module is used to receive the first remote sensing image data and to upload the optimized model parameters to the satellite.

[0027] Furthermore, this application also provides another remote sensing image processing apparatus for satellites, the apparatus comprising:

[0028] The data acquisition module is used to acquire remote sensing image data;

[0029] The satellite-side algorithm module is used to process the remote sensing image data in cooperation with the ground station using a first processing model according to the remote sensing data processing task indicated by the scheduling instruction. This includes: determining the processing results to be retained in the remote sensing data processing task and the first remote sensing image data, which only includes remote sensing image data for which the processing results are not retained.

[0030] The spaceborne communication module is used to receive the scheduling instructions and the model parameters after the ground station optimizes the first processing model, and is also used to transmit the first remote sensing image data and the retained processing results to the ground station that sent the scheduling instructions.

[0031] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0032] This application utilizes satellite scheduling to collaborate (relay) with ground stations in processing remote sensing data. This eliminates the need for the satellite to transmit all remote sensing image data to the ground station, reducing the amount of data to be transmitted, lowering the bandwidth requirements for the satellite-to-ground link, avoiding data backlog, and improving the real-time performance of data transmission and processing. This satellite-to-ground collaborative processing method reduces the satellite's computational load, lowers its energy consumption, extends its operational time and lifespan, avoids the impact of high satellite heat on computational performance, and improves satellite computational efficiency. Furthermore, by optimizing the initial processing model on the satellite at the ground station, the satellite's computational load is reduced, while simultaneously improving the utilization rate of the satellite's limited storage resources and increasing the efficiency of updating model parameters. Attached Figure Description

[0033] The present invention will be described by way of example and with reference to the accompanying drawings, wherein:

[0034] Figure 1 This is a flowchart of an embodiment of the remote sensing image processing method provided in this application, applied to a ground station.

[0035] Figure 2 This is a flowchart of another embodiment of the remote sensing image processing method provided in this application, applied to a ground station.

[0036] Figure 3 This is a flowchart of a remote sensing image processing method provided in this application, applied to a satellite.

[0037] Figure 4 This is a flowchart of another embodiment of the remote sensing image processing method provided in this application, applied to a satellite.

[0038] Figure 5 This is a flowchart of an embodiment of the remote sensing image processing method provided in this application for realizing satellite-ground cooperative control.

[0039] Figure 6 This is a data flow diagram of the remote sensing image processing method provided in this application in an embodiment of satellite-ground cooperative control.

[0040] Figure 7 This is a structural diagram of one embodiment of the remote sensing image processing system provided in this application.

[0041] Figure 8 This is a structural diagram of another embodiment of the remote sensing image processing system provided in this application. Detailed Implementation

[0042] All features disclosed in this specification, or all steps in all disclosed methods or processes, may be combined in any way, except for mutually exclusive features and / or steps.

[0043] Any feature disclosed in this specification (including any appended claims and abstract) may be replaced by other equivalent or similar features, unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is merely one example of a series of equivalent or similar features.

[0044] To address the current problems of high bandwidth requirements for satellite-to-ground link communication, poor real-time performance of data processing, and high satellite computing resource occupancy in remote sensing data processing, this application provides a remote sensing image processing method, apparatus, and storage medium, aiming to reduce the bandwidth requirements for satellite-to-ground link communication, improve the timeliness of data processing, and reduce the occupancy of satellite computing resources.

[0045] In one embodiment applied to a ground station, such as Figure 1 As shown, the remote sensing image processing method provided in this application includes the following steps:

[0046] Receive remote sensing data processing tasks. The remote sensing data processing task indicates what kind of remote sensing image data to be processed (i.e., what kind of remote sensing image data needs to be acquired), and what kind of processing to be performed on the remote sensing image data (i.e., the content of the processing of the remote sensing image data, such as classification, identification, processing, etc.).

[0047] Satellites are scheduled based on this remote sensing data processing task to cooperate with ground stations in processing remote sensing image data. Since remote sensing image data is acquired by satellites, the satellites typically perform the remote sensing data processing task first, and then the ground stations continue to perform the same task. Therefore, this cooperative relationship can also be called a relay relationship, where the ground stations take over from the satellites to continue the remote sensing data processing task after the satellites have completed theirs.

[0048] Optimize the primary processing model used in satellite remote sensing data processing tasks, such as a machine learning model, and upload the optimized model parameters to at least the satellite (i.e., the cooperating satellite).

[0049] By cooperating between satellites and ground stations to process remote sensing data, the amount of data transmitted from satellites can be reduced. This reduces the bandwidth requirements of the satellite-to-ground link while improving the real-time performance of data transmission, thereby enhancing the real-time performance of remote sensing data processing tasks.

[0050] The primary processing model configured on the satellite is a machine learning model, such as a basic image classification model or an image recognition model. Its parameters are updated by the ground station, reducing the occupancy rate of satellite computing resources, improving the utilization rate of satellite computing resources, and increasing the efficiency of updating the primary processing model. The initial model parameters of the primary processing model can also be obtained by the ground station training the initialized primary processing model based on sample data (such as historical remote sensing image data).

[0051] Remote sensing data processing tasks can be tasks involving the identification or classification of remote sensing image data. These tasks are typically initiated by ground station users, meaning the user (using appropriate terminal equipment) sends the task to the ground station. Correspondingly, the processing results of these tasks can ultimately be fed back to the user from the ground station.

[0052] As an optional implementation, the first processing model configured on the satellite is a lightweight model. A lightweight model, compared to a full machine learning model, has its parameter data volume significantly pruned, resulting in a corresponding sacrifice in performance. This allows the satellite to quickly perform preliminary processing of remote sensing image data with limited computing resources, filtering out the first remote sensing image data that requires further processing by the ground station, thereby improving the overall processing efficiency and real-time performance of the mission. It also conserves satellite energy and prevents a series of problems caused by excessive overheating.

[0053] For satellite scheduling, in one optional implementation, based on the current satellite status information of each satellite in the constellation connected to the ground station, including attribute information (such as orbit number, altitude, hardware configuration, etc.) and operational status information (such as load status, operating temperature, energy consumption, etc.), and based on the requirements of the remote sensing data processing task, at least one satellite is selected from the constellation for scheduling. By selecting satellites for task scheduling, tasks can be executed most efficiently, ensuring the real-time nature of task processing.

[0054] In some possible implementations, satellite status information may include attributes such as orbital position, altitude, status, and available resources (computing power, energy status, etc.). Remote sensing data processing tasks may require some or all of the satellite status information attributes.

[0055] For example, methods for satellite scheduling include:

[0056] 1) Filtering. Based on the attributes required for the remote sensing data processing task, filter out satellites in the constellation that do not meet the necessary attribute requirements. For example, the remote sensing data processing task requires satellites in a specified orbit / altitude; the task requires satellites to have a certain external device; the task requires a specific geographical location of the satellites; or other conditions required by the task. Filter out satellites that do not meet all the necessary attribute requirements based on the task's requirements.

[0057] 2) Optimal Selection. Among the filtered satellites, they are scored based on conventional attributes, and the one or more satellites with the highest scores are selected. For example, each attribute of a satellite is scored based on factors such as CPU, memory, disk availability, temperature, and energy status; the better the attribute, the higher the score. Finally, the one or more satellites with the highest total scores are selected for scheduling to achieve load balancing.

[0058] Furthermore, if a satellite no longer meets the requirements for remote sensing data processing after being rescheduled, other satellites will be rescheduled.

[0059] Considering that the constellation connected to the ground station may include more than one satellite, different satellites may be scheduled for processing remote sensing data received sequentially. However, the processing methods for remote sensing image data are the same. Therefore, in one optional implementation, the optimized model parameters (at the ground station) are uploaded to all satellites in the constellation that can communicate with it. This allows for batch updates of the model parameters of the first processing model on the satellites after each optimization of the first processing model, improving the efficiency and accuracy of satellite processing of remote sensing data. For example, after the ground station optimizes the first processing model, when scheduling other satellites for the next remote sensing data processing task, the optimized model parameters are first uploaded to that satellite to update the first processing model, and then that satellite processes the remote sensing data processing task.

[0060] For satellite-to-ground relay processing, in one optional implementation, such as Figure 2 As shown, methods for coordinating satellites to process acquired remote sensing image data include:

[0061] The system receives first remote sensing image data transmitted from the satellite, as well as the processing results retained after the satellite performs remote sensing data processing tasks. This first remote sensing image data includes at least the remote sensing image data for which the satellite has not transmitted processing results.

[0062] The first remote sensing image data is processed according to the remote sensing data processing task.

[0063] The satellite first processes the remote sensing data (using the first processing model) to obtain all processing results for all remote sensing image data corresponding to the task. Then, based on the corresponding screening conditions (such as reliability, accuracy, etc.), some processing results are selected and retained, while the remaining processing results are discarded. The retained processing results and the first remote sensing image data corresponding to the discarded processing results are transmitted to the ground station, which then takes over and processes the first remote sensing image data according to the remote sensing data processing task.

[0064] Taking the example of the ground station feeding back the processing results of the remote sensing data processing task to the user in the previous embodiment, the processing results transmitted from the satellite and the processing results of the first remote sensing image data are fused at the ground station and fed back to the user as the final processing result of the remote sensing data processing task.

[0065] In some specific embodiments, the processing of the first remote sensing image data involves using a configured second processing model at the ground station. This second processing model can also be a machine learning model, and can be trained using sample data from the initial second processing model.

[0066] In some specific embodiments, the initial configuration of the first processing model is obtained by pruning the second processing model, thereby running a lightweight model on the satellite, saving satellite energy and computing power consumption, extending satellite working time and service life, and avoiding excessive temperature rise.

[0067] As an optional implementation, the first processing model is optimized at the ground station based at least on the received first remote sensing image data for optimization of the first processing model.

[0068] For example, when there is no redundancy in the satellite-to-ground link, the satellite only transmits the first remote sensing image data to the ground station. At the ground station, in addition to processing the first remote sensing image data, the satellite optimizes the first processing model using the first remote sensing image data as samples and the processing results as labels, obtaining optimized model parameters. When there is redundancy in the satellite-to-ground link, the satellite can transmit the remaining (or as much as possible) remote sensing image data from the remote sensing data processing task to the ground station. The ground station uses all received remote sensing image data as samples and the processing results as labels to optimize the first processing model, obtaining optimized model parameters. Optimization of the first processing model can be achieved by directly optimizing its parameters, or by optimizing the second processing model and then pruning it to obtain the optimized first processing model. The pruned model parameters are the optimized model parameters. Optimizing the first processing model using real-time processed remote sensing image data improves its accuracy without consuming additional storage resources.

[0069] In some embodiments, such as Figure 3 As shown, this application provides another remote sensing image processing method, which is applied to satellites, and the method includes the following steps:

[0070] Receive scheduling instructions. These instructions are typically sent from the ground station.

[0071] In response to a received scheduling instruction, and according to the remote sensing data processing task indicated by the instruction, the satellite collaborates with the ground station using a first processing model to process the acquired remote sensing image data, thus completing the remote sensing data processing task in relay. The remote sensing image data is acquired by the satellite according to the instructions of the remote sensing data processing task, such as which area to acquire, what type of data to acquire, the quantity acquired, and the frequency.

[0072] Receive the model parameters after the ground station optimizes the first processed model.

[0073] In one optional implementation, the scheduling instruction is generated by the ground station based on the received remote sensing data processing task. The scheduling instruction carries an indication of the remote sensing image data required for the remote sensing data processing task, and the satellite acquires the collected remote sensing image data from the corresponding remote sensor according to this indication. The method by which the ground station schedules the satellite according to the remote sensing data processing task can be found in the previous embodiment and will not be repeated here. This relay processing method between the satellite and the ground station reduces the bandwidth requirements of the satellite-to-ground link, improves data processing efficiency, reduces satellite energy consumption and resource occupancy, prevents excessive overheating, and extends the satellite's operating time and lifespan. Optimization of model parameters by the ground station can further reduce satellite resource occupancy and improve the update efficiency of the first processing model.

[0074] As an alternative implementation, the first processing model on the satellite is a lightweight model, such as the lightweight model obtained by pruning the full machine learning model as described in the previous embodiment.

[0075] As an optional implementation method, such as Figure 4 As shown, the methods for processing acquired remote sensing image data in cooperation with ground stations include:

[0076] The first remote sensing image data, along with the processing results retained after the remote sensing data processing task, is transmitted to the ground station. This first remote sensing image data only includes remote sensing data for which processing results have not been transmitted.

[0077] In some specific embodiments, the satellite selects the processing results that reach the (confidence or accuracy) threshold from all processing results after performing the remote sensing data processing task on the acquired remote sensing image data based on a pre-configured confidence or accuracy threshold, and these are the retained processing results.

[0078] Alternatively, the threshold can be configured based on the satellite-to-ground link bandwidth and the satellite's computing resource usage. For example, when the satellite-to-ground link bandwidth is high, the threshold can be configured higher to retain fewer initial processing results and transmit more first-sensing image data to the ground station for processing, thus ensuring real-time data processing and higher accuracy. The same principle applies when satellite computing resource usage is high, transmitting more first-sensing image data to the ground station for processing. Conversely, when the satellite-to-ground link bandwidth is low, or satellite computing resource usage is low, the threshold can be configured lower to reduce the amount of first-sensing image data transmitted to the ground station. In other words, the threshold value is positively correlated with the satellite-to-ground link bandwidth or the satellite's computing resource usage.

[0079] Taking the classification of remote sensing image data as an example, the first processing model obtains several probability values ​​corresponding to the number of classification dimensions when processing the remote sensing data. The category corresponding to the highest probability value is taken as the classification result. A configured threshold is used: if the highest probability value reaches the threshold, the classification of the remote sensing image is considered to have high confidence; conversely, if the highest probability value is lower than the configured threshold, the classification of the remote sensing image is considered to have low confidence. Finally, all high-confidence classification results are retained, and the remote sensing image data with low confidence in the classification results is used as the first remote sensing image data, which is then transmitted to the ground station along with the high-confidence classification results.

[0080] Referring to the previous embodiments, the satellite transmits first remote sensing image data to the ground station, which can then optimize the first processing model based on this data. Alternatively, in an optional implementation, the satellite transmits all acquired remote sensing image data to the ground station when there is redundancy in the satellite-to-ground link. This allows the ground station to utilize all received remote sensing image data to optimize the first processing model. While optimizing the first processing model, the ground station can also optimize its own data processing model. For example, in an embodiment where the ground station uses a second processing model to process the first remote sensing image data, the ground station can optimize the second processing model based on the received remote sensing image data (either the first remote sensing image data or all remote sensing image data). Real-time optimization of the first and / or second processing models by the ground station can improve the accuracy of performing remote sensing data processing tasks.

[0081] In some embodiments, such as Figure 5 As shown, another remote sensing image processing method provided in this application includes:

[0082] Receive remote sensing data processing tasks. These tasks are typically input by the user / inputter.

[0083] The remote sensing data processing task control ground station coordinates with satellites to process the remote sensing image data acquired by the satellites.

[0084] The first processing model used by the control ground station to optimize satellite remote sensing data processing tasks.

[0085] The optimized model parameters should be uploaded to at least the satellite performing the remote sensing data processing task.

[0086] In the remote sensing image processing method of this application embodiment, a lightweight model can still be selected in the first processing model configured on the satellite to improve the efficiency of remote sensing data processing and quickly determine the remote sensing data to be transmitted.

[0087] As an optional implementation, similar to the previous embodiments, for satellite scheduling, based on the current satellite status information of each satellite in the constellation connected to the ground station, and based on the requirements of the remote sensing data processing task for the satellites, at least one satellite is selected from the constellation for scheduling. For specific feasible scheduling methods, please refer to the previous embodiments, which will not be repeated here.

[0088] It's possible to schedule more than one satellite to perform remote sensing data processing tasks. Furthermore, the different remote sensing data processing tasks received are often consecutive. Therefore, different satellites may be scheduled at the ground station for different remote sensing data processing tasks, and these satellites all process the remote sensing image data using a first processing model. To improve the efficiency of satellites updating the first processing model and ensure the accuracy of satellite processing of remote sensing image data in the next task, in one optional implementation, the optimized model parameters from the ground station are uploaded to all satellites in the constellation that can communicate with the ground station, so as to efficiently synchronize and update the model parameters of the first processing models of multiple satellites.

[0089] As an optional implementation method, such as Figure 6 As shown, the method for controlling ground stations to coordinate with satellites to process remote sensing image data acquired by satellites includes:

[0090] The control ground station receives the first remote sensing image data transmitted from the satellite, as well as the processing results retained by the remote sensing data processing task. The first remote sensing image data only includes remote sensing image data for which the satellite has not transmitted processing results. The control ground station processes the first remote sensing image data according to the remote sensing data processing task (the relevant operations indicated).

[0091] As an optional implementation method, the above-mentioned satellite scheduling process includes:

[0092] Based on the current satellite status information of each satellite in the constellation connected to the ground station, and based on the requirements of the remote sensing data processing task for the satellites, at least one satellite is selected from the constellation for the ground station to schedule.

[0093] Based on the remote sensing data processing task, a scheduling instruction is generated and sent to the scheduled satellite. This scheduling instruction controls the satellite to process the acquired remote sensing image data using a first processing model according to the remote sensing data processing task, and then transmits the retained processing results and the first remote sensing image data back to the ground station. This first remote sensing image data only includes remote sensing data for which processing results have not yet been transmitted.

[0094] The ground control station processes the first remote sensing image data;

[0095] The control ground station integrates the processing results from satellite downlink and its own processing results from the first remote sensing image data as the final processing result of the remote sensing data processing task. The integration method can be a simple combination of the two processing results, or it can involve separately labeling the processing object and the object being processed based on the combination.

[0096] In one feasible implementation, a confidence or accuracy threshold is pre-configured on the satellite. Based on this threshold, the retained processing results are selected from all processing results after the satellite processes remote sensing data. Examples of methods for filtering and retaining processing results can be found above and will not be repeated here.

[0097] This application embodiment also optimizes the first processing model of the satellite in real time. In an optional implementation, at least the ground station is controlled to optimize the first processing model based on the first remote sensing image data. This includes optimizing the first processing model with the first remote sensing image data, or optimizing the first processing model with all the downloaded remote sensing image data (including the corresponding processing results) when the satellite downloads more remote sensing image data. Similarly, when the ground station uses the second processing model to process the first remote sensing image data, the ground station can also be controlled to optimize the second processing model using the first remote sensing image data, or to optimize the second processing model using more remote sensing image data downloaded by the satellite.

[0098] Through the methods described above, this application achieves unified constellation management and mission scheduling. Faced with complex satellite constellation architectures and diverse onboard resources, unified mission scheduling and resource management effectively coordinate satellites with different configurations, avoiding conflicts in mission allocation. This unified management approach ensures optimal utilization of resources within the satellite constellation, improving mission scheduling efficiency and mission execution success rates.

[0099] Furthermore, for remote sensing data processing tasks, in some optional implementations, when controlling the ground station to schedule satellites to cooperate in completing remote sensing data processing tasks, a priority order is given to the specific matters of the satellites processing remote sensing data. It can be understood that the priority ordering of specific matters of the remote sensing data processing tasks, such as the priority ordering of the geographical location of remote sensing image data, the priority ordering of the data volume of different remote sensing image data, etc., is included in the remote sensing data processing tasks indicated by the scheduling instructions. Of course, the same priority processing requirements can also be formulated in the control of the ground station. In the embodiments of this application, the execution status of the remote sensing data processing tasks is also monitored, such as the satellite's operating status, the working status of the remote sensor, the progress of remote sensing image data processing, and the stability of the satellite-to-ground link, to ensure timely response in case of task execution anomalies.

[0100] In some alternative implementations, if any anomalies or malfunctions are identified during monitoring, the ground station will be instructed to take immediate remedial action. Fault handling may involve switching to a backup satellite, reconfiguring remote sensor parameters, or repairing the satellite-to-ground link. Timely fault handling reduces the risk of mission failure and ensures mission continuation.

[0101] In addition, necessary task adjustments and optimizations can be made based on monitoring data. This may include readjusting the execution priorities of specific task items, optimizing the utilization of satellite computing resources, or adjusting remote sensing data processing procedures, etc., to ensure that the mission can be executed smoothly and completed as planned.

[0102] In some embodiments, such as Figure 7 As shown, the system structure described above, which controls the ground station to schedule satellites to cooperate in completing remote sensing data processing tasks, can be achieved by the ground station simultaneously scheduling multiple satellites.

[0103] Based on the ideas of the above embodiments, in one feasible implementation, this application also provides a computer-readable storage medium storing computer instructions, which, when executed, can perform the remote sensing image processing method of any of the above embodiments.

[0104] As a possible implementation method, such as Figure 8 As shown, this application provides a remote sensing image processing device for ground stations. This device includes an information management module, a task scheduling module, a ground-based algorithm module, and a ground communication module. Additionally, a ground database can be configured to store data. In another embodiment of a remote sensing image processing device for satellites, the device includes an onboard communication module, a data acquisition module, and a satellite-based algorithm module. Alternatively, a satellite database can also be configured to store data. The two embodiments of the remote sensing image processing device will be described below.

[0105] For a remote sensing image processing device applied to a ground station, its information management module is used to receive remote sensing data processing tasks; the task scheduling module is used to schedule satellites to cooperate in processing remote sensing image data acquired by the satellites based on the remote sensing data processing tasks, including: instructing the scheduled satellites to download first remote sensing image data, and processing results retained by the remote sensing data processing tasks, wherein the first remote sensing image data only includes remote sensing image data for which the satellites have not downloaded processing results; the ground-side algorithm module is used to process the first remote sensing image data according to the remote sensing data processing tasks, and to optimize the first processing model used by the satellites when processing remote sensing data processing tasks; the ground communication module is used to receive the first remote sensing image data, and to upload the optimized model parameters to the satellites at least.

[0106] For a remote sensing image processing device applied to a satellite, its data acquisition module is used to acquire remote sensing image data (according to the requirements of the remote sensing data processing task indicated by the scheduling instruction); the satellite-side algorithm module is used to cooperate with the ground station to complete the processing of remote sensing image data according to the remote sensing data processing task indicated by the scheduling instruction, using a first processing model, including: determining the processing results to be retained in the remote sensing data processing task and the first remote sensing image data, which only includes remote sensing image data for which no processing results are retained; the onboard communication module is used to receive the scheduling instruction and the model parameters after the ground station optimizes the first processing model, and is also used to transmit the first remote sensing image data and the retained processing results to the ground station that sent the scheduling instruction.

[0107] The specific data that can be configured for each module in the two remote sensing image processing devices described above can be found in the corresponding process features in the remote sensing image processing method embodiments described above.

[0108] For ease of understanding, the operation flow of the system composed of the two remote sensing image processing devices described above is explained below:

[0109] The information management module receives task instructions from users that include remote sensing data processing tasks. It retrieves the current satellite status information of each satellite in the connected constellation from the ground database and transmits it to the task scheduling module for satellite scheduling. The scheduling instructions are then sent to the scheduled satellites via a space-to-ground link established by the ground communication module and the onboard communication module. Upon receiving the scheduling instructions, the satellite (or its onboard communication module) collects the required remote sensing image data from the corresponding remote sensors according to the requirements of the remote sensing data processing task indicated in the instructions, and stores it in the satellite database. In addition, the data acquisition module periodically collects satellite status information and stores it in the satellite database. During the space-to-ground communication window, the satellite transmits its status information to the ground station via its onboard communication module. The satellite-side algorithm module retrieves the remote sensing image data required for this remote sensing data processing mission from the satellite database. Following the mission instructions, it processes the image data using a configured first processing model, obtaining all processing results. Based on a configured threshold, it filters out the retained results. These retained results, along with the corresponding first remote sensing image data for the non-retained results, are then transmitted to the ground station via the onboard communication module. The transmitted processing results are saved in the ground database, while the transmitted first remote sensing image data is further processed by the ground-side algorithm module using a configured second processing model. The results are also saved in the ground database. The information management module retrieves the processing results from the ground database, combining them with the satellite-transmitted results to form the final processing result, which is then fed back to the user. In addition, the ground-side algorithm module also uses the first remote sensing image data (or all remote sensing image data received when the satellite is transmitting more remote sensing image data) to optimize the first processing model (the first processing model is also stored in the ground station), and uploads the optimized model parameters to the satellite (the currently scheduled satellite or all satellites in the connected constellation that can communicate) through the ground communication module to update the model parameters of the first processing model in the satellite-side algorithm module.

[0110] This invention is not limited to the specific embodiments described above. The invention extends to any new feature or combination disclosed in this specification, as well as any new method or process step or combination disclosed herein.

Claims

1. A method of processing a remote sensing image, characterized in that, The application is applied to a ground station, comprising: receiving a remote sensing data processing task; scheduling a satellite to cooperate to complete processing of remote sensing image data collected by the satellite based on the remote sensing data processing task, comprising: receiving first remote sensing image data downloaded by the satellite, and processing reserved processing results of the remote sensing data processing task, wherein the first remote sensing image data only includes remote sensing image data without downloaded processing results, and the processing of the reserved processing results of the remote sensing data processing task is part of all processing results of all remote sensing image data corresponding to the remote sensing data processing task, which is filtered according to corresponding filtering conditions by using a first processing model; processing the first remote sensing image data according to the remote sensing data processing task; optimizing the first processing model used by the satellite when processing the remote sensing data processing task; uploading the optimized model parameters to at least the satellite.

2. The remote-sensing image processing method of claim 1, wherein, The method for scheduling a satellite comprises: According to the current satellite state information of each satellite in the connected constellation, based on the requirements of the remote sensing data processing task for the satellite, at least one satellite is selected from the constellation for scheduling.

3. The remote sensing image processing method of claim 2, wherein, Uploading the optimized model parameters to at least the satellite, comprising: Uploading the optimized model parameters to all satellites in the constellation that can communicate with them.

4. The remote-sensing image processing method of claim 1, wherein, The method for scheduling a satellite to cooperate to complete processing of collected remote sensing image data further comprises: Fusing the processing results downloaded by the satellite and the processing results of the first remote sensing image data as the final processing results of the remote sensing data processing task.

5. The remote sensing image processing method of claim 1, wherein, Optimizing the first processing model used by the satellite when processing the remote sensing data processing task, comprising: Optimizing the first processing model based at least on the first remote sensing image data.

6. A method of processing a remote sensing image, characterized in that, The application is applied to a satellite, comprising: receiving a scheduling instruction; According to the remote sensing data processing task indicated by the scheduling instruction, cooperating with the ground station to complete the processing of the collected remote sensing image data by using a first processing model, which includes: downloading first remote sensing image data and processing results reserved for the remote sensing data processing task to the ground station, wherein the first remote sensing image data only includes remote sensing image data without downloaded processing results, and the processing results reserved for the remote sensing data processing task are part of all processing results of all remote sensing image data corresponding to the remote sensing data processing task, which are filtered according to corresponding filtering conditions by using a first processing model; receiving model parameters of the first processing model optimized by the ground station.

7. The remote sensing image processing method of claim 6, wherein, The method for selecting reserved processing results comprises: Based on a pre-configured threshold, selecting processing results reaching the threshold from all processing results after performing the remote sensing data processing task on collected remote sensing image data.

8. The remote sensing image processing method of claim 6, wherein, Further comprising: Downloading all remote sensing image data to the ground station when there is redundancy in the satellite-ground link.

9. A method of processing a remote sensing image, characterized in that, Comprising: receiving a remote sensing data processing task; The ground station is controlled based on the remote sensing data processing task to schedule a satellite to cooperatively complete processing of the remote sensing image data collected by the satellite, including: controlling the ground station to receive first remote sensing image data transmitted by the satellite, and processing reserved processing results of the remote sensing data processing task, wherein the first remote sensing image data only includes remote sensing image data that is not transmitted by the satellite with processing results, and the processing reserved processing results of the remote sensing data processing task is a part of all processing results of all remote sensing image data corresponding to the remote sensing data processing task, which is filtered from the all processing results by the satellite using a first processing model according to corresponding filtering conditions; and controlling the ground station to process the first remote sensing image data according to the remote sensing data processing task; controlling the ground station to optimize the first processing model used by the satellite to process the remote sensing data processing task; uploading the optimized model parameters to at least the satellite.

10. A computer readable storage medium storing computer instructions, characterized in that, The computer instructions are executed to perform the remote sensing image processing method of any one of claims 1-9.

11. A remote sensing image processing apparatus, characterized by comprising: Applied to a ground station, including: an information management module configured to receive a remote sensing data processing task; a task scheduling module configured to schedule a satellite to cooperatively complete processing of remote sensing image data collected by the satellite based on the remote sensing data processing task, including: instructing the scheduled satellite to transmit first remote sensing image data, and processing reserved processing results of the remote sensing data processing task, wherein the first remote sensing image data only includes remote sensing image data that is not transmitted by the satellite with processing results, and the processing reserved processing results of the remote sensing data processing task is a part of all processing results of all remote sensing image data corresponding to the remote sensing data processing task, which is filtered from the all processing results by the satellite using a first processing model according to corresponding filtering conditions; a ground algorithm module configured to process the first remote sensing image data according to the remote sensing data processing task, and optimize the first processing model used by the satellite to process the remote sensing data processing task; a ground communication module configured to receive the first remote sensing image data, and upload the optimized model parameters to at least the satellite.

12. A remote sensing image processing apparatus, characterized by comprising: For a satellite, including: a data acquisition module configured to acquire remote sensing image data; a satellite algorithm module configured to cooperatively complete processing of the remote sensing image data with a ground station according to a remote sensing data processing task indicated by a scheduling instruction using a first processing model, including: determining reserved processing results of the remote sensing data processing task and first remote sensing image data, wherein the first remote sensing image data only includes remote sensing image data that is not reserved with processing results, and the processing reserved processing results of the remote sensing data processing task is a part of all processing results of all remote sensing image data corresponding to the remote sensing data processing task, which is filtered from the all processing results by the satellite using a first processing model according to corresponding filtering conditions; a satellite communication module configured to receive the scheduling instruction and model parameters of the first processing model optimized by the ground station, and further configured to transmit the first remote sensing image data and the reserved processing results to the ground station that sends the scheduling instruction.

Images