Remote sensing image spatio-temporal query method based on multi-dimensional fusion grid index, medium and equipment

Abstract

The invention provides a remote sensing image spatio-temporal query method based on multi-dimensional fusion grid indexing, a medium and equipment, and the method comprises the steps: firstly carrying out the multi-dimensional spatio-temporal division of a global grid, and coding grid units, so as to obtain a space filling code which can be quickly converted between two-dimensional and three-dimensional grid information; secondly, designing a secondary index model based on grid unit coding, and proposing an acceleration scheme of large-scale space-time range query on this basis, that is, quickly generating a high-precision grid unit set through two-dimensional grid matching, and then performing three-dimensional replacement on the two-dimensional units to obtain a large-scale space-time range query; and a high-quality query code approximate to a three-dimensional grid matching result is generated at lower calculation cost. In addition, a query partitioning scheme is also introduced, which allows a query to be divided into a plurality of sub-queries according to grid units, and the sub-queries are distributed to different nodes to realize distributed loading of metadata. The query efficiency can be effectively improved, the query precision is improved, and meanwhile the calculation cost is reduced.

Description

Technical Field

[0001] The present invention belongs to the field of remote sensing image query, and in particular relates to a remote sensing image spatiotemporal query method, medium and equipment based on multi-dimensional fusion grid index. Background Art

[0002] As satellites launched by various earth observation agencies continue to collect information about ground objects, the amount of remote sensing data archives is growing continuously and rapidly. Massive remote sensing data provides sufficient knowledge for exploration. In recent years, with the emergence of large-scale remote sensing online analysis services, the knowledge in massive remote sensing data is being continuously explored, and application research in the fields of agriculture and food production, land development and urbanization, forest cover, sustainable development indicators, and water body changes is constantly emerging.

[0003] Remote sensing images are a typical type of spatiotemporal data, and their access is highly dependent on the query of spatiotemporal metadata. Therefore, in large-scale remote sensing data management, metadata is usually separated and extracted, and managed based on spatiotemporal databases. The remote sensing image data access and loading process can be summarized as the following steps: 1) Filter the data set based on metadata; 2) Divide the tasks according to the filtered data; 3) Distribute the tasks to the computing cluster, and each node loads the data relatively independently and performs analysis operations.

[0004] The extremely high usage frequency and query overhead make the query based on spatiotemporal metadata a performance bottleneck in remote sensing image query. Remote sensing image metadata is a small data with spatiotemporal characteristics, which is very suitable for management based on spatiotemporal database systems. However, the indexing schemes of existing mainstream spatiotemporal database systems have a lot of room for improvement in the performance of large-scale spatiotemporal data queries: 1) The multidimensional tree-based spatiotemporal index structure represented by PostGIS has serious performance degradation in massive remote sensing image scenarios; 2) The grid-based index structure represented by MongoDB cannot perform fine-grained control on grid division, and the performance still cannot meet the performance requirements of large-scale remote sensing data analysis, while target data screening based on geographic space range and time window is a necessary step in remote sensing analysis. Therefore, the query based on spatiotemporal metadata has become a major performance bottleneck in remote sensing data management. There are currently some two-dimensional or three-dimensional grid indexing schemes constructed using the spatiotemporal continuity characteristics of remote sensing image datasets. These schemes usually have excellent performance under some query conditions, but poor performance under other conditions. For example, the two-dimensional grid indexing scheme performs well only in spatial queries that do not involve the time dimension or spatiotemporal queries with a smaller time range; the three-dimensional grid indexing scheme performs well in range queries with similar scales in the spatiotemporal dimension, and it is difficult to guarantee correctness under larger range conditions.

[0005] Therefore, considering the data access characteristics of large-scale remote sensing analysis tasks, in order to achieve high-performance data access, it is very important and technically feasible to design an efficient spatiotemporal metadata query method. Summary of the invention

[0006] In view of the deficiencies in the prior art, the present invention provides a remote sensing image spatiotemporal query method, medium and device based on multi-dimensional fusion grid index.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] A remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index is characterized by comprising:

[0009] Extract the spatiotemporal center point of the imported remote sensing image, generate a unique identification code and build a grid index, and attach the identification code to the remote sensing image dataset to form a metadata record;

[0010] Accept user query requests and generate corresponding query statements;

[0011] The generated query statements are distributed to each computing node, and each computing node independently sends a query statement execution request to the database that records the metadata and obtains the query results.

[0012] To optimize the above technical solutions, the specific measures taken also include:

[0013] Furthermore, the extracting of the spatiotemporal center point of the imported remote sensing image, generating a unique identification code and constructing a grid index specifically includes:

[0014] Step 101: extracting the time-space center point coordinates (xc, yc, tc) of the imported remote sensing image, where (xc, yc) represents the two-dimensional space coordinates and tc represents time;

[0015] Step 102: Calculate the unique identification code of the three-dimensional grid cell corresponding to the center point of the remote sensing image based on the (xc, yc, tc) triple, construct a B+ tree index using the unique identification code of the three-dimensional grid cell, and attach the identification code to the remote sensing image data set to form a metadata record.

[0016] Furthermore, the accepting of a user query request and generating a corresponding query statement specifically includes:

[0017] Step 201: converting the spatiotemporal range of the user query request into a spatiotemporal range object R, where R includes a two-dimensional spatial query range and a time query range;

[0018] Step 202: matching the spatial query range of R with the two-dimensional grid cells of the three-dimensional grid cells obtained in step 102 in a certain time plane, generating a pending grid set and a hit grid set, wherein the hit grid set includes the determined matched two-dimensional grid cells, and the pending grid set includes the pending two-dimensional grid cells;

[0019] Step 203: Expand the grid units of the pending grid set, and use the two-dimensional grid units introduced by the expansion as the two-dimensional grid units to be determined; and merge the grid units of the pending grid set after the expansion.

[0020] Step 204: replacing the two-dimensional grid cells that meet the time span requirement in the hit grid set with three-dimensional grid cells;

[0021] Step 205: Divide the pending grid set and the hit grid set into a series of partitions;

[0022] Step 206: Convert the grid unit set corresponding to each partition into a series of query statements.

[0023] Furthermore, in step 202, the matching method of generating the pending grid set and the hit grid set is:

[0024] Select the smallest grid unit that can accommodate R, extract its subunits, and judge the spatial position relationship between each subunit and R: add the subunits completely contained by R to the hit grid set, and add the intersecting subunits to the pending grid set.

[0025] Furthermore, in step 203, the outward expansion of the grid unit specifically includes: incorporating the two-dimensional grid unit of the outer layer of the undetermined grid set into the undetermined grid set;

[0026] The merging of the grid units is specifically as follows: all grid units in the pending grid set are traversed, and for grid units whose parent units all have child units in the pending grid set at the same time, all child units are replaced with the parent unit.

[0027] Furthermore, in step 204, the time span requirement is: a time query range whose step length is less than R in the time dimension.

[0028] Furthermore, in step 205, the partitions are divided as follows:

[0029] Find the smallest circumscribed square that contains all current grid cells in two-dimensional space;

[0030] Calculate the square root of the number of computing nodes, then round up the square root to find the root n of the smallest square number, and divide the smallest circumscribed square into n*n, along the edges of the grid units;

[0031] The divided blocks are numbered, and the numbered blocks are partitioned for query tasks according to the computing nodes.

[0032] Furthermore, in step 206, the query statement is converted as follows:

[0033] For the three-dimensional grid cells in the hit grid set, a query condition is generated using the corresponding identification code range;

[0034] For the two-dimensional grid cells in the hit grid set, the corresponding identification code range and the time query range of R are used to jointly generate the query conditions;

[0035] For the two-dimensional grid cells in the undetermined grid set, the spatial query range and the temporal query range of R are used to jointly generate query conditions.

[0036] Accordingly, the present invention proposes a computer-readable storage medium storing a computer program, characterized in that the computer program enables a computer to execute the remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index as described above.

[0037] Correspondingly, the present invention proposes an electronic device, characterized in that it includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index as described above is implemented.

[0038] The beneficial effects of the present invention are as follows: when constructing an index, the present invention uses a three-dimensional grid division and encoding method to extract the spatiotemporal features of remote sensing images and construct an index; when querying, the spatiotemporal query range conditions are converted into a series of grid units, and then a series of integer range query conditions are generated using grid unit encoding, and some slow spatiotemporal range queries are replaced with faster integer range queries to improve query efficiency. In view of the problem that the spatiotemporal range three-dimensional grid matching process has a large amount of calculation and is difficult to achieve high precision, the present invention adopts a multi-dimensional fusion grid matching method, first uses a two-dimensional matching method to achieve a high-precision conversion, and then performs a three-dimensional replacement to generate a query code close to the quality of the three-dimensional matching result, which can significantly improve the performance and accuracy of large-scale spatiotemporal range queries while reducing the amount of calculation. Based on the above work, the present invention also proposes a query partitioning scheme that allows for specified granularity to support multi-node concurrent queries in distributed computing scenarios, directly load metadata to each node for subsequent calculations, avoid the I / O overhead of data distribution in centralized queries, and reduce query overhead. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a flow chart of a remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index.

[0040] Figure 2 This is a schematic diagram of global mesh division.

[0041] Figure 3 It is a schematic diagram of the grid coding method.

[0042] Figure 4 This is a schematic diagram of the index structure.

[0043] Figure 5 It is the grid matching flowchart of the spatiotemporal query range.

[0044] Figure 6 This is a schematic diagram of the expansion of the grid set.

[0045] Figure 7 This is a schematic diagram of the grid unit merging results.

[0046] Figure 8 It is a schematic diagram of three-dimensional grid unit replacement.

[0047] Fig. 9 It is a schematic diagram of the grid unit set partitioning.

[0048] Fig.10 It is a schematic diagram of the coding query range.

[0049] Fig.11 This is a schematic diagram of a typical geographical range.

[0050] Fig.12a This is a schematic diagram of the query results of T1 and various spatial range combinations.

[0051] Figure 12b This is a schematic diagram of the query results of T2 and various spatial range combinations.

[0052] Fig.12c This is a schematic diagram of the query results of T3 and various spatial range combinations.

[0053] Fig.12d This is a schematic diagram of the query results of Tn and various spatial range combinations. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0055] In one embodiment, the present invention proposes a remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index, the method is summarized as follows: Figure 1 As shown, the following will be for Figure 1 Each step in is described in detail.

[0056] 1. Space-time division and grid unit coding

[0057] First, the global space-time range is recursively divided, and the grid cells are one-dimensionally encoded while constructing the space-time division grid to prepare for the subsequent construction of the grid index.

[0058] This method selects EPSG:4326 as the coordinate reference system (CRS) for all geographic locations in the spatiotemporal grid partitioning model. In this coordinate system, geographic location information is expressed in the form of longitude and latitude. Unlike geographic spatial coordinates, there are no recognized upper and lower bounds for time attributes. Therefore, this model uses the Unix system Epoch time as the lower bound of time. The longitude range of the spatiotemporal grid partitioning model is -180° to +180°, the latitude range is -90° to +90°, and the time range is 0:0:00 on January 1, 1970 to 0:0:00 on January 1, 2070.

[0059] Then the global space-time range is recursively divided, such as Figure 2 As shown. The grid division level is level, and the upper and lower limits of the longitude range are ub respectively. x lb x (ub and lb are the abbreviations of upper_bound and lower_bound respectively), the upper and lower bounds of the latitude range are ub y lb y The upper and lower bounds of the time range are ub t lb t ; Then the step length calculation formula of the grid unit in the three dimensions of longitude, latitude and time is The subscripts of the grid cells in the three dimensions of longitude, latitude, and time are i, j, and k respectively. The calculation formula for the starting coordinate offset of the global grid cell division is (ofs x , ofs y , ofs t )=(lb x +i*sz x , lb y +j*sz y , lb t +k*sz t ). According to the starting coordinate offset and the step size of each dimension, the vertex coordinates of the grid unit can be calculated, so as to strictly determine the position of the grid unit. The point with the minimum value in the three dimensions of longitude, latitude and time is the starting point of the grid unit, and its three-dimensional coordinate calculation formula is (x min ,y min , t min )=(ofs x , ofs y , ofs t); The point with the maximum value in longitude, latitude and time is the end point of the grid unit, and its three-dimensional coordinate calculation formula is (x max ,y max , t max )=(ofs x +sz x , ofs y +sz y , ofs t +sz t ).

[0060] The Z space filling curve can be quickly constructed into a low-dimensional projection through binary masking. Therefore, this method uses 64-bit integer Morton coding to construct a three-dimensional Z space filling curve, and generates a code based on the cross-permutation and combination of the binary bits of the upper and lower scripts in the three dimensions. The code is used as the unique identifier of the grid unit. The code generation process is as follows: Figure 3 As shown in the figure. In order to represent information of up to three dimensions, this model limits the highest grid division level to 21, and each dimension occupies 21 bits, thereby maximizing the use of bit space. In addition, 64-bit integer data is supported by most database management systems and has the advantage of easy integration.

[0061] 2. Index Model

[0062] Different from the multidimensional tree-like spatiotemporal index structure used in traditional database systems, this method uses a grid index built based on spatiotemporal partitioning. Each grid cell divided by the grid partitioning scheme proposed above represents a unique three-dimensional spatiotemporal cube within the global spatiotemporal range. The cube does not overlap or intersect with any other cells and is used to identify a specific spatiotemporal area. Remote sensing images are usually stored in a tiled manner, so they have a relatively fixed and regular spatiotemporal range, and their spatiotemporal center point can retain its spatiotemporal information with the lowest possible error. This method extracts the spatiotemporal center point of the remote sensing image when importing metadata, and records the grid cell where the point is located, and then builds a B+ tree index for the grid cell field. The number of index nodes under this model will not exceed the number of grid partitioning units. When the newly inserted data falls into the grid cell where the existing data exists, the index structure will not change, which is very suitable for remote sensing big data scenarios with frequent updates. The data query process under this index model is as follows: the grid cell secondary index is used first to quickly reduce the search space, and then the spatiotemporal range index is used to scan part of the data, thereby greatly reducing the amount of calculation. This index model is as follows: Figure 4 As shown, the query process based on it will be described in detail below.

[0063] 3. Spatial-temporal range query method

[0064] The optimized spatiotemporal query processing flow aims to quickly narrow down the search space by converting the spatiotemporal query range into a series of spatiotemporal grid cells and using the secondary index built on the spatiotemporal grid cells. Among them, the most important work is to match the irregular spatiotemporal query range into a set of spatiotemporal grid cells. Since the complexity of the grid matching process is related to the number of dimensions, this method adopts a two-dimensional matching process that only uses spatial information to quickly achieve high accuracy. Subsequently, the generated two-dimensional grid cells are adjusted and partially replaced with three-dimensional grid cells to improve the quality of the generated query statements. Finally, in the query statement generation stage, the query is allowed to be split by specifying the number of partitions. The query process under the multi-dimensional secondary grid index scheme can be summarized into the following steps: spatiotemporal query range grid matching, grid cell set adjustment, grid cell three-dimensional replacement, grid set partitioning and query statement generation, which will be explained in detail below.

[0065] (1) Generation of 2D grid cell sets

[0066] In order to match a higher grid level in the shortest possible time, this method uses the spatial properties of the space-time range to convert the space-time range into a series of two-dimensional grid cells. Subsequently, this method performs three-dimensional replacement based on the two-dimensional grid cell set to fully utilize the advantages of three-dimensional grid division.

[0067] Let R be the input spatial query range, C be the smallest grid unit that can just accommodate R, and the spatial position relationship between R and C is as follows: Figure 5 The specific steps of query range grid matching are as follows: 1) Divide C into four sub-grid units to obtain a sub-unit set; 2) Iterate the sub-unit set to determine the spatial relationship between each sub-unit and R. If the sub-unit is included in R, add it to the hit set; if the sub-unit intersects with R, add it to the pending set and repeat step 1; if the sub-unit does not intersect with R, discard it.

[0068] (2) Grid cell set adjustment

[0069] In some datasets, remote sensing image data may have different spatiotemporal range sizes. Therefore, the spatiotemporal grid unit set generated by the spatiotemporal query range matching process cannot be directly used in subsequent processing procedures, and some processing is required for the pending units before they can be used.

[0070] Since the index model only records the center point of the object, when there is a large spatiotemporal volume in the data set, the grid unit where the object center is located may not intersect with the matching result set, resulting in the loss of correctness of the query result. Therefore, this method needs to perform a round of expansion on the pending result set, such as Figure 6As shown. The units in the outer layer are included in the pending set for subsequent judgment. For large-volume objects that still cannot guarantee the correctness of the query results after executing the external expansion process, this method ensures the correctness of the query results by storing them in a separate table or collection in the database and relying on the database's native spatial index for querying. Among them, the condition for judging large-volume objects is: if the current tile size exceeds twice the median size of the tiles or remote sensing images in the dataset (either the length or width exceeds), it is considered a large-volume object.

[0071] It is worth noting that in order to control the access concurrency granularity and data loading granularity, remote sensing image data usually needs to be tiled before accessing the system. Tile segmentation refers to dividing the entire large image into several smaller image slices according to the grid. The resulting slices are called tiles, and the tiles have a relatively consistent size. Therefore, super large data is considered abnormal data in the data set after tile segmentation, and the probability of occurrence is extremely small. Directly using data-oriented spatial indexes with relatively low query efficiency will not lead to a significant increase in query response time.

[0072] After the above-mentioned external expansion process has been corrected, the grid cell set still cannot be directly used for query statement generation. The reason is that a large number of grid cells will cause the generated query code to contain complicated range conditions, which will affect the execution efficiency of the query code. Therefore, this method introduces a grid cell merging stage to reduce the number of grid cells in the set and improve the quality of subsequent query statement generation. The specific steps of grid merging are as follows: 1) Traverse all grid cells in the set, and for grid cells whose parent cells all child cells (i.e., brother cells) are also present in the set, use the parent cell to replace all child cells; 2) Repeat step 1) until the elements in the set no longer change. The result of the grid cell merging process is shown in the figure. Figure 7 shown.

[0073] (3) Three-dimensional replacement of grid cells

[0074] After generating a set of available two-dimensional grid cells, this method screens the two-dimensional grid cell set, selects suitable parts, and generates three-dimensional grid cells in combination with the time conditions of the query range. The query code generated by the three-dimensional grid cell uses fewer composite indexes, scans less data than the query code generated by the two-dimensional grid cell, and has better execution efficiency. The specific steps for three-dimensional replacement of grid cells are as follows: 1) Find the lowest-level grid cell whose time span is smaller than the time range; 2) Combine the two-dimensional grids of the corresponding level in the result set with the time query range into a three-dimensional grid cell and replace it. The results of the three-dimensional replacement process of grid cells are shown in the figure below. Figure 8 shown.

[0075] (4) Grid set partitioning

[0076] Through the above steps, a grid subunit set that retains the spatiotemporal information of the query range can be obtained. The results of large-scale remote sensing image data range queries usually contain a large number of records, which are not suitable for subsequent processing by a single node. Therefore, this method introduces a query code partitioning stage to divide the grid subunit set into a specified number of partitions, such as Fig. 9 As shown, this supports sending query results to different computing nodes for subsequent distributed processing.

[0077] The steps of the grid set partitioning algorithm are described as follows: 1) Find the smallest circumscribed square containing the current grid; 2) Find the smallest square number greater than the current total number based on the input computing nodes to divide the circumscribed square, that is, find the square root of the computing nodes, and then round the square root up to find the smallest square number. For example, if the number of computing nodes is 5, the smallest square number greater than the current total number is 9, so it is divided into 3*3; when dividing, the edges of the internal grid will be used as the dividing line; 3) The divided blocks are numbered from left to right, and then the numbered blocks are partitioned according to the computing nodes for query tasks.

[0078] (5) Query statement generation

[0079] After all the above steps, a series of partitioned and highly optimized grid division sub-unit sets can be obtained, including several replaced three-dimensional grid units and unreplaced two-dimensional grid units. Each grid unit in the set has a unique (Code, Level) tuple, where Code represents the grid unit code and Level is the grid division level of the unit. Each tuple can be converted into a series of continuous highest-level grid sub-unit codes through the following strategy to generate an integer range condition for generating query codes: using the shared prefix feature of Morton coding, the range of grid sets of the specified sub-level is generated by filling all 0s and all 1s in the low bits, such as Fig.10 shown.

[0080] The replaced three-dimensional grid units and the unreplaced two-dimensional grid units generated after the above steps are mainly divided into three categories: the first category is the three-dimensional hit grid set contained in the input query conditions (time range and space range), and the query conditions can be directly generated using the coding range; the second category is for the two-dimensional hit grid set that has not been three-dimensionally expanded and whose spatial dimension is within the input spatial range, and the query conditions are jointly generated using the coding range and the time query range; the third category is for the two-dimensional pending grid set that has not been three-dimensionally expanded and intersects with the input spatial range in terms of spatial dimension, and the query conditions need to be jointly formed with the time query range and space query range entered by the user.

[0081] According to the above steps, the specific implementation steps of this method are divided into the following three stages:

[0082] In stage 1, to build a grid index, when importing data, it is necessary to extract the spatiotemporal center point of the remote sensing image, generate the code based on it and build the index. The specific implementation steps are as follows:

[0083] Step 101: The system extracts the coordinates (xc, yc, tc) of the time-space center point of the remote sensing image to be imported.

[0084] Step 102: Calculate the unique identification code of the grid unit corresponding to the center point of the image based on the (xc, yc, tc) triplet obtained in step 101, and construct a B+ tree index using the grid unit identification code. The index structure is as follows: Figure 4 As shown; and append the identification code to the image dataset table to construct the metadata record.

[0085] Phase 2: accepting user query requests and generating corresponding query statements. The specific implementation steps are as follows:

[0086] Step 201: extract the spatiotemporal range in the user query request and convert it into a spatiotemporal range object R.

[0087] Step 202: Using the two-dimensional matching method mentioned above, recursively match R extracted in step 201 with grids at all levels to generate a pending grid set Gridspending-2d and a hit grid set Gridscertain-2d. The specific matching algorithm is: select the smallest grid unit that can accommodate R, extract its subunits, judge the spatial position relationship between each subunit and R, add the subunits that are completely contained by R to the Gridscertain-2d set, and add the intersecting ones to the Gridspending-2d set; then repeat the above matching process for the subunits in the Gridspending-2d set until the target division level is reached.

[0088] Step 203: For the purpose of correctness assurance, the pending grid set is adjusted, and a round of expansion is performed on the grid units. The grid units in the new set introduced by the expansion are regarded as pending units, such as Figure 6 As shown; the grid units of the undetermined grid set after external expansion are then merged.

[0089] Step 204: In order to improve the execution performance of the generated query statement, the hit grid set is adjusted, and the appropriate two-dimensional grid cells are replaced with three-dimensional grid cells; the specific method is: select all the two-dimensional grid cells on the division level whose step size Tstep in the time dimension is just smaller than the time component T in the given space-time range R, and upgrade them to three-dimensional grid cells, such as Figure 8 shown.

[0090] Step 205: For the purpose of correctness assurance, the two grid unit sets are adjusted and divided into a series of partitions for generating block query sub-statements; the specific method is: select the largest circumscribed square of the corresponding grid, and then partition according to the computing nodes, such as Fig. 9 and Fig.10 shown.

[0091] Step 206: Convert the adjusted grid unit set in each partition in the above step into a high-performance query statement set, first using the Z grid coding bit arrangement characteristics introduced above, such as Fig.10 As shown, the low bits of the unique identifier of each unit are filled, and each unit is expanded to a range from all low bits being 0 to all low bits being 1. The ranges are linked using AND to generate integer range query conditions. Finally, the spatiotemporal query clause replaced by the integer range condition is combined with other conditions to generate a series of query statements. In this way, most complex and time-consuming spatial range queries can be converted into simple integer range queries.

[0092] Phase 3: Distribute the generated query statements to each computing node and request a query from the metadata database storage system. The specific implementation steps are as follows:

[0093] Step 301: The task initiator uses a series of query statements generated in stage 2 to divide the subtasks and distribute the tasks to the nodes on the computing cluster.

[0094] Step 302: The tasks on each computing node independently send query statement execution requests to the metadata repository system, and then obtain results to execute subsequent computing logic.

[0095] In order to verify the effect of this method, the remote sensing image metadata of the Landsat series of satellites is used as the query target experimental set. The Landsat series of satellites is a group of satellites jointly managed by the National Aeronautics and Space Administration (NASA) and the United States Geological Survey (USGS). Since the launch of the first Landsat satellite in 1972, this series of satellites has continuously collected remote sensing image data for earth observation, providing the world's longest continuous air-based observation record. Landsat satellite data has relatively rich metadata and relatively complete geographic spatial coverage and time range coverage. Using it as an experimental verification data set has certain practical reference significance.

[0096] The dataset used in this experiment comes from the Landsat satellite metadata service provided by USGS (US Geological Survey, 2024). For details, see Table 1. The Landsat-MSS-C2-L1 dataset contains the Multi-Spectral Scanner (MSS) data collected by the Landsat series of satellites from 1972 to 1999; the Landsat-TM-C2-L1 dataset contains the Thematic Mapper (TM) data collected by the Landsat series of satellites from 1982 to 2012; the Landsat-ETM-C2-L1 dataset contains the Enhanced Thematic Mapper Plus (ETM+) data collected by the Landsat series of satellites from 1999 to 2023; the Landsat-OT-C2-L1 dataset contains the Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS) data collected by Landsat satellites from 2013 to 2023.

[0097] Table 1 Landsat dataset information

[0098] Dataset name Collection period Number of images Landsat-MSS-C2-L1 1972-1999 1842292 Landsat-TM-C2-L1 1982-2012 2923623 Landsat-ETM-C2-L1 1999-2023 3355632 Landsat-OT-C2-L1 2013-2023 3271868

[0099] In order to verify the performance improvement effect of this method, the experiment constructed Fig.11 Several typical geographic space ranges are shown in Figure 2 and several time query ranges are shown in Table 2. The experiment will construct spatiotemporal queries based on these space and time ranges.

[0100] Table 2 Time query range

[0101] Range Name Start time End Time Tn / / T0 1983-01-01 2023-01-01 T1 2013-01-01 2023-01-01 T2 2020-01-01 2023-01-01

[0102] In order to verify the performance of this method under different query cases, this experiment constructed a series of spatiotemporal range query requests based on the combination of spatiotemporal ranges. The above queries were executed using MongoDB's built-in 2dSphere index, 2D grid index, 3D grid index, and this method to compare and verify the performance. The experimental results are shown in Table 3.

[0103] Table 3 Time consumption of each time and space query

[0104]

[0105]

[0106] like Figures 12a to 12dAs shown in the performance comparison experiment based on MongoDB, by comparing different spatial range query strategies, this method can reduce the query overhead by up to 93.11% compared with the 2dSphere index, significantly improving the performance of large-scale spatiotemporal range queries, and providing an efficient query method for large-scale remote sensing data queries.

[0107] In another embodiment, the present invention proposes a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index as described in the first embodiment.

[0108] In another embodiment, the present invention proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index as described in Example 1 is implemented.

[0109] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. More specific examples of computer storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0110] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0111] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index, characterized in that: include: Extract the spatiotemporal center point of the imported remote sensing image, generate a unique identification code and build a grid index, and attach the identification code to the remote sensing image dataset to form a metadata record; Accept user query requests and generate corresponding query statements; The generated query statements are distributed to each computing node, and each computing node independently sends a query statement execution request to the database that records the metadata and obtains the query results.

2. The remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index according to claim 1, characterized in that: The step of extracting the spatiotemporal center point of the imported remote sensing image, generating a unique identification code and constructing a grid index specifically includes: Step 101: extracting the time-space center point coordinates (xc, yc, tc) of the imported remote sensing image, where (xc, yc) represents the two-dimensional space coordinates and tc represents time; Step 102: Calculate the unique identification code of the three-dimensional grid cell corresponding to the center point of the remote sensing image based on the (xc, yc, tc) triple, construct a B+ tree index using the unique identification code of the three-dimensional grid cell, and attach the identification code to the remote sensing image data set to form a metadata record.

3. The remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index as claimed in claim 2, characterized in that: The receiving of a user query request and generating a corresponding query statement specifically includes: Step 201: converting the spatiotemporal range of the user query request into a spatiotemporal range object R, where R includes a two-dimensional spatial query range and a time query range; Step 202: matching the spatial query range of R with the two-dimensional grid cells of the three-dimensional grid cells obtained in step 102 in a certain time plane, generating a pending grid set and a hit grid set, wherein the hit grid set includes the determined matched two-dimensional grid cells, and the pending grid set includes the pending two-dimensional grid cells; Step 203: Expand the grid units of the pending grid set, and use the two-dimensional grid units introduced by the expansion as the two-dimensional grid units to be determined; and merge the grid units of the pending grid set after the expansion. Step 204: replacing the two-dimensional grid cells that meet the time span requirement in the hit grid set with three-dimensional grid cells; Step 205: Divide the pending grid set and the hit grid set into a series of partitions; Step 206: Convert the grid unit set corresponding to each partition into a series of query statements.

4. The remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index as claimed in claim 3, characterized in that: In step 202, the matching method of generating the pending grid set and the hit grid set is: Select the smallest grid unit that can accommodate R, extract its subunits, and judge the spatial position relationship between each subunit and R: add the subunits completely contained by R to the hit grid set, and add the intersecting subunits to the pending grid set.

5. The remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index as claimed in claim 3, characterized in that: In step 203, the expansion of the grid unit specifically includes: incorporating the two-dimensional grid unit in the outer layer of the pending grid set into the pending grid set; The merging of the grid units is specifically as follows: all grid units in the pending grid set are traversed, and for grid units whose parent units all have child units in the pending grid set at the same time, all child units are replaced with the parent unit.

6. The remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index as claimed in claim 3, characterized in that: In step 204, the time span requirement is: a time query range whose step length is less than R in the time dimension.

7. The remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index as claimed in claim 3, characterized in that: In step 205, the partitions are divided as follows: Find the smallest circumscribed square that contains all current grid cells in two-dimensional space; Calculate the square root of the number of computing nodes, then round up the square root to find the root n of the smallest square number, and divide the smallest circumscribed square into n*n, along the edges of the grid units; The divided blocks are numbered, and the numbered blocks are partitioned for query tasks according to the computing nodes.

8. The remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index as claimed in claim 3, characterized in that: In step 206, the query statement is converted as follows: For the three-dimensional grid cells in the hit grid set, a query condition is generated using the corresponding identification code range; For the two-dimensional grid cells in the hit grid set, the corresponding identification code range and the time query range of R are used to jointly generate the query conditions; For the two-dimensional grid cells in the undetermined grid set, the spatial query range and the temporal query range of R are used to jointly generate query conditions.

9. A computer-readable storage medium storing a computer program, characterized in that: The computer program enables a computer to execute the remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index as described in any one of claims 1 to 8.

10. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the remote sensing image spatiotemporal query method based on multi-dimensional fusion grid index as described in any one of claims 1 to 8 is implemented.

Images