Data storage method and device, electronic equipment, storage medium and program product

Abstract

The application discloses a data storage method and device, electronic equipment, a storage medium and a program product, and relates to the technical field of remote sensing data processing. The method comprises the following steps: acquiring remote sensing embedded data stored in a first bit depth format, the data being composed of at least two time slices; analyzing the remote sensing embedded data to obtain a plurality of numerical components belonging to different semantic dimensions, the bit width of the components being smaller than the first bit depth; splicing at least two numerical components into a composite byte according to a preset semantic pairing relationship; gathering each composite byte of the same semantic dimension and the same time slice into a storage band organized in a second bit depth format, the second bit depth being smaller than the first bit depth; and packaging each storage band to obtain a target storage file. Through implementation of the technical scheme, high-bit-depth remote sensing embedded data can be reorganized into a low-bit-depth storage structure independent of semantic dimensions, information integrity is maintained, storage bit width is reduced, and flexible access according to semantic dimensions is realized.

Description

Technical Field

[0001] This application relates to the field of remote sensing data processing technology, specifically to data storage methods, devices, electronic devices, storage media, and program products. Background Technology

[0002] With the rapid development of satellite remote sensing technology, remote sensing data processing is entering a new stage centered on large-scale model representation. Raw data is mapped to a high-dimensional feature space via deep neural networks, forming high-dimensional embedded data consisting of at least two time slices. This embedded data is gradually becoming a key carrier for remote sensing information representation and downstream computational tasks. When storing this type of embedded remote sensing data, related technologies typically follow traditional cloud-optimized encapsulation approaches, organizing and storing the embedded data in its original first-order deep format, and then processing it using general compression algorithms.

[0003] However, the aforementioned storage methods have significant shortcomings when handling such embedded data. On the one hand, when embedded data is encapsulated in a first-bit-depth format, the inherent high-bit-depth nature of the data often makes effective compression difficult, resulting in a still large file size after encapsulation and wasting storage resources. On the other hand, in actual access, when only specific semantic dimension information from the embedded data is needed, the relevant storage methods, by using high-bit-depth data units as the smallest access granularity, force the system to acquire the entire high-bit-depth data and then extract the required portion. This results in a large amount of data unrelated to the current need occupying transmission bandwidth, significantly reducing access efficiency. Summary of the Invention

[0004] In view of this, this application provides a data storage method, apparatus, electronic device, storage medium, and program product to solve the problem of low data storage and access efficiency.

[0005] In a first aspect, this application provides a data storage method, comprising: acquiring remote sensing embedded data stored in a first-bit-depth format, the remote sensing embedded data consisting of at least two time slices; parsing the remote sensing embedded data to obtain multiple numerical components belonging to different semantic dimensions, the bit width of the numerical components being smaller than the first-bit depth corresponding to the remote sensing embedded data; concatenating at least two numerical components into composite bytes according to a preset semantic pairing relationship; aggregating the composite bytes belonging to the same semantic dimension and the same time slice into storage bands organized in a second-bit-depth format, the second-bit depth corresponding to the storage bands being smaller than the first-bit depth; and encapsulating each storage band to obtain a target storage file.

[0006] In some optional implementations, parsing remote sensing embedded data to obtain multiple numerical components belonging to different semantic dimensions includes: obtaining the energy level parameter sequence corresponding to the remote sensing embedded data; using the energy level parameter sequence, decomposing a first numerical component and at least one second numerical component from each pixel value of the remote sensing embedded data level by level; wherein, the multiple numerical components include a first numerical component and a second numerical component, and the first numerical component and the second numerical component belong to different semantic dimensions.

[0007] In some optional implementations, the preset semantic pairing relationship is the pairing relationship between the first numerical component and the second numerical component; according to the preset semantic pairing relationship, at least two numerical components are concatenated into a composite byte, including: performing bit-level concatenation on the first numerical component and the second numerical component that satisfy the preset semantic pairing relationship to obtain a composite byte.

[0008] In some optional implementations, encapsulating each storage band to obtain a target storage file includes: for any time slice, arranging each storage band within the time slice adjacently to obtain a slice band group corresponding to the time slice; arranging the slice band groups corresponding to each time slice in time sequence to obtain a time-series band sequence; and encapsulating the time-series band sequence to obtain the target storage file.

[0009] In some optional implementations, encapsulating each storage band to obtain a target storage file includes: arranging storage bands belonging to the same time slice in a band-interleaved pattern to obtain an arranged band sequence; dividing the arranged band sequence according to a preset size to obtain storage tiles corresponding to the band sequence; and compressing the storage tiles to obtain the target storage file.

[0010] In some alternative implementations, in response to a data request for a target semantic dimension in a target storage file, a storage band corresponding to the target semantic dimension is extracted from the target storage file.

[0011] Secondly, this application provides a data storage device, comprising: an acquisition module for acquiring remote sensing embedded data stored in a first-bit-depth format, the remote sensing embedded data consisting of at least two time slices; a parsing module for parsing the remote sensing embedded data to obtain multiple numerical components belonging to different semantic dimensions, the bit width of the numerical components being smaller than the first-bit depth corresponding to the remote sensing embedded data; a splicing module for splicing at least two numerical components into composite bytes according to a preset semantic pairing relationship; a convergence module for converging composite bytes belonging to the same semantic dimension and the same time slice into storage bands organized in a second-bit-depth format, the second-bit depth corresponding to the storage bands being smaller than the first-bit depth; and a packaging module for packaging each storage band to obtain a target storage file.

[0012] Thirdly, this application provides an electronic device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the data storage method of the first aspect or any corresponding embodiment described above.

[0013] Fourthly, this application provides a computer-readable storage medium storing computer instructions that cause a computer to perform the data storage method of the first aspect or any corresponding embodiment described above.

[0014] Fifthly, this application provides a computer program product, including computer instructions for causing a computer to execute the data storage method of the first aspect or any corresponding embodiment described above.

[0015] The data storage method provided in this application, for remote sensing embedded data consisting of at least two time slices and stored in a first-bit-depth format, decomposes the original high-bit-depth data into multiple low-bit-width numerical components belonging to different semantic dimensions through parsing operations. This breaks the strong coupling of multiple semantic information within a single wide-byte numerical unit, allowing each semantic dimension to be explicitly separated and expressed independently. Based on this, at least two numerical components are concatenated into composite bytes according to a preset semantic pairing relationship. This enables the compact reorganization of dispersed semantic information at the bit level, significantly improving the density and logic of data expression and effectively avoiding the occupation of invalid or redundant bits in storage space. Subsequently, composite bytes belonging to the same semantic dimension and the same time slice are aggregated into storage bands organized in a second-bit-depth format, where the second bit depth is less than the first bit depth. This process equivalently reconstructs high-bit-depth data into multiple independent low-bit-depth bands at the physical storage level, greatly reducing the storage volume of a single band and making the data organization structure more refined and lightweight. The final encapsulated target storage file achieves independent encapsulation of time slices and semantic dimensions, thereby supporting selective reading and access according to specific semantic dimensions. It eliminates the need to load the full amount of high-bit-depth data in order to obtain partial semantic information, greatly improving the flexibility and efficiency of data retrieval, while also reducing bandwidth overhead during data transmission. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the specific embodiments or related technologies of this application, the drawings used in the description of the specific embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a schematic flowchart of a data storage method according to an embodiment of this application; Figure 2 This is a schematic diagram of a second data storage method according to an embodiment of this application; Figure 3 This is a schematic diagram of bit semantic dimension decoupling and bit weaving reassembly according to an embodiment of this application; Figure 4 This is a structural block diagram of a data storage device according to an embodiment of this application; Figure 5 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. 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.

[0019] It should be noted that the information (including but not limited to user input information, such as information entered by the user into input boxes), data (including but not limited to data used for analysis, stored data, and displayed data, such as context code, all code of the current project, the service pressure corresponding to operations performed on all code of the current project, and the code development status of the current project), and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with relevant laws, regulations, and standards. For example, the context code, operations performed on all code of the current project, the corresponding service pressure, and the code development status involved in this application were all obtained with full authorization.

[0020] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0021] With the rapid development of satellite remote sensing technology, remote sensing data processing is entering a new stage centered on large-scale model representation. Remote sensing image processing has gradually shifted from traditional pixel analysis to high-dimensional feature representation based on deep learning. Raw data is mapped to a high-dimensional feature space via deep neural networks, forming high-dimensional embedded data consisting of at least two time slices. This embedded data has gradually become a key carrier for remote sensing information expression and downstream task computation. To support efficient access to massive, multi-scale remote sensing data in a cloud environment, large-scale front-end visualization rendering and analysis, and tile-free on-demand rendering and computation, the common approach in this field is to use Cloud Optimized GeoTIFF (COG) as the standard format for data storage and access.

[0022] When storing remote sensing embedded data, the traditional cloud-optimized GeoTIFF technology is used for data organization and storage. It typically encapsulates the embedded data in its original first-order deep format and performs direct cloud optimization processing in conjunction with general compression algorithms such as ZSTD and differential prediction technology.

[0023] However, traditional COG encapsulation technology suffers from significant compatibility issues and performance bottlenecks when dealing with this type of embedded data. First, COG encapsulation causes a significant increase in file size. To support fast multi-scale retrieval, COG requires the additional construction of tiles and pyramids. Since embedded data is typically stored in a wide bit-depth format, traditional COG drivers, when encapsulating such high dynamic range data, often fail to effectively utilize differential predictors, resulting in COG files that are more than twice the size of the original data. This optimization burden severely negates the original storage advantages of the embedded data. Second, the computational load for large-scale front-end rendering is excessive. When performing global-scale rendering based on WebGL, browsers need to parse large amounts of high-bit-depth data over the network. For the front-end, processing bit-depth data not only consumes higher memory bandwidth but also involves complex decoding logic during real-time semantic mapping and shading, easily leading to frame drops or memory overflows on low-end devices. Furthermore, the coupling between storage architecture and semantic dimensions limits on-demand retrieval efficiency. Current storage architectures forcibly encapsulate multiple numerical components belonging to different semantic dimensions into single-band high-bit-depth logical units, resulting in strong coupling of data in physical storage. However, terminal rendering scenarios typically only require access to data in specific semantic dimensions. This "logically compact, physically redundant" feature arrangement forces the front-end to retrieve the full amount of high-bit-depth data over the network. The lack of a decoupling mechanism to reconstruct high-bit-depth data into low-bit-depth, semantically independent storage units results in a large amount of invalid data unrelated to the current rendering task occupying limited transmission bandwidth, significantly increasing network I / O latency and weakening the large-scale data response capabilities in cloud-native environments.

[0024] The data storage method provided in this application, for high-bit-depth remote sensing embedded data, first parses it into multiple numerical components belonging to different semantic dimensions and with smaller bit widths, then concatenates them into composite bytes according to a preset semantic pairing relationship, and then aggregates composite bytes of the same semantic dimension and the same time slice into storage bands organized with a second bit depth (less than the first bit depth) and encapsulates them separately. This results in lower bit depth and narrower dynamic range of the final storage bands, reducing internal data redundancy, and making them easier to be effectively utilized by differential prediction and compression algorithms during cloud-optimized encapsulation, thereby mitigating the file size expansion caused by directly encapsulating high-bit-depth data. During front-end rendering, the browser can directly obtain and process these low-bit-depth storage bands without decoding high-bit-depth units, significantly reducing memory bandwidth consumption and decoding complexity, and alleviating the risk of frame drops and memory overflow during large-scale rendering. Moreover, since the data is physically aggregated into different storage bands independently according to semantic dimensions and time slices, it breaks the traditional layout of strong coupling of multi-dimensional semantics in high-depth single-band data. It can request only low-depth data of specific semantic dimensions and time slices according to actual needs, without being forced to download the full high-depth data mixed with irrelevant semantics. This reduces the amount of invalid data transmitted in the network, reduces I / O latency, and improves the on-demand retrieval and response efficiency of large-scale data in the cloud environment.

[0025] According to an embodiment of this application, a data storage method embodiment is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0026] This embodiment provides a data storage method that can be used in electronic devices, such as cloud storage servers and remote sensing data servers. Figure 1 This is a flowchart of a data storage method according to an embodiment of this application, such as... Figure 1 As shown, the process includes the following steps: Step S101: Obtain remote sensing embedding data stored in first-bit deep format, the remote sensing embedding data consisting of at least two time slices.

[0027] The first bit depth format refers to a higher bit depth storage format used in the original remote sensing embedding data. For example, it can refer to a wide-byte format like Uint16. It means that each pixel value in the data is represented and stored using 16 bits (2 bytes).

[0028] Remote sensing embedded data refers to high-dimensional feature representation data formed by mapping the original remote sensing images to a high-dimensional feature space after processing by a deep learning model (deep neural network).

[0029] A time slice refers to a complete data layer corresponding to a specific point in time in time-series remote sensing data. For example, a year's remote sensing embedding data may consist of 12 time slices, corresponding to January to December. Each time slice contains high-dimensional features of all spatial locations at that time.

[0030] Specifically, a dataset with a unique internal structure is received from the data storage system or upstream processing flow. This dataset is not the raw satellite imagery, but rather embedded data that has been abstracted and refined by a deep learning model. It inherently contains various high-level semantic features about the ground scene. This complex semantic information is encoded and encapsulated in a numerical structure with a relatively large bit depth (i.e., first-bit deep format), for example, a single value occupies 16 bits. Simultaneously, this dataset is continuous in the temporal dimension, consisting of a series of complete data layers (i.e., time slices) observed or generated at different points in time, stacked sequentially.

[0031] Step S102: Parse the remote sensing embedded data to obtain multiple numerical components belonging to different semantic dimensions. The bit width of the numerical components is smaller than the first bit depth corresponding to the remote sensing embedded data.

[0032] Semantic dimension refers to each independent feature direction or attribute with specific meaning in the high-dimensional feature space embedded in data. It can be understood as multiple abstract feature categories describing a geographic location, and these dimensions are coupled and stored in a wide-byte pixel value.

[0033] Numerical components refer to the independent numerical values ​​that are parsed and separated from the original, high-bit-depth pixel values, each belonging to a different semantic dimension. These components are the decoupled results, each representing an independent semantic feature, and their numerical range is relatively small, so they can be represented using fewer bits.

[0034] The first bit depth refers to the original bit depth used when storing remote sensing embedding data. The bit width of a numerical component refers to the number of binary bits required to represent a numerical component, which is significantly smaller than the first bit depth of the original remote sensing embedding data. For example, a raw Uint16 (16-bit deep) pixel value may be resolved into 6 low-bit-width components, such as 4 bits, 5 bits, etc.

[0035] Specifically, based on a deep understanding of the original data encoding rules, high-bit-depth numerical values ​​at each location in the remote sensing embedded dataset are decoded in reverse. That is, a large, complex numerical value is treated as a container containing multiple independent information fragments with different meanings, each requiring only a few bits to fully express. Through a series of data processing steps, these different semantic elements, originally tightly bound within a single wide byte, can be separated layer by layer, ultimately yielding a series of discrete numerical components representing specific semantic features.

[0036] Step S103: According to the preset semantic pairing relationship, at least two numerical components are concatenated into a composite byte.

[0037] Predefined semantic pairing relationships refer to a set of predefined rules that specify which semantic dimension numerical components should be paired and combined together.

[0038] A composite byte refers to a new, compact data unit formed by combining two or more low-bit-width numerical components into one byte (or other smaller storage unit) through bitwise operations (such as left shift, bitwise OR, etc.) according to a preset pairing relationship.

[0039] Specifically, based on predefined pairing logic (i.e., preset semantic pairing relationships), specific numerical components that should be processed collaboratively in terms of computational tasks or physical meaning are selected. Then, at the bit operation level, the selected two or more low-bit components are sequentially placed into different positions within a new, standard low-bit-depth storage unit through techniques such as shifting and merging. In this way, multiple independent semantic information units are compactly packaged within the capacity of this low-bit-depth storage unit. This splicing process integrates semantically related but physically separate information units into a composite information unit optimized for computation, laying the foundation for constructing compact data bands.

[0040] Step S104: Combine the composite bytes belonging to the same semantic dimension and the same time slice into storage bands organized in a second bit depth format, where the second bit depth of the storage band is less than the first bit depth.

[0041] The second bit depth format refers to a lower bit depth format used to store storage bands after reassembly. For example, it can refer to a compact byte format like Uint8. Each pixel value is represented using only 8 bits (1 byte), which is half the bit depth compared to the original Uint16, making it more suitable for WebGL textures and efficient GPU access.

[0042] Storage bands are the product of data recombination. They refer to a physically independent raster data layer (i.e., a band) formed by bringing together all composite bytes that belong to the same time slice and are generated by a specific pairing relationship. This layer is organized in a second-bit depth format.

[0043] The second bit depth refers to the lower bit depth value used in the storage band. It is a specific quantization of the second bit depth format, such as 8 bits.

[0044] Specifically, all generated composite bytes are traversed and categorized and aggregated based on two core attributes: first, their origin from a specific point in time (i.e., which time slice); and second, the type of paired semantic information they carry. All composite bytes that simultaneously satisfy these two conditions are collected and arranged to spatially reconstruct a complete and ordered data array. Because the numerical units in each array are low-bit-depth composite bytes, this array naturally becomes a physically independent storage band organized with a relatively small bit depth (i.e., second-bit-depth format).

[0045] Step S105: Encapsulate each storage band to obtain the target storage file.

[0046] The target storage file refers to the optimized data file generated after all storage bands have been packaged and stored according to a specific organization method. For example, it can refer to the final packaged GeoTIFF file for cloud storage. Its internal data structure has been completely reconstructed from the original high-bit-depth, semantically coupled bands to a low-bit-depth, semantically decoupled set of bit-plane bands. This file is the final product for cloud access and front-end rendering. Specifically, according to a preset data organization strategy, all storage bands are arranged holistically, ensuring that semantic information at the same time can be efficiently and centrally read in the physical storage layout, while maintaining a clear and orderly correspondence between data at different times. Based on this, cloud-optimized file packaging technology is further employed to process the arranged data in blocks and perform efficient compression encoding. Finally, all this optimized and compressed data, along with its internal directory index, spatial reference, and other metadata information, is packaged and written into a standard cloud-optimized file container, thereby generating the final target storage file that can be distributed, accessed, and rendered. This encapsulation process ensures that the reconstructed data structure not only saves space but also allows front-end applications to request and parse only the specific bands required for their tasks, enabling on-demand and efficient network data transmission.

[0047] The data storage method provided in this application, for remote sensing embedded data consisting of at least two time slices and stored in a first-bit-depth format, decomposes the original high-bit-depth data into multiple low-bit-width numerical components belonging to different semantic dimensions through parsing operations. This breaks the strong coupling of multiple semantic information within a single wide-byte numerical unit, allowing each semantic dimension to be explicitly separated and expressed independently. Based on this, at least two numerical components are concatenated into composite bytes according to a preset semantic pairing relationship. This enables the compact reorganization of dispersed semantic information at the bit level, significantly improving the density and logic of data expression and effectively avoiding the occupation of invalid or redundant bits in storage space. Subsequently, composite bytes belonging to the same semantic dimension and the same time slice are aggregated into storage bands organized in a second-bit-depth format, where the second bit depth is less than the first bit depth. This process equivalently reconstructs high-bit-depth data into multiple independent low-bit-depth bands at the physical storage level, greatly reducing the storage volume of a single band and making the data organization structure more refined and lightweight. The final encapsulated target storage file achieves independent encapsulation of time slices and semantic dimensions, thereby supporting selective reading and access according to specific semantic dimensions. It eliminates the need to load the full amount of high-bit-depth data in order to obtain partial semantic information, greatly improving the flexibility and efficiency of data retrieval, while also reducing bandwidth overhead during data transmission.

[0048] This embodiment provides a data storage method that can be used in electronic devices, such as cloud storage servers and remote sensing data servers. Figure 2 This is a flowchart of a data storage method according to an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps: Step S201: Obtain the remote sensing embedding data stored in first-order deep format. The remote sensing embedding data consists of at least two time slices. See details below. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0049] Step S202: Parse the remote sensing embedded data to obtain multiple numerical components belonging to different semantic dimensions. The bit width of the numerical components is smaller than the first bit depth corresponding to the remote sensing embedded data.

[0050] Specifically, step S202 includes: Step S2021: Obtain the energy level parameter sequence corresponding to the remote sensing embedded data.

[0051] The energy level parameter sequence refers to a specific set of parameters defined in the pre-quantization encoding stage of remote sensing embedding data. It specifies the encoding span and cardinality of different semantic dimensions in the original high-bit-depth pixel values. Specifically, when acquiring remote sensing embedding data, a numerical matrix is ​​not obtained in isolation; instead, a set of parameters, which is fixed during the data generation stage, is simultaneously acquired or read from the dataset's metadata. This parameter sequence precisely describes how the original high-bit-depth numerical values ​​couple multiple semantic dimensions together, specifically defining the bit range, encoding cardinality, and hierarchical progression relationship occupied by each semantic dimension within the numerical value. Therefore, the essence of acquiring the energy level parameter sequence is to access and parse the encoding scheme definition associated with the source of the remote sensing embedding data, thereby accurately grasping all the mathematical rules required for subsequent reverse decomposition.

[0052] The obtained energy level parameter sequence is not arbitrarily set in terms of cardinality and span step size, but is strictly consistent with the energy level parameters defined in the pre-quantization encoding stage of the input remote sensing embedding data. Therefore, the subsequent decomposition process is not a general numerical splitting operation, but a targeted reconstruction process that maintains a strict correspondence with the original semantic encoding structure.

[0053] In a specific embodiment, the energy level parameter sequence specifically defines a set of cardinality sequences used during quantization encoding, such as 8, 8, 8, 5, 5, 5, etc., as well as the encoding span and hierarchical progression relationship corresponding to each semantic dimension. The corresponding span step sizes include 8, 64, 512, 2560, 12800, etc. These parameters are strictly aligned with the energy level parameters in the remote sensing embedding data generation stage to ensure that the numerical components obtained by decomposition are completely consistent with the original semantic structure.

[0054] Step S2022: Using the energy level parameter sequence, the first numerical component and at least one second numerical component are decomposed step by step from each pixel value of the remote sensing embedded data.

[0055] Among them, multiple numerical components include a first numerical component and a second numerical component, and the first numerical component and the second numerical component belong to different semantic dimensions.

[0056] A pixel value refers to a single high-bit depth numerical value at each spatial location in remote sensing embedded data. It is the smallest operational unit of the data. For example, in a Uint16 format band, each pixel value is a 16-bit integer that couples information from multiple semantic dimensions.

[0057] The first and second numerical components refer to the specific numerical components belonging to different semantic dimensions, which are decomposed step by step from the pixel value using the energy level parameter sequence.

[0058] Specifically, each high-bit-depth pixel value in the remote sensing embedded data is traversed and used as the original input to be decomposed. Using a pre-acquired energy level parameter sequence, each pixel value is decomposed using a step-by-step extraction method: at the first level of decomposition, through a combination of modulo and integer division operations, the first low-bit-width independent semantic feature value is separated from the original pixel value based on the first set of energy level parameters; this is the first numerical component. Subsequently, the remaining intermediate values, excluding the influence of the first component, are used as new processing objects, and decomposition is continued using the next set of energy level parameters to extract the subsequent second numerical components. This process progresses layer by layer according to the length of the energy level parameter sequence until all implicit semantic dimensions are extracted individually, forming a set of physically independent numerical components with extremely small bit widths. The above decomposition process is not a general numerical splitting operation, but a directional reconstruction process that maintains a strict correspondence with the original semantic coding structure. Each numerical component obtained from the decomposition corresponds to an independent latent semantic dimension, which can completely preserve the original semantic information without any loss of precision.

[0059] For example, when remote sensing embedding data incorporates 6 latent semantic dimensions to Coupled to a Uint16 cell value In the middle stage, using the cardinality and hierarchical relationship provided by the energy level parameter sequence, six numerical components can be decomposed step by step according to the following formula:

[0060]

[0061]

[0062]

[0063]

[0064]

[0065] In this process, the divisors used in each step—such as 8, 64, 512, 2560, and 12800—as well as the modulo radix sequence of 8, 8, 8, 5, 5, 5, are all predefined by the energy level parameter sequence. These parameters strictly correspond to the energy level parameters defined in the pre-quantization encoding stage of the input embedded data, thus ensuring lossless reconstruction of the semantic components. Through this inverse mapping, the original pixel value is mapped to a set of low-bit-depth, semantically relatively independent feature components, completely eliminating high-bit redundancy within the original wide byte.

[0066] The data storage method provided in this application introduces an energy level parameter sequence consistent with the original coding system, providing a precise inverse mapping basis for the parsing process. This ensures that the decomposition of high-bit-depth remote sensing embedded data is not arbitrary, but strictly follows the structural decoupling operation defined by semantic dimensions. Using this energy level parameter sequence, each pixel value is decomposed level by level into a first numerical component and at least one second numerical component belonging to different semantic dimensions. This allows for the deterministic systematic extraction of multiple independent and low-bit-width semantic components from a single wide-byte value, ensuring that the parsing process corresponds to the original quantization coding logic. Thus, at the physical level, lossless and ordered decomposition of information in each semantic dimension is achieved.

[0067] Step S203: According to the preset semantic pairing relationship, at least two numerical components are concatenated into a composite byte.

[0068] Specifically, the preset semantic pairing relationship is the pairing relationship between the first numerical component and the second numerical component. The above step S203 includes: performing bit-level concatenation on the first numerical component and the second numerical component that satisfy the preset semantic pairing relationship to obtain a composite byte.

[0069] The system has predefined rules for pairing numerical components with different semantic dimensions. These rules specify which independent components need to be combined into a cooperating composite unit in the new data structure. In practice, for each pair of first and second numerical components that satisfy the pairing relationship, spatial rearrangement is performed at the bit level. Specifically, a new, smaller-width storage unit is created. Then, one component is left-shifted into the higher-order segment of this unit, and the other component is bitwise ORed into the lower-order segment. In this way, the two originally separate numerical components are seamlessly and compactly merged into a new, low-bit-depth storage unit—the composite byte. By sequentially arranging feature components with different semantic dimensions through bit-level concatenation, a composite representation can be formed within a single byte, achieving a compact expression of multiple semantic information in low-bit-depth space and effectively improving bit utilization efficiency.

[0070] The data storage method provided in this application establishes a clear and deterministic basis for semantic dimension reorganization by explicitly defining the preset semantic pairing relationship as the pairing relationship between the first and second numerical components. This allows the originally disassembled heterogeneous semantic components to be paired in an orderly manner according to preset logic, avoiding semantic mismatch problems that may be caused by random combinations. Furthermore, only the first and second numerical components within this pairing relationship are concatenated at the bit level to obtain a composite byte. This targeted concatenation method can directly fuse information from two specific semantic dimensions within a single byte space, thereby achieving efficient co-location and compact expression of heterogeneous semantic information on physical storage units while maintaining the low-bit attributes of each semantic component.

[0071] In some optional implementations, the preset semantic pairing relationship can be as follows: the six semantic components obtained from the above decomposition are paired in pairs to form three independent bit-plane channels, respectively named Alpha, Beta, and Gamma. The bit-level concatenation operation is performed according to the following formula to obtain their respective composite bytes:

[0072]

[0073]

[0074] The numerical calculation process for the above decomposition and bit weaving recombination can be referred to accordingly. Figure 3 Examples are shown in the text. Figure 3 This diagram illustrates bit semantic dimension decoupling and bit weaving reconstruction. The left side shows the visualization results of remote sensing embedded data tensors at different spatial scales, specifically embedded images at dimensions (13, 3600, 3600) and (13, 512, 512), where 13 corresponds to the number of physical bands in the data. The progressively enlarging yellow boxes show the hierarchical correspondence from the overall global image to local areas, and then to the temporal values ​​of individual pixels, intuitively demonstrating the data processing objects and multi-scale adaptability of this scheme. The right side of the diagram shows the complete bit-level reconstruction operation flow for temporal pixels at a single spatial location: the top first lists... to The original pixel values ​​corresponding to 12 time slices are all stored in Uint16 format, forming a temporal embedding feature sequence for this spatial location, and are presented in... Taking the pixel value of 60678 at time 60678 as an example, the subsequent calculations are demonstrated; then, through a step-by-step parsing logic combining modulo operation and floor division, the original Uint16 pixel value is decoupled into... to There are a total of 6 low-bit-width discrete semantic feature components, each of which is simultaneously labeled with a decimal value and its corresponding binary representation, completing the physical separation of multiple semantic dimensions; based on this, through ( 3)| The bit-weaving operation rules concatenate the six semantic components in pairs according to a preset pairing relationship, and reassemble them into three composite bytes, Alpha, Beta, and Gamma, stored in Uint8 format. The figure simultaneously marks the binary code and decimal value of each composite byte, showing the compact arrangement of multiple semantic information within a single byte. At the bottom of the figure, the final temporal organization method is presented in the form of a cube group, that is, each time slice corresponds to a set of independent data units composed of three bit planes, Alpha, Beta, and Gamma. The time slices are arranged sequentially according to time, forming a bit plane storage structure with physical isolation in the semantic dimension and continuous and ordered in the temporal dimension.

[0075] This process does not change the discrete value attribute of the original semantics, and realizes the equivalent reconstruction of a single Uint16 band into three physically independent compact Uint8 bands. Through the above bit weaving, three composite bytes expressed in Uint8 format will be generated on each time slice. They belong to three orthogonal semantic dimensions. The reconstructed bit plane data structure has higher bit utilization efficiency and can directly participate in the subsequent storage and graphics processing in the form of standard byte stream, laying the foundation for subsequent independent band convergence.

[0076] Step S204: Combine the composite bytes belonging to the same semantic dimension and the same time slice into storage bands organized in a second bit depth format, where the second bit depth of the storage band is less than the first bit depth.

[0077] Specifically, step S204 includes: Step S2041: According to the spatial location of each composite byte in the remote sensing embedded data, the composite bytes belonging to the same semantic dimension and the same time slice are aggregated into a pixel array.

[0078] Spatial location refers to the row and column coordinates of the original two-dimensional remote sensing image raster to which each composite byte belongs in the remote sensing embedded data.

[0079] A pixel array refers to a two-dimensional data grid formed by arranging all composite bytes belonging to the same semantic dimension and the same time slice according to their respective spatial positions.

[0080] Specifically, during the previous decomposition and stitching process, each generated composite byte carries an immutable attribute: its original row and column coordinates in the original remote sensing image—its spatial location information. During aggregation, a blank two-dimensional grid frame is created that perfectly matches the spatial extent of the original remote sensing image. Then, all generated composite bytes are traversed, undergoing a dual filtering process based on their time slice identifier and semantic dimension identifier. Once a composite byte meets the conditions of belonging to the same time slice and the same semantic dimension, its spatial location coordinates are read and filled into the corresponding row and column intersection in that two-dimensional grid frame. Once all qualified composite bytes have been placed, the grid is filled, forming a spatially ordered array of pixels.

[0081] Step S2042: The pixel array is determined as a storage band organized in a second bit-depth format.

[0082] Because each grid cell constituting the aforementioned cell array no longer stores the original high-bit-depth cell value, but rather a recombined, smaller composite byte, the overall data precision and bit-depth attributes of the array have fundamentally changed. This complete cell array is treated as an independent raster data layer, and a new data type identifier (i.e., a lower bit-depth format matching the bit width of the internal composite byte) is assigned to it. Essentially, this operation formally registers a set of composite bytes conforming to specific spatial and semantic affiliations within the system as a physically independent standard raster band defined in the new format, thus qualifying it as an independent data unit for all subsequent operations.

[0083] Step S205: Encapsulate each storage band to obtain the target storage file.

[0084] In some optional implementations, step S205 above includes: Step a1: For any time slice, arrange the stored bands within the time slice adjacently to obtain the slice band group corresponding to the time slice.

[0085] A slice band group refers to a set of data formed by arranging the stored bands within a given time slice adjacent to each other. Specifically, for a particular time slice, after all the preceding operations, several physically independent stored bands, each representing different composite semantics, are generated. At this stage, the system uses the time slice as a container, treating all the stored bands within that time slice as an indivisible whole. During the data stream organization process before writing to the file, the order of these bands is adjusted, arranging them closely together one after another without inserting bands from other times. This collection of all physically adjacent stored bands belonging to the same time moment is thus organized into the slice band group for that time slice, ensuring that the complete semantic information of the same moment is also centralized on the storage medium. This arrangement method is also called a time-interleaved layout, which can guarantee the continuity of multi-dimensional semantic information at the same moment in physical storage and improve access efficiency during continuous reading.

[0086] For example, for each time slice, the three storage bands Alpha, Beta, and Gamma are arranged closely in adjacent order to form the slice band group for that time slice.

[0087] Step a2: Arrange the slice band groups corresponding to each time slice in chronological order to obtain the time-series band sequence.

[0088] A time-series band sequence refers to the overall band sequence formed by arranging the band groups corresponding to each time slice in their own chronological order. Specifically, the system has already organized a tightly structured band group for each time slice. Now, these groups need to be arranged according to the natural chronological order of the time points they represent. The time identifier associated with each band group is read, and these groups are concatenated from beginning to end according to a linear ascending order of time. The earliest band group is placed at the beginning, and the latest band group is placed at the end, thus forming a total band sequence that spans the entire time dimension and is logically and physically ordered, namely the time-series band sequence, which is ready for final writing.

[0089] Step a3: Encapsulate the time-series band sequence to obtain the target storage file.

[0090] The pre-arranged time-series band sequences are used as the effective data payload to be written, supplemented with metadata information for correct reading and parsing in cloud environments and various software. This includes geospatial reference coordinates, projection information, transformation matrices, and the file's internal directory index. During encapsulation, the geographic coordinate transformation parameters and spatial reference information of the original remote sensing embedded data are fully inherited, ensuring complete spatial consistency between the reconstructed data and the original data. Simultaneously, the Image File Directory (IFD) is placed at the beginning of the file to support high-performance range reading in cloud environments. Subsequently, a standard cloud-optimized raster file driver is invoked to encode and write these band sequence data and metadata together according to a predefined cloud-optimized file format specification, generating a single, self-describing file entity that supports efficient range reading. This final generated entity is the target storage file available for cloud distribution and client access. For valid data mask bands attached to the original data, their original bit-depth format is retained and encapsulated together with the reconstructed bit-plane bands according to the above rules, ensuring the complete transmission of data validity identification.

[0091] In the above implementation, by clearly defining the encapsulation logic of the storage bands, a dual organizational system is constructed, with orderly layer-by-layer arrangement from within a time slice to between slices. First, within a single time slice, the storage bands at the same moment are arranged adjacently to form a slice band group, ensuring that all semantic information at a point in time is tightly aggregated in physical storage, eliminating the discrete addressing overhead that may occur when reading across time. On this basis, the slice band groups corresponding to each time slice are arranged sequentially according to time order to form a time-series band sequence, so that the overall data structure presents a linear progressive characteristic in the time dimension that is completely consistent with the original acquisition order. This provides a high degree of physical continuity and spatial locality for subsequent range access or streaming reading by time series, which can significantly improve I / O efficiency in time-series data access mode.

[0092] In some optional implementations, step S205 above includes: Step b1: Arrange the stored bands belonging to the same time slice in a band-interleaved pattern to obtain the arranged band sequence.

[0093] Band interleaving mode refers to a specific band arrangement method that organizes stored bands belonging to the same time slice in an alternating pattern.

[0094] The arranged band sequence refers to the order of the data stream that is finally used for the next step of partitioning and compression after all the stored bands are rearranged according to the band interleaving pattern.

[0095] Specifically, when forming the arranged band sequence, instead of storing all data of the same band continuously, it is organized alternately in units of pixel rows or preset tiny pixel blocks. Specifically, all stored bands belonging to the same time slice are traversed simultaneously, and in each round of writing, a small data unit from each band is extracted and arranged intermittently. This ensures that composite bytes belonging to different semantic bands but located in the same spatial position are extremely close together in the final physical storage sequence, greatly enhancing the pattern representation of the data within a local range and providing better input for subsequent compression algorithms.

[0096] Step b2: Divide the arranged band sequence according to the preset size to obtain the storage tiles corresponding to the band sequence.

[0097] The preset size refers to a fixed pixel range that is pre-defined when dividing the arranged band sequence. This size determines the spatial coverage of the final generated storage tiles and is a fundamental parameter for implementing block access in cloud optimization files. For example, in practical applications, a large tile size of 512 pixels can be used to reduce the additional overhead of tile indexing while ensuring access flexibility.

[0098] Storage tiles refer to individual, regularly shaped data blocks obtained by dividing an arranged band sequence according to a preset size. Each tile contains data for all bands within its pixel block area.

[0099] Specifically, a pre-defined, fixed-size pixel window (i.e., a preset size) is used as a segmenter to regularly divide the entire spatial range covered by the pre-arranged band sequence. Starting from the spatial origin of the data, this spatial range is cut into a series of equal-area, non-overlapping rectangular grids. All data stream segments spanning all bands within each grid are extracted and stored as independent data blocks; these encapsulated data blocks are called storage tiles. This mechanism allows the system to directly locate and read a few corresponding tiles based on the requested spatial range, without parsing the entire file.

[0100] Step b3: Compress the storage tiles to obtain the target storage file.

[0101] For each partitioned storage tile, a specific, efficient, lossless compression algorithm is selected, and the binary data stream within the tile is encoded and compressed. Because the preceding band interleaving arrangement already gives the data within the tile greater local regularity, the compression algorithm can more effectively identify redundant patterns, thus achieving a higher compression ratio. The compressed tile data, along with the tile's spatial index information, is written to a file. Finally, a file package containing all compressed tiles, tile spatial indexes, and complete metadata information is encapsulated into a standard cloud-optimized format; this final output is the target storage file.

[0102] Optionally, during the compression process, the ZSTD lossless compression algorithm can be employed, and a differential predictor optimized for raster data can be enabled. This predictor can pre-encode the numerical correlation between adjacent cells, enabling the ZSTD algorithm to more effectively capture data redundancy, thereby achieving a higher compression ratio and further suppressing the volume expansion of the cloud-native packaging process.

[0103] In the above implementation, by arranging the storage bands within the same time slice in a band-interleaved pattern, physical interweaving of multi-semantic dimension data at the same moment is achieved. This allows heterogeneous semantic information belonging to the same time point to be stored alternately and compactly on the storage medium, thereby significantly enhancing the data locality at that time point and providing a centralized access unit for subsequent processing. Based on this, the arranged band sequence is further divided into storage tiles according to a preset size, and then the tiles are compressed to obtain the target storage file. This process organizes the physically densely interwoven semantic data into regular data blocks of fixed size, enabling the compression algorithm to more fully capture and utilize the data redundancy and statistical regularities between adjacent heterogeneous semantics within each tile. This effectively improves compression performance and suppresses storage volume expansion without loss.

[0104] In some alternative implementations, in response to a data request for a target semantic dimension in a target storage file, a storage band corresponding to the target semantic dimension is extracted from the target storage file.

[0105] The target semantic dimension refers to the specific semantic feature that a front-end application or user requests data from a target storage file.

[0106] A data request refers to a network access request initiated by a client (such as a front-end WebGIS application) to an electronic device for specific data content in a target storage file.

[0107] Specifically, when an external data request arrives, it explicitly specifies the precise temporal and semantic location of the required data, i.e., which target semantic dimension it needs. Upon receiving the request, instead of reading the entire target storage file, the system first parses the request to determine the corresponding target semantic dimension label. Next, it consults the file's internal directory index to directly locate the specific storage band corresponding to the target semantic dimension, completely ignoring other storage bands in the same file related to other semantic dimensions. Subsequently, only the compressed data of the requested band is extracted from the file, decompressed, and a clean, low-bit-depth data stream is returned to the requester, thus truly minimizing the amount of data transmitted.

[0108] In the above implementation, based on the prior technical premise of organizing information of different semantic dimensions into independent storage bands, a precise on-demand response mechanism is established. When a data request for a specific target semantic dimension is received, it is not necessary to read and transmit the full amount of data containing other irrelevant semantic dimensions. Instead, it is possible to directly extract the single storage band corresponding to that semantic dimension from the target storage file. This direct mapping extraction method completely avoids the transmission overhead and bandwidth occupation of irrelevant semantic data at the physical I / O level, achieving atomic-level precise supply of information of specific semantic dimensions, and significantly improving data access efficiency and system response speed.

[0109] In this embodiment, a typical scenario using Uint16 as the first bit depth format is illustrated. In this scenario, the original remote sensing embedded data is stored in a Uint16 high-bit deep format, with multiple semantic dimensions coupled within a single 16-bit pixel value. After processing by the method of this application, the data is equivalently reconstructed into multiple independently stored bands organized in a Uint8 low-bit deep format. At this point, if the front-end application only needs to obtain a specific target semantic dimension (e.g., the semantic components carried by the Alpha channel) for rendering, it only needs to initiate a request for that specific Uint8 band. Compared to traditional solutions that require obtaining the full Uint16 data and parsing it on the client side, this application achieves physical isolation of semantic granularity at the data source, ensuring that the amount of data transmitted is only a subset of the original data. Furthermore, the Uint8 format itself has half the bit width of Uint16, and the combination of these two factors significantly reduces the bandwidth consumption for on-demand transmission.

[0110] The technical effectiveness of files stored using the above method was verified through comparative experiments. The experiments, based on the same scene embedding data, constructed two sets of cloud-optimized storage encapsulation schemes for compression testing. The experimental data came from a publicly available remote sensing embedding dataset, which contained 13 physical bands: the first 12 were time-series feature index data in Uint16 format, corresponding to 12 consecutive time slices; the 13th band was an effective data mask record.

[0111] In the control group, the high-bit-depth (Uint16) wide-byte format of the original data was directly retained, and the underlying storage topology optimization strategies, such as band interleaving mode, differential predictor, and ZSTD deep compression, were simultaneously configured to construct a standard cloud-optimized GeoTIFF file. The experimental group, however, used the method described in this application to equivalently reconstruct the first 12 high-bit-depth bands into a compact Uint8 underlying feature tensor of shape (36, H, W). In terms of spatiotemporal mapping logic, the 36 independent bands of the physical layer precisely correspond to 12 consecutive latent time steps, meaning that the semantic features of each latent time step are jointly represented by three dedicated Uint8 bit-plane components: Alpha, Beta, and Gamma. In terms of spatial dimension, the originally mixed high-dimensional features of each pixel are physically "compacted," directly decomposed, and mapped to this three-dimensional orthogonal semantic vector composed of bit-plane components. During this bit-level remapping process, the original latent semantic information was strictly lossless, without any loss of precision or semantic change. The experimental group employed a temporally interleaved layout during the writing process, storing three bit-plane bands adjacently at the same time to ensure the continuity of semantic information in the physical layout. The COG driver was activated for file layout, placing the Image File Directory (IFD) at the beginning of the file to support high-performance range reading. Band interleaving mode and large 512-pixel tiles were configured, along with ZSTD compression and differential prediction technology. Both schemes used consistent underlying compression and layout optimization strategies to ensure the fairness of the comparative experiment.

[0112] Experimental results show that, compared to traditional cloud-optimized GeoTIFF directly encapsulated in Uint16 high-bit-depth format, the size of the Uint8 target storage file generated by this application is significantly reduced under the same compression parameters. Experiments verify that the lossless compression performance of this application consistently outperforms traditional wide-byte direct encapsulation schemes at different data scales, reducing the file size by 3.9% to 7.3% (as shown in Table 1), effectively verifying the technical superiority of this application in suppressing the expansion of cloud-native data volume. Furthermore, with the increase in image scale and total data volume, the compression benefits brought by the bit-plane reconstruction scheme exhibit a non-linear amplification trend. This is because the dense byte stream reconstructed by this application, with the increase in the global pixel sample size, allows compression engines such as ZSTD to more accurately and efficiently capture the statistical regularities of latent semantics in space and time, thereby further improving compression performance. Meanwhile, since the file is organized in low-bit-depth, semantically independent bands such as Uint8, the data can be directly read by the texture unit of the front-end graphics processing unit without additional bit operations or dynamic decoding, which significantly reduces the memory bandwidth usage and computational complexity, and improves the rendering efficiency and system responsiveness in large-scale, high-concurrency scenarios.

[0113] Table 1

[0114] This embodiment also provides a data storage device for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0115] This embodiment provides a data storage device, such as... Figure 4 As shown, it includes: The acquisition module 401 is used to acquire remote sensing embedding data stored in first-bit deep format, the remote sensing embedding data consisting of at least two time slices; The parsing module 402 is used to parse the remote sensing embedded data to obtain multiple numerical components belonging to different semantic dimensions. The bit width of the numerical components is smaller than the bit depth of the corresponding remote sensing embedded data. The splicing module 403 is used to splice at least two numerical components into a composite byte according to a preset semantic pairing relationship; The aggregation module 404 is used to aggregate composite bytes belonging to the same semantic dimension and the same time slice into storage bands organized in a second bit depth format, wherein the second bit depth of the storage band is less than the first bit depth. The encapsulation module 405 is used to encapsulate each storage band to obtain the target storage file.

[0116] In some alternative implementations, the parsing module 402 includes: The acquisition submodule is used to acquire the energy level parameter sequence corresponding to the remote sensing embedded data; The decomposition submodule is used to decompose each pixel value of the remote sensing embedded data into a first numerical component and at least one second numerical component by utilizing the energy level parameter sequence. Among them, multiple numerical components include a first numerical component and a second numerical component, and the first numerical component and the second numerical component belong to different semantic dimensions.

[0117] In some optional implementations, the preset semantic pairing relationship is the pairing relationship between the first numerical component and the second numerical component; the splicing module 403 includes: The splicing submodule is used to perform bit-level splicing on the first and second numerical components that satisfy the preset semantic pairing relationship to obtain a composite byte.

[0118] In some alternative implementations, the encapsulation module 405 includes: The first arrangement submodule is used to arrange the stored bands within any time slice adjacently to obtain the slice band group corresponding to the time slice for any given time slice. The second arrangement submodule is used to arrange the slice band groups corresponding to each time slice in time sequence to obtain the time band sequence. The encapsulation submodule is used to encapsulate time-series band sequences to obtain the target storage file.

[0119] In some alternative implementations, the encapsulation module 405 includes: The arrangement submodule is used to arrange the stored bands belonging to the same time slice in a band staggered pattern to obtain the arranged band sequence. The partitioning submodule is used to partition the arranged band sequence according to a preset size to obtain the storage tiles corresponding to the band sequence; The compression submodule is used to compress storage tiles to obtain the target storage file.

[0120] In some alternative implementations, the data storage device further includes: The extraction module is used to extract the storage band corresponding to the target semantic dimension from the target storage file in response to a data request for the target semantic dimension.

[0121] The data storage device provided in this application embodiment can execute the data storage method provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the various modules and units described above are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0122] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0123] The following is a detailed reference. Figure 5 The diagram illustrates a structural schematic suitable for implementing the electronic device described in the embodiments of this application. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 501, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 502 or a program loaded from memory 508 into random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the electronic device. The processor 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0124] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 508 including, for example, magnetic tapes, hard disks, etc.; and communication devices 509. Communication device 509 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0125] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a memory 508, or installed from a ROM 502. When the computer program is executed by the processor 501, it performs the functions defined in the data storage method of embodiments of this application.

[0126] Figure 5 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0127] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the data storage method shown in the above embodiments is implemented.

[0128] A portion of this application can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to this application through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0129] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A data storage method, characterized in that, The method includes: Acquire remote sensing embedding data stored in first-order deep format, the remote sensing embedding data consisting of at least two time slices; The remote sensing embedded data is parsed to obtain multiple numerical components belonging to different semantic dimensions. The bit width of the numerical components is smaller than the bit depth of the corresponding first bit of the remote sensing embedded data. According to the preset semantic pairing relationship, at least two of the numerical components are concatenated into a composite byte; The composite bytes belonging to the same semantic dimension and the same time slice are aggregated into a storage band organized in a second bit depth format, wherein the second bit depth of the storage band is less than the first bit depth. Each of the aforementioned storage bands is encapsulated to obtain the target storage file.

2. The method according to claim 1, characterized in that, The process of parsing the remote sensing embedded data yields multiple numerical components belonging to different semantic dimensions, including: Obtain the energy level parameter sequence corresponding to the remote sensing embedded data; Using the energy level parameter sequence, a first numerical component and at least one second numerical component are decomposed step by step from each pixel value of the remote sensing embedded data. The plurality of numerical components include the first numerical component and the second numerical component, wherein the first numerical component and the second numerical component belong to different semantic dimensions.

3. The method according to claim 2, characterized in that, The preset semantic pairing relationship is the pairing relationship between the first numerical component and the second numerical component; The step of concatenating at least two numerical components into a composite byte according to a preset semantic pairing relationship includes: The first and second numerical components that satisfy the preset semantic pairing relationship are concatenated bit-wise to obtain the composite byte.

4. The method according to claim 1, characterized in that, The encapsulation of each of the storage bands to obtain the target storage file includes: For any given time slice, the stored bands within the time slice are arranged adjacently to obtain the slice band group corresponding to the time slice; Arrange the slice band groups corresponding to each time slice in time sequence to obtain a time-series band sequence. The time-series band sequence is encapsulated to obtain the target storage file.

5. The method according to claim 1, characterized in that, The encapsulation of each of the storage bands to obtain the target storage file includes: The stored bands belonging to the same time slice are arranged in a band staggered pattern to obtain the arranged band sequence. The arranged band sequence is divided according to a preset size to obtain the storage tiles corresponding to the band sequence. The storage tiles are compressed to obtain the target storage file.

6. The method according to claim 1, characterized in that, The method further includes: In response to a data request for a target semantic dimension in the target storage file, a storage band corresponding to the target semantic dimension is extracted from the target storage file.

7. A data storage device, characterized in that, The device includes: The acquisition module is used to acquire remote sensing embedded data stored in first-bit deep format, the remote sensing embedded data consisting of at least two time slices; The parsing module is used to parse the remote sensing embedded data to obtain multiple numerical components belonging to different semantic dimensions. The bit width of the numerical components is smaller than the bit depth corresponding to the remote sensing embedded data. The splicing module is used to splice at least two of the numerical components into a composite byte according to a preset semantic pairing relationship; The aggregation module is used to aggregate the composite bytes that belong to the same semantic dimension and the same time slice into a storage band organized in a second bit depth format, wherein the second bit depth of the storage band is less than the first bit depth. The encapsulation module is used to encapsulate each of the storage bands to obtain the target storage file.

8. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory stores computer instructions, and the processor executes the computer instructions to perform the data storage method of any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to perform the data storage method according to any one of claims 1 to 6.

10. A computer program product, characterized in that, It includes computer instructions for causing a computer to perform the data storage method according to any one of claims 1 to 6.

Images