Remote sensing data processing methods and systems, electronic devices and storage media

Abstract

This application provides a remote sensing data processing method and system, electronic device, and storage medium, relating to the technical field of remote sensing data processing. The remote sensing data processing method includes: acquiring target remote sensing data from a target satellite constellation based on received user data requests; parsing the sensor type corresponding to the target remote sensing data to determine at least one target type of remote sensing data; determining the metadata information of the target type remote sensing data; performing block processing on the target type remote sensing data based on the metadata information to obtain multiple sub-target type remote sensing data; and performing distributed parallel computation on the multiple sub-target type remote sensing data to obtain the target output result. This application enables online access and acquisition of remote sensing data by parsing user data requests online, avoiding local downloads and offline processing. By migrating remote sensing data and algorithms to the cloud, it solves problems such as long processing time, low processing efficiency, and high local resource consumption costs.

Description

Technical Field

[0001] This application relates to the technical field of remote sensing data processing, and more specifically, to a remote sensing data processing method and system, electronic device and storage medium. Background Technology

[0002] In recent years, with the rapid development of commercial aerospace technology and the continuous deployment of remote sensing satellite constellations, Earth observation capabilities have achieved a leapfrog improvement, and multi-source, multi-scale, and multi-temporal remote sensing data have accumulated rapidly. For example, a single high-resolution remote sensing satellite can generate tens of terabytes of raw image data every day, and the total amount of remote sensing data globally has grown rapidly from petabytes to exabytes.

[0003] Currently, remote sensing data processing generally employs the traditional offline processing model. The basic workflow of this offline model involves end users first downloading massive amounts of data from geographically dispersed data centers to their local machine. Then, they use appropriate software in their local computing environment to perform a series of complex data processing tasks, including radiometric correction, geometric correction, orthorectification, stitching, fusion, and information extraction. However, the inventors of this application have discovered that remote sensing data processing based on the offline processing model suffers from at least the following problems:

[0004] 1. The contradiction between processing efficiency and ultra-large-scale remote sensing data: The transmission and computation of EB-level data will exceed the capacity limit of most local computing nodes, resulting in a long cycle from data preparation to result output. There are problems such as huge data volume, long time consumption and low processing efficiency, making it difficult to support real-time, parallel or multi-user collaborative application scenarios.

[0005] 2. High local resource consumption and cost: Building and maintaining high-performance computing clusters and storage systems capable of processing ultra-large-scale remote sensing data requires significant investment and ongoing operation and maintenance costs.

[0006] 3. Data silos exist: Algorithm resources, storage resources, and data resources are fixed in users' local environments, forming data silos. This makes it difficult to achieve cross-regional and multi-departmental data sharing and collaborative processing, and limits the updating of processing algorithms and the sharing of model resources.

[0007] The content in the background section is merely technology known to the public and does not necessarily represent existing technology in this field. Summary of the Invention

[0008] This application provides a remote sensing data processing method and system, electronic device and storage medium, which aim to solve at least one of the above-mentioned technical problems.

[0009] According to one aspect of this application, a remote sensing data processing method is provided, comprising: acquiring target remote sensing data of a target satellite constellation based on received user data requirements; parsing the sensor type corresponding to the target remote sensing data to determine at least one target type of remote sensing data; determining metadata information of the target type of remote sensing data; performing block processing on the target type of remote sensing data based on the metadata information to obtain multiple sub-target type remote sensing data; and performing distributed parallel computing on the multiple sub-target type remote sensing data to obtain a target output result.

[0010] According to some embodiments of this application, obtaining target remote sensing data of a target satellite constellation based on received user data requests includes: obtaining satellite cataloging data from at least one satellite manufacturer; determining corresponding remote sensing data parameter information based on user data requests; and obtaining target remote sensing data based on the remote sensing data parameter information and the satellite cataloging data.

[0011] According to some embodiments of this application, obtaining target remote sensing data of a target satellite constellation based on received user data requirements includes: determining whether the target remote sensing data meets a first preset requirement; if the target remote sensing data does not meet the first preset requirement, replacing the target remote sensing data according to user data requirements until the replaced target remote sensing data meets the first preset requirement.

[0012] According to some embodiments of this application, segmenting target type remote sensing data into blocks based on metadata information to obtain multiple sub-target type remote sensing data includes: segmenting the target type remote sensing data into blocks according to bands when performing different calculations on different bands of the target type remote sensing data to obtain multiple sub-target type remote sensing data.

[0013] According to some embodiments of this application, segmenting target type remote sensing data into blocks based on metadata information to obtain multiple sub-target type remote sensing data includes: performing the same calculation on different bands of the target type remote sensing data, segmenting the target type remote sensing data into blocks as a whole to obtain multiple sub-target type remote sensing data.

[0014] According to some embodiments of this application, the remote sensing data processing method further includes: when there are multiple user data demands, balancing the multiple user data demands to multiple computing clusters based on load balancing processing.

[0015] According to some embodiments of this application, the remote sensing data processing method further includes: determining whether the target output result meets the second preset requirement; if the target output result does not meet the second preset requirement, adjusting the metadata information based on the determination result until the adjusted target output result meets the second preset requirement.

[0016] According to another aspect of this application, a remote sensing data processing system is provided, including a data acquisition module, a data parsing module, a block processing module, and a data processing module. The data acquisition module acquires target remote sensing data from a target satellite constellation based on received user data requirements; the data parsing module parses the sensor type corresponding to the target remote sensing data to determine at least one target type of remote sensing data, and to determine the metadata information of the target type of remote sensing data; the block processing module performs block processing on the target type of remote sensing data based on the metadata information to obtain multiple sub-target type remote sensing data; the data processing module performs distributed parallel computation on the multiple sub-target type remote sensing data to obtain the target output result.

[0017] According to another aspect of this application, an electronic device is also provided. The electronic device includes: one or more processors; and a storage device for storing one or more programs, which, when executed by the one or more processors, enable the one or more processors to implement the remote sensing data processing method described above.

[0018] According to another aspect of this application, a non-volatile computer-readable storage medium is also provided. This storage medium stores a computer program that, when executed by a processor, can implement the remote sensing data processing method described above.

[0019] According to another aspect of this application, this application also provides a computer program product. The computer program product includes: a computer program stored on a computer-readable storage medium; the computer program includes program instructions that, when executed by a computer, cause the computer to perform the remote sensing data processing method as described above.

[0020] Beneficial effects

[0021] This application can acquire target remote sensing data from a target satellite constellation based on received user data requirements, analyze the sensor type corresponding to the target remote sensing data, and determine at least one target type of remote sensing data. Then, this application can determine the metadata information of the target type remote sensing data, perform block processing based on the metadata information to obtain multiple sub-target type remote sensing data, and perform distributed parallel computing on the multiple sub-target type remote sensing data to obtain the target output result.

[0022] This application enables online access and acquisition of remote sensing data by parsing user data requirements online, avoiding local downloads and offline processing. By migrating remote sensing data and algorithms to the cloud, it solves problems such as long processing times, low efficiency, and high local resource consumption costs. This application also addresses the processing of ultra-large-scale remote sensing data, which cannot be handled by a single machine, by segmenting remote sensing data into blocks. Furthermore, this application improves the processing efficiency and stability of ultra-large-scale remote sensing data through distributed parallel computing and a high-performance storage system. Finally, this application overcomes the limitations of data silos through standardized service interfaces and a unified data management mechanism, enabling collaborative analysis and resource sharing among multiple users and departments. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A flowchart illustrating a remote sensing data processing method according to an embodiment of this application is shown.

[0025] Figure 2 This diagram illustrates the structure of a job scheduling system according to an embodiment of this application.

[0026] Figure 3 This diagram illustrates yet another flow chart of the remote sensing data processing method according to an embodiment of this application.

[0027] Figure 4 This diagram illustrates yet another flow chart of the remote sensing data processing method according to an embodiment of this application.

[0028] Figure 5 This diagram illustrates a block processing method based on wavebands, according to an embodiment of this application.

[0029] Figure 6 This diagram illustrates a block-based processing method according to an embodiment of the present application.

[0030] Figure 7 A schematic diagram illustrating load balancing in an embodiment of this application is shown;

[0031] Figure 8 This diagram illustrates a load balancing optimization embodiment of the present application.

[0032] Figure 9 This application shows a time comparison chart of embodiments;

[0033] Figure 10A schematic diagram of the structure of a remote sensing data processing system according to an embodiment of this application is shown.

[0034] Explanation of reference numerals in the attached figures:

[0035] Remote sensing data processing system 1; data acquisition module 11; data parsing module 12; block processing module 13; data processing module 14; data quality inspection module 15. Detailed Implementation

[0036] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0037] The inventors of this application have also discovered that some existing technologies offer online processing modes for remote sensing data based on general-purpose cloud platforms, which can migrate computing tasks to the cloud. While this online processing mode can alleviate the pressure on local resources to some extent, it lacks optimization for the characteristics of large-scale remote sensing data and computationally intensive tasks. For example, when facing the processing needs of ultra-large-scale remote sensing data or multiple concurrent tasks, it still has shortcomings in areas such as scheduling computing resources, coordination between storage and computing units, and network optimization across supercomputing nodes, and its processing efficiency and resource utilization need to be improved.

[0038] According to one aspect of this application, a remote sensing data processing method is provided. Figure 1 This diagram illustrates a flow chart of a remote sensing data processing method according to an embodiment of this application. Figure 1 As shown, the remote sensing data processing method may include steps S100-S500.

[0039] For example, this remote sensing data processing method can be performed by a remote sensing data processing system with computing capabilities.

[0040] According to the example embodiment, in step S100, the remote sensing data processing system acquires the target remote sensing data of the target satellite constellation based on the received user data requirements.

[0041] For example, a user can input their data request (essentially an online data processing order) into the remote sensing data processing system via an interactive page. In response to this request, the system can access the target remote sensing data collected by the target satellite constellation from at least one satellite vendor's data platform, thereby acquiring the target remote sensing data.

[0042] For example, user data requirements include, but are not limited to, data requirements in multiple dimensions such as geographical coverage, time range, required type and specifications, and target quality of remote sensing data.

[0043] For example, the target satellite constellation includes, but is not limited to, satellite constellations with remote sensing data acquisition capabilities such as the Gaofen constellation, Jilin constellation, resource series satellite constellation, environmental series satellite constellation, and radar satellite constellation.

[0044] In step S200, the remote sensing data processing system analyzes the sensor type corresponding to the target remote sensing data to determine at least one type of target remote sensing data.

[0045] For example, the sensor type refers to the type of payload equipment used by the target satellite constellation to acquire remote sensing data. Different sensor types will generate different types of remote sensing data. The remote sensing data processing system analyzes the sensor type of the acquired target remote sensing data and identifies the target remote sensing data as the corresponding target type remote sensing data.

[0046] For example, the types of payload equipment include, but are not limited to, spectroscopic cameras, synthetic aperture radars, and imaging spectrometers. Correspondingly, the target type remote sensing data includes, but is not limited to, payload data of types such as optical remote sensing data, microwave remote sensing data, and hyperspectral remote sensing data.

[0047] In step S300, the remote sensing data processing system determines the metadata information of the target type remote sensing data.

[0048] For example, a remote sensing data processing system can parse metadata information from remote sensing data of a target type, thereby enabling the precise extraction of key parameters. Exemplarily, metadata information includes, but is not limited to, spatial reference system, image row and column count, bands, data type, pixel depth, and time.

[0049] In step S400, the remote sensing data processing system divides the target type remote sensing data into blocks based on metadata information to obtain multiple sub-target type remote sensing data.

[0050] In step S500, the remote sensing data processing system performs distributed parallel computation on remote sensing data of multiple sub-target types to obtain target output results.

[0051] For example, a remote sensing data processing system can determine the quantity and size of sub-target type remote sensing data based on metadata information, thereby dividing the target type remote sensing data into multiple sub-target type remote sensing data. The system decomposes ultra-large-scale remote sensing data, which cannot be processed by a single machine, into multiple sub-target type remote sensing data through block processing. These sub-target type remote sensing data are then distributed to multiple nodes in a distributed computing cluster for parallel processing. After all computing nodes have completed their calculations, the system merges the results from all nodes into the final output (i.e., the target output result).

[0052] For example, the target output results include, but are not limited to, diverse satellite results including raw data, standard products, orthophoto products, advanced products, and application products, and this application does not limit them.

[0053] It is understandable that ultra-large-scale remote sensing data (i.e., the total number of pixels in the remote sensing data) is characterized by its large size and high resolution. For processing ultra-large-scale remote sensing data, single-machine processing methods struggle to overcome the limitations of memory and processing power. This application, through distributed parallel processing of sub-target type remote sensing data, employs a "block-computation-merging" computational logic. Based on spatial indexing and a parallel computing framework, it can achieve efficient processing of ultra-large-scale remote sensing data.

[0054] Figure 2 A schematic diagram of the structure of the job scheduling system according to an embodiment of this application is shown.

[0055] According to the example embodiment, the remote sensing data processing system can perform distributed parallel computing on remote sensing data of multiple sub-target types based on the supercomputing cloud network to obtain target output results.

[0056] For example, remote sensing data processing systems can build distributed parallel computing nodes based on supercomputing power (such as supercomputing processor architecture) and construct high-performance computer storage that meets business scenarios. Then, the remote sensing data processing system is based on a job scheduling system (such as... Figure 2 The Slurm job scheduling system shown submits remote sensing data of multiple sub-target types to a supercomputing cloud network formed by distributed parallel computing nodes for parallel computing.

[0057] Slurm Job Scheduling System: Simple Linux Utility for Resource Management, is an open-source, scalable job scheduling system.

[0058] For example, such as Figure 2As shown, the Slurm job scheduling system can uniformly receive user data requests (i.e., user-submitted data tasks) based on multiple algorithm plugins, and dynamically schedule and distribute tasks according to the resource status of each computing node (such as computing node 1, computing node 2, and computing node 3), so that user requests can be allocated to various computing nodes. Furthermore, each computing node can provide real-time feedback on its job status (i.e., supercomputing job status information) to the Slurm job scheduling system, thus forming a closed-loop process from receiving user data requests, distributed scheduling to execution monitoring.

[0059] Optionally, the remote sensing data processing system can also feed back the target output results to the user in a preset format.

[0060] For example, a remote sensing data processing system can automatically upload target output results to a user-specified storage space (such as a cloud drive), or it can provide a preview to the user based on a visual page.

[0061] For example, the remote sensing data processing system can publish and aggregate target output results through pre-defined standard geographic information services, thereby forming an interoperable service cluster. Furthermore, the remote sensing data processing system can also leverage WebGL (Web Graphics Library) 3D earth engine or 2D front-end library to achieve seamless overlay, real-time rendering, and multi-dimensional dynamic display of massive raster and vector data for the visualization of target output results.

[0062] For example, the preset standard can be the OGC standard. OGC: Open Geospatial Consortium.

[0063] Optionally, the remote sensing data processing system can also support interactive querying of metadata, spatial range, and spectral bands of the target output results, and can provide data support such as time series analysis and roll-up comparison, thereby transforming the target output results into dynamic spatial information that can be explored and analyzed in depth.

[0064] Through the above embodiments, this application can acquire target remote sensing data of a target satellite constellation based on the received user data requirements, parse the sensor type corresponding to the target remote sensing data, and determine at least one target type of remote sensing data. Then, this application can determine the metadata information of the target type remote sensing data, perform block processing on the target type remote sensing data based on the metadata information to obtain multiple sub-target type remote sensing data, and perform distributed parallel computing on the multiple sub-target type remote sensing data to obtain the target output result.

[0065] This application enables online access and acquisition of remote sensing data by parsing user data requirements online, avoiding local downloads and offline processing. By migrating remote sensing data and algorithms to the cloud, it solves problems such as long processing times, low efficiency, and high local resource consumption costs. This application also addresses the processing of ultra-large-scale remote sensing data, which cannot be handled by a single machine, by segmenting remote sensing data into blocks. Furthermore, this application improves the processing efficiency and stability of ultra-large-scale remote sensing data through distributed parallel computing and a high-performance storage system. Finally, this application overcomes the limitations of data silos through standardized service interfaces and a unified data management mechanism, enabling collaborative analysis and resource sharing among multiple users and departments.

[0066] Through the above embodiments, this application can realize preprocessing operations such as data cropping, coordinate reprojection, image mosaicking, cloud removal and enhancement; it can realize intelligent analysis tasks such as land feature classification, target detection, change detection, and time series analysis; it can realize unified resource modeling and dynamic scheduling based on supercomputing cloud network; and it can also realize user-oriented visualization display and result service publishing.

[0067] This application can analyze user data requirements online and, combined with dynamic scheduling and automated execution mechanisms, construct a closed-loop remote sensing data processing system integrating "data acquisition, data processing, data analysis, data quality inspection, and data delivery." This application can meet the high-precision, high-efficiency, automated, and intelligent processing needs of various payload data, including optical remote sensing data, microwave remote sensing data, and hyperspectral remote sensing data. This application can adapt different data results to different target types of remote sensing data, thereby providing accurate aerospace big data results. This application can couple multiple types of computing power, such as intelligent computing, supercomputing, big data, and cloud computing, to improve the efficiency and intelligence level of remote sensing data processing.

[0068] This application enables end-to-end, one-stop automated processing of online remote sensing data orders. By integrating remote sensing data from multiple satellite constellations, it constructs a fully intelligent system covering data security management, intelligent quality inspection, cleaning and screening, processing, results archiving, and delivery. Relying on continuous feedback and status monitoring mechanisms, this application can automatically generate diverse satellite results, from raw data to advanced application products, meeting the data processing needs of multiple scenarios and payloads.

[0069] This application leverages artificial intelligence technology and a unified resource scheduling framework for multi-task parallel processing to automate the entire process of processing massive, multi-source, heterogeneous imagery data. It provides a one-stop service covering task-driven automation, data verification automation, data processing automation, intelligent analysis automation, output quality inspection automation, and output delivery automation. For example, the task-driven module can monitor and automatically push target remote sensing data in real time. After the data verification module completes integrity verification and quality checks, the data processing module generates standardized data. Subsequently, the intelligent analysis module produces target output results based on deep learning algorithms. After passing dual verification by the output quality inspection module, the results are finally intelligently uploaded to a designated delivery cloud drive via the output delivery module. The various functional modules in this application can collaborate through message queues to form a closed-loop remote sensing data processing system integrating "data acquisition - data processing - data analysis - data quality inspection - data delivery."

[0070] Figure 3 This is another schematic flowchart illustrating the remote sensing data processing method according to an embodiment of this application.

[0071] Optionally, such as Figure 3 As shown, step S100 may also include steps S110-S130.

[0072] In step S110, the remote sensing data processing system acquires satellite cataloging data from at least one satellite manufacturer.

[0073] For example, satellite cataloging data serves as an index describing the core characteristics and status of remote sensing data. Satellite cataloging data includes at least the remote sensing data ID, coverage information, time information, cloud cover information, and storage location information.

[0074] Remote sensing data processing systems can automatically import satellite catalog data from the data platforms of at least one satellite manufacturer through standardized data interfaces, and can also aggregate and preview satellite catalog data.

[0075] In step S120, the remote sensing data processing system determines the corresponding remote sensing data parameter information based on the user's data requirements.

[0076] In step S130, the remote sensing data processing system acquires target remote sensing data based on remote sensing data parameter information and satellite cataloging data.

[0077] For example, remote sensing data parameters include, but are not limited to, time parameters, coverage parameters, and cloud cover information.

[0078] Upon receiving a user's data request, the remote sensing data processing system automatically parses the request, converting it into specific remote sensing data parameter information. Based on this parameter information, it determines the most suitable target satellite constellation. Then, using previously acquired and aggregated satellite catalog data, the system performs rapid queries and matching within the database based on the remote sensing data parameter information, thereby obtaining the required target remote sensing data.

[0079] Through the above embodiments, this application can automatically extract multi-source satellite catalog data, convert structured user data requirements into accurate remote sensing data parameter information, and perform analysis and matching based on remote sensing data parameter information and satellite catalog data, thereby achieving accurate acquisition of target remote sensing data.

[0080] Figure 4 This is another schematic flowchart illustrating the remote sensing data processing method according to an embodiment of this application.

[0081] Optionally, such as Figure 4 As shown, step S100 may also include steps S140 and S150.

[0082] In step S140, the remote sensing data processing system determines whether the target remote sensing data meets the first preset requirements.

[0083] In step S150, if the target remote sensing data does not meet the first preset requirement, the remote sensing data processing system replaces the target remote sensing data according to the user's data requirements until the replaced target remote sensing data meets the first preset requirement.

[0084] For example, the first preset requirement can be a data requirement for the target remote sensing data that is customized according to user needs. The first preset requirement includes, but is not limited to, data integrity requirements, imaging quality requirements, and specification requirements.

[0085] For example, data integrity can include whether the data is complete and whether it is corrupted. Imaging quality requirements can include cloud cover (e.g., for optical remote sensing data, whether the cloud cover is below a preset threshold), noise level (e.g., whether there is noise), and data quality (e.g., whether it is stable or whether there is distortion). Specification requirements can include whether the resolution, number of bands, etc., meet the user's data needs.

[0086] For example, a remote sensing data processing system can perform quality checks on the acquired target remote sensing data based on a first preset requirement. If the target remote sensing data does not meet the first preset requirement, the remote sensing data processing system can call the user's data request again, change the data source, until a replacement target remote sensing data that meets the preset requirement is determined.

[0087] Through the above embodiments, this application can perform quality inspection on the acquired target remote sensing data based on the first preset requirements. When the target remote sensing data is unqualified, a feedback loop can be performed based on the quality inspection results. By dynamically adjusting the input data, the reliability of the finally acquired target remote sensing data can be ensured.

[0088] Figure 5 This diagram illustrates a block processing method based on wavebands, according to an embodiment of this application. Figure 6 This diagram illustrates a block-based processing method according to an embodiment of the present application.

[0089] Optionally, in step S400, the remote sensing data processing system performs different calculations on different bands of the target type remote sensing data, and then divides the target type remote sensing data into blocks according to the bands to obtain multiple sub-target type remote sensing data.

[0090] For example, when a user's data requirement is to perform different calculation logics on different bands of remote sensing data of a target type, such as image radiometric calibration (different gain / bias coefficients for each band), enhancement of a specific band (such as enhancement of the green band), band operations, etc., the remote sensing data processing system will divide the remote sensing data of the target type into blocks according to bands.

[0091] For example, such as Figure 5 As shown (different colors represent different bands), the remote sensing data processing system traverses the bands of the target type remote sensing data. For the currently being processed band, the system divides it spatially according to preset dimensions, generating multiple sub-target type remote sensing data sets containing only that band. These single-band sub-target type remote sensing data sets are then distributed to distributed computing nodes for parallel computation. After computation, the results are written back sequentially to the corresponding band and spatial location.

[0092] Through the above embodiments, this application can segment remote sensing data of a target type according to its bands when performing different calculations on different bands. This configuration allows the application of different parameters to the corresponding bands during data processing, thereby ensuring the accuracy of band-specific processing. Furthermore, the segmentation method, which only requires loading data blocks of a single band, reduces the instantaneous demand on single-machine memory.

[0093] Optionally, in step S400, the remote sensing data processing system performs the same calculations on different bands of the target type remote sensing data, and then divides the target type remote sensing data into blocks to obtain multiple sub-target type remote sensing data.

[0094] For example, when a user's data requirement is to perform the same computational logic on different bands of remote sensing data of a target type, the remote sensing data processing system will process the target remote sensing data in blocks as a whole.

[0095] For example, such as Figure 6 As shown (different colors represent different bands), the remote sensing data processing system can divide the overall target type remote sensing data (including the entire image data across all bands) into several regular sub-target type remote sensing data sets covering all bands, based on a preset size and spatial range. The system then distributes these multi-band sub-target type remote sensing data sets to distributed computing nodes for parallel computation. Each computing node performs a unified operation on all bands within the sub-target type remote sensing data set and writes back the complete result data block.

[0096] Through the above embodiments, this application can divide the target type remote sensing data into blocks for overall processing while performing the same calculations on different bands of the target type remote sensing data. This configuration allows all band data within a spatial region to be assigned to a single computing node at once. This computing node can perform unified vectorization or parallel computation on all pixels (regardless of band) of the sub-target type remote sensing data without switching between bands, maximizing spatial locality and improving processing efficiency.

[0097] Figure 7 A schematic diagram illustrating load balancing in an embodiment of this application is shown; Figure 8 This diagram illustrates a load balancing optimization embodiment of the present application. Figure 9 A time comparison chart of embodiments of this application is shown.

[0098] Optionally, the remote sensing data processing method may also include: when there are multiple user data demands, the remote sensing data processing system distributes the multiple user data demands to multiple computing clusters (or servers) based on load balancing.

[0099] For example, such as Figure 7 As shown, when the remote sensing data processing system receives multiple user data requests, it can distribute the multiple user data requests to different servers (server A, server B, server C) based on load balancing.

[0100] Through the above embodiments, this application can distribute user data requests to different servers based on load balancing processing, and distribute the computing load evenly, thus avoiding the problem of overload on a single server.

[0101] Optionally, such as Figure 8As shown, remote sensing data processing systems can optimize batch task submission scripts, thereby reducing time wasted due to serial task submissions. Remote sensing data processing systems can also distribute multiple user data requests (dateHyper) to multiple computing clusters (such as HAProxy load balancers) based on load balancing (e.g., HAProxy load balancers). Figure 8 In the various computing nodes (such as iFactory shown), and in remote sensing data processing systems, the Least-Conn strategy (scheduled scheduling with the fewest connections) can be selected to reduce cluster task waiting caused by some time-consuming tasks.

[0102] For example, such as Figure 8 As shown, for application scenarios with five computing clusters, this application can significantly reduce computing time (e.g., in the case of 1000 order tasks, the time taken by a single cluster is optimized from 36 minutes to 1.5 minutes for five clusters).

[0103] Among them, Figure 8 In this architecture, Storage 1 is shared storage for multiple clusters (such as single-cluster, triple-cluster, and five-cluster clusters), and all compute nodes' programs can access Storage 1. 1000 represents the number of order tasks. In the first phase of optimization, the HAProxy load balancer split the 1000 order tasks into three groups, each containing 334, 333, and 333 order tasks respectively. In the latest optimized architecture, the HAProxy load balancer splits the 1000 order tasks into multiple groups, each containing n order tasks.

[0104] For example, Figure 9 The chart shows a comparison of processing times for 500, 1000, 1500, 2000, and 5000 orders (i.e., user demand quantities) concurrently in real time. Figure 9 As shown, for a request for 500 sets of user data, the shortest processing time is 29 seconds, the longest is 77 seconds, and the average is 47.4 seconds; for a request for 1000 sets of user data, the shortest is 54 seconds, the longest is 146 seconds, and the average is 87 seconds; for a request for 1500 sets of user data, the shortest is 116 seconds, the longest is 227 seconds, and the average is 196 seconds; for a request for 2000 sets of user data, the shortest is 146 seconds, the longest is 276 seconds, and the average is 239 seconds; and for a request for 5000 sets of user data, the shortest is 324 seconds, the longest is 483 seconds, and the average is 457 seconds. Therefore, for a large number of user data requests, this application can achieve high concurrency processing efficiency.

[0105] Optionally, the remote sensing data processing method may further include: the remote sensing data processing system determining whether the target output result meets a second preset requirement. If the target output result does not meet the second preset requirement, the remote sensing data processing system adjusts the metadata information based on the determination result until the adjusted target output result meets the second preset requirement.

[0106] For example, a remote sensing data processing system can adjust the algorithm parameters corresponding to the metadata information based on the judgment result until the adjusted target output meets the second preset requirement.

[0107] For example, if the user's data requirement corresponds to the output data in the CGCS2000 coordinate system, but the algorithm parameter information of the remote sensing data processing system is set to the WGS84 coordinate system by default, the remote sensing data processing system will determine that the target output result does not meet the second preset requirement and will prompt the user to change the algorithm parameter information to meet the user's needs.

[0108] CGCS2000 coordinate system: China Geodetic Coordinate System 2000; WGS84 coordinate system: World Geodetic System 1984.

[0109] According to another aspect of this application, a remote sensing data processing system is also provided. Figure 10 A schematic diagram of the structure of a remote sensing data processing system according to an embodiment of this application is shown. Figure 10 As shown, the remote sensing data processing system 1 may include a data acquisition module 11, a data parsing module 12, a block processing module 13, and a data processing module 14.

[0110] According to the example embodiment, the data acquisition module 11 acquires the target remote sensing data of the target satellite constellation based on the received user data requirements.

[0111] For example, a user can input their data request (which can be understood as an online data processing order) into the remote sensing data processing system via an interactive page. The data acquisition module 11 can respond to this user data request by importing target remote sensing data collected by the target satellite constellation from at least one satellite manufacturer's data platform, thereby acquiring the target remote sensing data.

[0112] For example, user data requirements include, but are not limited to, data requirements in multiple dimensions such as geographical coverage, time range, required type and specifications, and target quality of remote sensing data.

[0113] For example, the target satellite constellation includes, but is not limited to, satellite constellations with remote sensing data acquisition capabilities such as the Gaofen constellation, Jilin constellation, resource series satellite constellation, environmental series satellite constellation, and radar satellite constellation.

[0114] The data parsing module 12 parses the sensor type corresponding to the target remote sensing data to determine at least one type of target remote sensing data.

[0115] For example, the sensor type refers to the type of payload equipment used by the target satellite constellation to acquire remote sensing data. Different sensor types will generate different types of remote sensing data. The remote sensing data processing system analyzes the sensor type of the acquired target remote sensing data and identifies the target remote sensing data as the corresponding target type remote sensing data.

[0116] For example, the types of payload equipment include, but are not limited to, spectroscopic cameras, synthetic aperture radars, and imaging spectrometers. Correspondingly, the target type remote sensing data includes, but is not limited to, payload data of types such as optical remote sensing data, microwave remote sensing data, and hyperspectral remote sensing data.

[0117] The data parsing module 12 determines the metadata information of the target type remote sensing data.

[0118] For example, the data parsing module 12 parses the metadata information of the target type remote sensing data, thereby enabling the accurate extraction of key parameters. Exemplarily, the metadata information includes, but is not limited to, data information such as spatial reference system, image row and column count, bands, data type, pixel depth, and time.

[0119] The block processing module 13 performs block processing on the target type remote sensing data according to the metadata information to obtain multiple sub-target type remote sensing data.

[0120] The data processing module 14 performs distributed parallel computation on remote sensing data of multiple sub-target types to obtain target output results.

[0121] For example, the block processing module 13 can determine the quantity and size of the sub-target type remote sensing data based on metadata information, thereby dividing the target type remote sensing data into multiple sub-target type remote sensing data. The block processing module 13 decomposes the ultra-large-scale remote sensing data that cannot be processed by a single machine into multiple sub-target type remote sensing data through block processing, and distributes the multiple sub-target type remote sensing data to multiple nodes in the distributed computing cluster for parallel processing. After all computing nodes have completed their calculations, the data processing module 14 merges the calculation results of all computing nodes into the final output (i.e., the target output result).

[0122] For example, the target output results include, but are not limited to, diverse satellite results including raw data, standard products, orthophoto products, advanced products, and application products, and this application does not limit them.

[0123] It is understandable that ultra-large-scale remote sensing data (i.e., the total number of pixels in the remote sensing data) is characterized by its large size and high resolution. For processing ultra-large-scale remote sensing data, single-machine processing methods struggle to overcome the limitations of memory and processing power. This application, through distributed parallel processing of sub-target type remote sensing data, employs a "block-computation-merging" computational logic. Based on spatial indexing and a parallel computing framework, it can achieve efficient processing of ultra-large-scale remote sensing data.

[0124] According to the example embodiment, the data processing module 14 can perform distributed parallel computing on remote sensing data of multiple sub-target types based on the supercomputing cloud network to obtain the target output results.

[0125] For example, data processing module 14 can construct distributed parallel computing nodes based on supercomputing power (such as supercomputing processor architecture) and build high-performance computer storage that meets business scenarios. Then, data processing module 14 uses a job scheduling system (such as...) Figure 2 The Slurm job scheduling system shown submits remote sensing data of multiple sub-target types to a supercomputing cloud network formed by distributed parallel computing nodes for parallel computing.

[0126] For example, such as Figure 2 As shown, the Slurm job scheduling system can uniformly receive user data requests (i.e., user-submitted data tasks) based on multiple algorithm plugins, and dynamically schedule and distribute tasks according to the resource status of each computing node (such as computing node 1, computing node 2, and computing node 3), so that user requests can be allocated to various computing nodes. Furthermore, each computing node can provide real-time feedback on its job status (i.e., supercomputing job status information) to the Slurm job scheduling system, thus forming a closed-loop process from receiving user data requests, distributed scheduling to execution monitoring.

[0127] Optionally, the data processing module 14 can also feed back the target output results to the user in a preset format.

[0128] For example, the data processing module 14 can automatically upload the target output results to the user-specified storage space (such as a cloud drive), or the data processing module 14 can also provide a preview to the user based on a visual page.

[0129] For example, the data processing module 14 can publish and aggregate the target output results through geographic information services based on preset standards (such as OGC standards), thereby forming an interoperable service cluster. Furthermore, the data processing module 14 can also, based on a WebGL 3D earth engine or a 2D front-end library, achieve seamless overlay, real-time rendering, and multi-dimensional dynamic display of massive raster and vector data for the visualization of the target output results.

[0130] Optionally, the data processing module 14 can also support interactive querying of the target output results' metadata, spatial range, and spectral bands, and can provide data support such as time series analysis and roll-up comparison, thereby transforming the target output results into dynamic spatial information that can be explored and analyzed in depth.

[0131] Through the above embodiments, this application can acquire target remote sensing data of a target satellite constellation based on the received user data requirements, parse the sensor type corresponding to the target remote sensing data, and determine at least one target type of remote sensing data. Then, this application can determine the metadata information of the target type remote sensing data, perform block processing on the target type remote sensing data based on the metadata information to obtain multiple sub-target type remote sensing data, and perform distributed parallel computing on the multiple sub-target type remote sensing data to obtain the target output result.

[0132] This application enables online access and acquisition of remote sensing data by parsing user data requirements online, avoiding local downloads and offline processing. By migrating remote sensing data and algorithms to the cloud, it solves problems such as long processing times, low efficiency, and high local resource consumption costs. This application also addresses the processing of ultra-large-scale remote sensing data, which cannot be handled by a single machine, by segmenting remote sensing data into blocks. Furthermore, this application improves the processing efficiency and stability of ultra-large-scale remote sensing data through distributed parallel computing and a high-performance storage system. Finally, this application overcomes the limitations of data silos through standardized service interfaces and a unified data management mechanism, enabling collaborative analysis and resource sharing among multiple users and departments.

[0133] Through the above embodiments, this application can realize preprocessing operations such as data cropping, coordinate reprojection, image mosaicking, cloud removal and enhancement; it can realize intelligent analysis tasks such as land feature classification, target detection, change detection, and time series analysis; it can realize unified resource modeling and dynamic scheduling based on supercomputing cloud network; and it can also realize user-oriented visualization display and result service publishing.

[0134] This application can analyze user data requirements online and, combined with dynamic scheduling and automated execution mechanisms, construct a closed-loop remote sensing data processing system integrating "data acquisition, data processing, data analysis, data quality inspection, and data delivery." This application can meet the high-precision, high-efficiency, automated, and intelligent processing needs of various payload data, including optical remote sensing data, microwave remote sensing data, and hyperspectral remote sensing data. This application can adapt different data results to different target types of remote sensing data, thereby providing accurate aerospace big data results. This application can couple multiple types of computing power, such as intelligent computing, supercomputing, big data, and cloud computing, to improve the efficiency and intelligence level of remote sensing data processing.

[0135] This application enables end-to-end, one-stop automated processing of online remote sensing data orders. By integrating remote sensing data from multiple satellite constellations, it constructs a fully intelligent system covering data security management, intelligent quality inspection, cleaning and screening, processing, results archiving, and delivery. Relying on continuous feedback and status monitoring mechanisms, this application can automatically generate diverse satellite results, from raw data to advanced application products, meeting the data processing needs of multiple scenarios and payloads.

[0136] This application leverages artificial intelligence technology and a unified resource scheduling framework for multi-task parallel processing to automate the entire process of processing massive, multi-source, heterogeneous imagery data. It provides a one-stop service covering task-driven automation, data verification automation, data processing automation, intelligent analysis automation, output quality inspection automation, and output delivery automation. For example, the task-driven module can monitor and automatically push target remote sensing data in real time. After the data verification module completes integrity verification and quality checks, the data processing module generates standardized data. Subsequently, the intelligent analysis module produces target output results based on deep learning algorithms. After passing dual verification by the output quality inspection module, the results are finally intelligently uploaded to a designated delivery cloud drive via the output delivery module. The various functional modules in this application can collaborate through message queues to form a closed-loop remote sensing data processing system integrating "data acquisition - data processing - data analysis - data quality inspection - data delivery."

[0137] Optionally, the data acquisition module 11 acquires satellite cataloging data from at least one satellite manufacturer.

[0138] For example, satellite cataloging data serves as an index describing the core characteristics and status of remote sensing data. Satellite cataloging data includes at least the remote sensing data ID, coverage information, time information, cloud cover information, and storage location information.

[0139] The data acquisition module 11 can automatically import satellite cataloging data from the data platform of at least one satellite manufacturer through a standardized data interface, and can also complete the aggregation and preview of satellite cataloging data.

[0140] The data acquisition module 11 determines the corresponding remote sensing data parameter information based on the user's data requirements. The data acquisition module 11 then acquires the target remote sensing data based on the remote sensing data parameter information and satellite cataloging data.

[0141] For example, remote sensing data parameters include, but are not limited to, time parameters, coverage parameters, and cloud cover information.

[0142] Upon receiving a user's data request, the data acquisition module 11 automatically parses the request into specific remote sensing data parameter information. Based on this parameter information, it determines the most suitable target satellite constellation. Then, using previously acquired and aggregated satellite catalog data, the data acquisition module 11 performs rapid queries and matching in the database based on the remote sensing data parameter information to obtain the required target remote sensing data.

[0143] Through the above embodiments, this application can automatically extract multi-source satellite catalog data, convert structured user data requirements into accurate remote sensing data parameter information, and perform analysis and matching based on remote sensing data parameter information and satellite catalog data, thereby achieving accurate acquisition of target remote sensing data.

[0144] Optionally, such as Figure 10 As shown, the remote sensing data processing system 1 may also include a data quality inspection module 15.

[0145] The data quality inspection module 15 determines whether the target remote sensing data meets the first preset requirement. If the target remote sensing data does not meet the first preset requirement, the data quality inspection module 15 replaces the target remote sensing data according to the user's data requirements until the replaced target remote sensing data meets the first preset requirement.

[0146] For example, the first preset requirement can be a data requirement for the target remote sensing data that is customized according to user needs. The first preset requirement includes, but is not limited to, data integrity requirements, imaging quality requirements, and specification requirements.

[0147] For example, data integrity can include whether the data is complete and whether it is corrupted. Imaging quality requirements can include cloud cover (e.g., for optical remote sensing data, whether the cloud cover is below a preset threshold), noise level (e.g., whether there is noise), and data quality (e.g., whether it is stable or whether there is distortion). Specification requirements can include whether the resolution, number of bands, etc., meet the user's data needs.

[0148] For example, the data quality inspection module 15 can perform quality inspection on the acquired target remote sensing data based on a first preset requirement. If the target remote sensing data does not meet the first preset requirement, the remote sensing data processing system can call the user data request again, change the data source, until a replacement target remote sensing data that meets the preset requirement is determined.

[0149] Through the above embodiments, this application can perform quality inspection on the acquired target remote sensing data based on the first preset requirements. When the target remote sensing data is unqualified, a feedback loop can be performed based on the quality inspection results. By dynamically adjusting the input data, the reliability of the finally acquired target remote sensing data can be ensured.

[0150] Optionally, the block processing module 13 can perform different calculations on different bands of the target type remote sensing data, and then divide the target type remote sensing data into blocks according to the bands to obtain multiple sub-target type remote sensing data.

[0151] For example, when the user's data requirement is to perform different calculation logic on different bands of remote sensing data of the target type, such as image radiometric calibration (different gain / bias coefficients for each band), enhancement of specific bands (such as enhancement of the green band), band calculation, etc., the block processing module 13 will divide the remote sensing data of the target type into blocks according to the bands.

[0152] For example, such as Figure 5 As shown (different colors represent different bands), the block processing module 13 traverses the bands of the target type remote sensing data. For the currently processed band, the block processing module 13 divides it according to preset dimensions in a spatial dimension, generating multiple sub-target type remote sensing data containing only that band. Then, the block processing module 13 distributes these single-band sub-target type remote sensing data to distributed computing nodes for parallel computation. After the computation is complete, the results are written back to the corresponding band and spatial location in sequence.

[0153] Through the above embodiments, this application can segment remote sensing data of a target type according to its bands when performing different calculations on different bands. This configuration allows the application of different parameters to the corresponding bands during data processing, thereby ensuring the accuracy of band-specific processing. Furthermore, the segmentation method, which only requires loading data blocks of a single band, reduces the instantaneous demand on single-machine memory.

[0154] Optionally, when performing the same calculations on different bands of the target type remote sensing data, the block processing module 13 divides the target type remote sensing data into blocks as a whole to obtain multiple sub-target type remote sensing data.

[0155] For example, when the user's data requirement is to perform the same calculation logic on different bands of the target type remote sensing data, the block processing module 13 performs block processing according to the overall target remote sensing data.

[0156] For example, such as Figure 6As shown (different colors represent different bands), the block processing module 13 can divide the entire target type remote sensing data (including the entire image data of all bands) into several regular sub-target type remote sensing data covering all bands according to a preset size, based on the spatial range. Then, the block processing module 13 distributes these multi-band sub-target type remote sensing data to distributed computing nodes for parallel computing. Each computing node performs a unified operation on all bands within the sub-target type remote sensing data and writes back the complete result data block.

[0157] Through the above embodiments, this application can divide the target type remote sensing data into blocks for overall processing while performing the same calculations on different bands of the target type remote sensing data. This configuration allows all band data within a spatial region to be assigned to a single computing node at once. This computing node can perform unified vectorization or parallel computation on all pixels (regardless of band) of the sub-target type remote sensing data without switching between bands, maximizing spatial locality and improving processing efficiency.

[0158] Optionally, when there are multiple user data requests, the data processing module 14 can distribute the multiple user data requests to multiple computing clusters (or servers) based on load balancing.

[0159] For example, such as Figure 7 As shown, when the remote sensing data processing system receives multiple user data requests, the data processing module 14 can distribute the multiple user data requests to different servers (server A, server B, server C) based on load balancing.

[0160] Through the above embodiments, this application can distribute user data requests to different servers based on load balancing processing, and distribute the computing load evenly, thus avoiding the problem of overload on a single server.

[0161] Optionally, such as Figure 8 As shown, the data processing module 14 can optimize the batch task submission script, thereby reducing the time wasted due to serial task submission. The data processing module 14 can also distribute multiple user data requests to multiple computing clusters (such as HAProxy load balancers) based on load balancing processing (such as HAProxy load balancers). Figure 8 (See the iFactory and other computing nodes shown). The data processing module 14 can select the Least-Conn strategy (schedule with the fewest connections) to reduce cluster task waiting caused by some time-consuming tasks.

[0162] For example, such as Figure 8As shown, for application scenarios with five computing clusters, this application can significantly reduce computing time (e.g., in the case of 1000 order tasks, the time taken by a single cluster is optimized from 36 minutes to 1.5 minutes for five clusters).

[0163] For example, Figure 9 The chart shows a comparison of the time taken to process 500, 1000, 1500, 2000, and 5000 sets of requests concurrently in real time. Figure 9 As shown, this application can achieve high concurrency processing efficiency for large user data requirements.

[0164] Optionally, the data quality inspection module 15 determines whether the target output result meets the second preset requirement. If the target output result does not meet the second preset requirement, the remote sensing data processing system adjusts the metadata information based on the determination result until the adjusted target output result meets the second preset requirement.

[0165] For example, a remote sensing data processing system can adjust the algorithm parameters corresponding to the metadata information based on the judgment result until the adjusted target output meets the second preset requirement.

[0166] For example, if the user's data requirement corresponds to the output data in the CGCS2000 coordinate system, but the algorithm parameter information of the remote sensing data processing system is set to the WGS84 coordinate system by default, the remote sensing data processing system will determine that the target output result does not meet the second preset requirement and will prompt the user to change the algorithm parameter information to meet the user's needs.

[0167] According to another aspect of this application, an electronic device is also provided. The electronic device includes: one or more processors; and a storage device for storing one or more programs, which, when executed by the one or more processors, enable the one or more processors to perform the methods described above.

[0168] According to another aspect of this application, a non-volatile computer-readable storage medium is also provided. This storage medium stores a computer program that, when executed by a processor, can perform the methods described above.

[0169] According to another aspect of this application, this application also provides a computer program product. The computer program product includes: a computer program stored on a computer-readable storage medium; the computer program includes program instructions that, when executed by a computer, cause the computer to perform the methods described above.

[0170] Finally, it should be noted that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions of the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method of processing remote sensing data, characterized in that, include: Based on the received user data requirements, acquire the target remote sensing data of the target satellite constellation; The sensor type corresponding to the target remote sensing data is analyzed to determine at least one type of target remote sensing data; The metadata information of the remote sensing data of the target type is determined; The target type remote sensing data is segmented according to the metadata information to obtain multiple sub-target type remote sensing data, including: segmenting the target type remote sensing data according to bands when different calculations are performed on different bands of the target type remote sensing data to obtain the multiple sub-target type remote sensing data; or, segmenting the target type remote sensing data as a whole when the same calculations are performed on different bands of the target type remote sensing data to obtain the multiple sub-target type remote sensing data. Distributed parallel computation is performed on the remote sensing data of the multiple sub-target types to obtain the target output results; When there are multiple user data requests, load balancing is used to distribute these multiple user data requests across multiple computing clusters.

2. The remote sensing data processing method of claim 1, wherein, The process of acquiring target remote sensing data for the target satellite constellation based on received user data requests includes: Obtain satellite cataloging data from at least one satellite manufacturer; Based on the user's data requirements, the corresponding remote sensing data parameter information is determined; The target remote sensing data is obtained based on the remote sensing data parameter information and the satellite cataloging data.

3. The method of claim 1, wherein, The process of acquiring target remote sensing data for the target satellite constellation based on received user data requests includes: Determine whether the target remote sensing data meets the first preset requirement; If the target remote sensing data does not meet the first preset requirement, the target remote sensing data is replaced according to the user data requirements until the replaced target remote sensing data meets the first preset requirement.

4. The method of claim 1, wherein, The remote sensing data processing method further includes: Determine whether the target output meets the second preset requirement; If the target output does not meet the second preset requirement, the metadata information is adjusted based on the judgment result until the adjusted target output meets the second preset requirement.

5. A remote sensing data processing system, characterized in that, The remote sensing data processing system is used to execute the remote sensing data processing method as described in any one of claims 1-4, and the remote sensing data processing system includes: The data acquisition module acquires target remote sensing data of the target satellite constellation based on the received user data requirements; The data parsing module parses the sensor type corresponding to the target remote sensing data to determine at least one type of target remote sensing data, and to determine the metadata information of the target type remote sensing data; The block processing module performs block processing on the target type remote sensing data according to the metadata information to obtain multiple sub-target type remote sensing data, including: when different calculations are performed on different bands of the target type remote sensing data, the target type remote sensing data is block processed according to bands to obtain the multiple sub-target type remote sensing data; or, when the same calculations are performed on different bands of the target type remote sensing data, the target type remote sensing data is block processed as a whole to obtain the multiple sub-target type remote sensing data. The data processing module performs distributed parallel computing on the remote sensing data of the multiple sub-target types to obtain the target output results, and, in the case of multiple user data demands, balances the multiple user data demands to multiple computing clusters based on load balancing processing.

6. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the remote sensing data processing method as described in any one of claims 1-4.

7. A non-volatile computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the remote sensing data processing method as described in any one of claims 1-4.

Images