A remote sensing image unified identification method based on GeoS0T composite coding

Abstract

The present application relates to remote sensing image index retrieval technical field, especially to a kind of remote sensing image unified identification method based on GeoSOT composite coding, comprising: extracting remote sensing image spatial range, resolution, imaging time and the like metadata and constructing structured record;Generate the GeoSOT composite coding containing fixed length position code, resolution code, timestamp and variable length field;Binding coding and image storage address and metadata, derive multidimensional index features such as space and time;Distributed multidimensional inverted index is constructed and is synchronously maintained;Analysis search request, after set operation, fine filtering, backtracking to obtain result.The present application expresses image multidimensional attribute by structured GeoSOT composite coding, and constructs collaborative multidimensional inverted index based on coding, converts complex space-time-resolution-sensor combination retrieval into multiple one-dimensional inverted lookup and set operation, significantly improves the retrieval efficiency and system expansibility of mass multi-source remote sensing image.

Description

Technical Field

[0001] This invention relates to the field of remote sensing image indexing and retrieval technology, and in particular to a unified identification method for remote sensing images based on GeoSOT composite coding. Background Technology

[0002] With the rapid development of remote sensing technology, multi-source remote sensing platforms such as high-resolution satellites, aerial photography, and drones are constantly emerging. Remote sensing images exhibit characteristics of multi-source, multi-temporal, multi-resolution, and massive data volume, and have been widely applied in many fields such as land surveys, urban planning, and disaster emergency response. Currently, mainstream remote sensing data management systems generally adopt a multi-field distributed storage mode, recording the spatial location, resolution, time, sensor, and other attributes of the image in different fields or data tables. Data identification relies on file paths, file names, or database auto-incrementing IDs, lacking a unified structured coding system.

[0003] In this traditional model, the labeling rules for remote sensing images from different sources and platforms are inconsistent, making it difficult to establish a globally unified spatial, temporal, resolution, and sensor mapping relationship, which leads to significant challenges in data integration and sharing. At the same time, remote sensing image retrieval relies on multi-field joint queries, complex SQL conditions, or multi-table joins, resulting in lengthy query paths. In scenarios with massive amounts of data, performance bottlenecks are easily encountered, failing to meet the business needs for rapid positioning and unified management of multi-source, multi-temporal, and multi-resolution images. This has become a core issue restricting the efficient utilization of remote sensing data. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides a unified identification method for remote sensing images based on GeoSOT composite coding, aiming to improve the inefficiency of existing technologies where multi-source remote sensing images are not uniformly identified and retrieval relies on joint queries of multiple fields.

[0005] In a first aspect, the present invention provides the following technical solution: a unified identification method for remote sensing images based on GeoSOT composite coding, comprising the following steps: S1. Remote sensing image metadata extraction: Preprocess the original remote sensing images to extract image spatial range, image resolution, imaging time, sensor type and satellite platform information; S2, GeoSOT composite code generation: Based on the metadata extracted from S1, combined with the GeoSOT grid system, a GeoSOT composite code is generated as a globally unique identifier. S3. Encoding Binding and Multidimensional Index Feature Annotation: The GeoSOT composite code generated in S2 is used as the primary key and bound to the physical storage address of the corresponding remote sensing image and the structured metadata record built based on the metadata in S1. At the same time, the GeoSOT composite code is parsed and combined with the structured metadata to derive multidimensional features for constructing the inverted index. S4. Multidimensional Inverted Index Construction and Storage: In the distributed search and analysis engine, a document index is constructed using the GeoSOT composite code generated in S2 as the document identifier; corresponding inverted indexes are established for the multidimensional features derived from S3; when image data is added, updated, or deleted, each inverted index is updated synchronously using the GeoSOT composite code as the clue. S5. Fast retrieval based on multidimensional inverted index: Parse the user-submitted retrieval request containing spatial range, time interval, resolution conditions and sensor conditions, and map each retrieval condition to the corresponding dimension key; query the inverted index built in S4 based on the dimension key to obtain the candidate GeoSOT composite code set under each dimension; perform set operations, fine filtering and sorting on the candidate code set, and backtrack the document index to obtain the retrieval results.

[0006] By adopting the above technical solution: constructing GeoSOT composite coding that integrates spatial location, resolution, time, and sensor multi-dimensional information, a unified identification system is established to eliminate identification barriers between data from different sources. Combined with a multi-dimensional inverted index design, retrieval is transformed into a set operation of indexes of various dimensions. With a distributed sharding routing strategy, the overhead of cross-shard access is reduced. The accuracy of results is improved by relying on standardized multi-dimensional features and fine filtering processes. At the same time, it is adapted to multi-scale and multi-source image data, effectively solving the inefficiency problem of traditional retrieval relying on multi-field joint queries. It realizes rapid positioning of multi-source, multi-temporal, and multi-resolution images, providing support for the unified management and efficient retrieval of massive remote sensing images.

[0007] Preferably, step S1 further includes: extracting product level and cloud cover information; the image spatial range is defined by the coordinates of the lower left corner and the upper right corner of the image; all extracted metadata is recorded in the structured metadata and serves as the basis for S2 encoding generation, S3 feature derivation and S4 index construction.

[0008] Preferably, in step S2, the generated GeoSOT composite code is sequentially combined with a fixed-length location code, a fixed-length resolution code, a fixed-length time stamp, and a variable-length field in a preset order; wherein, the fixed-length location code is a 27-bit fixed-length code, which is composed of a 17-bit 16-level GeoSOT grid code located at the lower left corner of the image, a 5-bit latitude grid span code, and a 5-bit longitude grid span code, which are sequentially spliced ​​together, and the spatial resolution of the 16-level GeoSOT grid is 1 kilometer level.

[0009] Preferably, the fixed-length resolution code is a 5-bit fixed-length code, which has a structure of three integer bits, one decimal bit, and one unit character; the unit character includes c, m, and k, which represent centimeters, meters, and kilometers, respectively; the three integer bits and one decimal bit together constitute the resolution value, and the fixed-length resolution code can represent a resolution ranging from 1 millimeter to 999 kilometers.

[0010] Preferably, the fixed time stamp is a 14-bit fixed-length code, with the encoding order being 4 bits for year, 2 bits for month, 2 bits for day, 2 bits for hour, 2 bits for minute, and 2 bits for second; if the imaging time information in the metadata is not precise enough, it is padded with digits 0 from the lowest missing bit to 14 bits.

[0011] Preferably, the variable-length field includes a sensor type identifier and a satellite platform identifier, which are connected by a hyphen; the combination order of the GeoSOT composite code is: fixed-length location code, fixed-length resolution code, fixed-length time stamp, and variable-length field.

[0012] Preferably, in step S3, the derived multidimensional features include: Using the 16-level GeoSOT grid encoding with fixed-length position codes in the GeoSOT composite encoding generated by S2 as a method, spatial dimension features of at least one spatial grid cell identifier and which can be further derived into upper-level region IDs are obtained. Using a fixed time stamp in the GeoSOT composite code generated by a predefined granularity mapping S2 based on seconds, days, months, and quarters, the time dimension features of the corresponding time slice identifier are obtained. By mapping the values ​​of the fixed-length resolution codes in the GeoSOT composite encoding generated according to S2 to predefined resolution ranges of sub-meter, meter, and ten-meter levels, the resolution dimension features of the corresponding resolution level identifiers are obtained. By parsing the variable-length field in the GeoSOT composite code generated by S2, the sensor dimension features of the sensor type identifier and satellite platform identifier are obtained; The derivation of the multidimensional features is accomplished by combining structured metadata constructed based on metadata extracted from S1.

[0013] Preferably, in step S4, establishing the corresponding inverted index includes: A spatial inverted index with the spatial grid cell identifier in the spatial dimension feature as the key and the GeoSOT composite code set containing the spatial grid cell identifier as the value; A time inverted index with time slice identifiers in the time dimension features as keys and GeoSOT composite encoding sets containing those time slice identifiers as values; A resolution inverted index with the resolution level identifier in the resolution dimension feature as the key and the GeoSOT composite code set containing that resolution level identifier as the value; A sensor inverted index with the sensor type identifier or satellite platform identifier from the sensor dimension features as the key and the GeoSOT composite code set containing that identifier as the value.

[0014] Preferably, in step S4, when constructing the document index, GeoSOT composite encoding is used as the primary key and written into the distributed search engine. The preset length prefix of the fixed-length position code in the encoding is used for sharding routing, so that spatially adjacent grid cells fall into the same or adjacent sharding nodes, reducing the cross-sharding access overhead during spatial queries.

[0015] Preferably, in step S5, the retrieval results obtained by backtracking the document index specifically include: Set operations are performed on the candidate GeoSOT composite encoding sets under each dimension, starting with the dimension with the smallest number of candidates, to obtain a preliminary result set; The preliminary result set is finely filtered. The actual coverage of the image is calculated by the number of grids spanned by the latitude and longitude in the fixed long location code. Images that do not completely fall within the target spatial range are removed. Images that do not meet the preset resolution range are filtered by the precise value of the fixed long resolution code. Finally, the images are sorted in chronological order or in reverse order based on the fixed long time stamp to obtain the final code set. The document index built by S4 is traced back from the final encoded set to return the matching image data.

[0016] The present invention has the following beneficial effects:

[0017] 1. In this invention, by constructing GeoSOT composite coding to integrate spatial location, resolution, time, and sensor multi-dimensional information, and combining it with multi-dimensional inverted index design, remote sensing image retrieval is transformed into a set operation of indexes of various dimensions. This solves the inefficiency problem of traditional retrieval relying on multi-field joint queries, and enables rapid positioning of multi-source, multi-temporal, and multi-resolution images. At the same time, the unified coding system eliminates the barriers of inconsistent data identification from different sources, laying the foundation for unified management and efficient retrieval of massive remote sensing images.

[0018] 2. In this invention, based on the distributed fragmentation routing strategy, the fixed-length position code prefix in GeoSOT composite coding is used for fragmentation, so that spatially adjacent grid cells fall into the same or adjacent fragments, which greatly reduces the cross-fragment access overhead during spatial queries. Combined with the wide-range coverage capability of resolution codes, it is adapted to multi-scale image data from millimeter to kilometer level, significantly improving the storage scalability and retrieval response speed of massive image data in a distributed environment.

[0019] 3. In this invention, based on derived standardized multidimensional features and a fine filtering process, the actual coverage of the image is calculated by parsing the fixed-length position code, and images that do not match completely are eliminated. The images that meet the requirements are filtered with the precise value of the resolution code, which effectively improves the accuracy of the search results and avoids the redundancy of results caused by traditional indexes that are based on only a single dimension for filtering. This ensures that the search results are highly consistent with the user's needs.

[0020] 4. In this invention, relying on variable-length field design and sensor inverted index, sensor type and satellite platform identifier are included in the index dimension. Combined with the whole process design, it realizes integrated retrieval of multi-source images from different sensors and different satellite platforms. There is no need to build an additional heterogeneous data adaptation mechanism, which reduces the difficulty of integrating and sharing multi-source remote sensing data and improves the utilization efficiency of data assets. Attached Figure Description

[0021] Figure 1 This is a flowchart of a unified identification method for remote sensing images based on GeoSOT composite coding proposed in this invention;

[0022] Figure 2 This is a schematic diagram of the GeoSOT fixed-length position code generation proposed in this invention. Detailed Implementation

[0023] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] This invention provides a unified identification method for remote sensing images based on GeoSOT composite coding, such as... Figure 1 As shown, it includes the following steps: S1. Remote sensing image metadata extraction: Preprocess the original remote sensing images to extract image spatial range, image resolution, imaging time, sensor type and satellite platform information; Furthermore, step S1 also includes: extracting product level and cloud cover information; defining the image spatial range using the coordinates of the lower left and upper right corners of the image; recording all extracted metadata in structured metadata, which serves as the basis for S2 encoding generation, S3 feature derivation, and S4 index construction.

[0025] Specifically, after preprocessing the raw remote sensing images, including denoising and geometric correction, the image header file and related metadata documents are parsed to extract core metadata information, including image spatial extent, image resolution, imaging time, sensor type, and satellite platform information. The image spatial extent is defined by decimal latitude and longitude coordinates of the lower left and upper right corners, the imaging time is accurate to the second, and the sensor type and satellite platform information are directly traced back to the data source. Further, product level and cloud cover information are extracted. Product levels are divided according to the degree of preprocessing (e.g., L1A raw data, L2 atmospheric correction data), and cloud cover is expressed as a percentage. All extracted metadata is organized as relational table fields, JSON documents, or key-value pairs of structured records, serving as the foundation for subsequent encoding generation, feature derivation, and index construction.

[0026] For example, the structured metadata of a GF1B satellite PMS sensor image is as follows: lower left corner E111.35°, N22.95°, upper right corner E111.56°, N23.12°, resolution 2.2 meters, imaging time 10:30:25 on April 3, 2022, product level L1A, cloud cover 12%; as another example, in the structured metadata of a remote sensing image of a certain area, the code for the lower left corner of the image spatial range corresponding to the 16th level GeoSOT grid is G0011321001002230, and the image spans 17 16th level GeoSOT grids in the latitude direction and 21 16th level GeoSOT grids in the longitude direction.

[0027] S2, GeoSOT composite code generation: Based on the metadata extracted from S1, combined with the GeoSOT grid system, a GeoSOT composite code is generated as a globally unique identifier. Further, in step S2, the generated GeoSOT composite code is combined in a preset order with a fixed-length location code, a fixed-length resolution code, a fixed-length time stamp, and a variable-length field. The fixed-length location code is a 27-bit fixed-length code, which is composed of a 17-bit 16-level GeoSOT grid code located at the lower left corner of the image, a 5-bit latitude grid span code, and a 5-bit longitude grid span code. The spatial resolution of the 16-level GeoSOT grid is 1 kilometer. Furthermore, the fixed-length resolution code is a 5-bit fixed-length code, which consists of three integer bits, one decimal bit, and one unit character. The unit characters include c, m, and k, which represent centimeters, meters, and kilometers, respectively. The three integer bits and one decimal bit together constitute the resolution value. The fixed-length resolution code can represent a resolution range from 1 millimeter to 999 kilometers.

[0028] Furthermore, the fixed time stamp is a 14-bit fixed-length code, with the encoding order being 4 bits for year, 2 bits for month, 2 bits for day, 2 bits for hour, 2 bits for minute, and 2 bits for second; if the imaging time information in the metadata is not precise enough, it is padded with digits 0 from the lowest missing bit to 14 bits.

[0029] Furthermore, the variable-length field contains sensor type identifier and satellite platform identifier, which are connected by a hyphen; the combination order of GeoSOT composite encoding is: fixed-length location code, fixed-length resolution code, fixed-length time stamp, and variable-length field.

[0030] Specifically, based on the metadata extracted by S1 and combined with the GeoSOT global mesh system, a GeoSOT composite code is generated as a globally unique identifier. This code is constructed according to a preset order of "fixed-length location code → fixed-length resolution code → fixed-length timestamp → variable-length field". The fixed-length location code is a 27-bit fixed-length code: it consists of a 17-bit 16-level GeoSOT grid code located at the lower left corner of the image. The 16-level grid was chosen because of its 1-kilometer-level resolution, balancing expression accuracy and coding efficiency. The code is composed of 5 bits for the number of grids spanned in the latitude direction and 5 bits for the number of grids spanned in the longitude direction. The lower left corner was chosen as the reference point because the span of the corresponding grid in the latitude and longitude directions is a positive integer, which facilitates unified coding rules and simple parsing logic. The 5-bit span field can theoretically support up to 99999×99999 grids, meeting the needs of global spatial range expression.

[0031] The fixed-length resolution code is a 5-bit fixed-length code: it adopts the structure of "three integer bits + one decimal bit + one unit character". The unit characters c, m, and k represent centimeters, meters, and kilometers, respectively. The three integer bits and one decimal bit together constitute the resolution value, covering a resolution range from 1 millimeter to 999 kilometers.

[0032] The fixed time stamp is a 14-bit fixed-length code: arranged in the order of 4 bits for year, 2 bits for month, 2 bits for day, 2 bits for hour, 2 bits for minute, and 2 bits for second. If the imaging time information in the metadata is not precise enough, it is padded with 0 bits from the lowest missing bit to 14 bits.

[0033] Variable-length field: Contains sensor type identifier and satellite platform identifier. The reason for using a hyphen to connect them is that the length of sensor / satellite names varies greatly, which facilitates parsing and adaptation to different data sources.

[0034] Specifically, the generation logic of the fixed-length location code is as follows: Based on the four corner coordinates of the image obtained by S1, the number of 16-level GeoSOT grids spanned by the image in the latitude and longitude directions is calculated using a formula. The calculation adopts an upward rounding rule to ensure complete coverage of the image range. The formula is as follows: ; ;

[0035] Among them, lat max with lat min These are the latitude coordinates of the upper right and lower left corners of the image extracted by S1, respectively. max with lon min These are the longitude coordinates of the upper right and lower left corners of the image, respectively. Δlat and Δlon are both side lengths of a 16-level GeoSOT grid, approximately 0.008993°. N lat With N lon The calculation results are represented by 5-digit fixed-length values, with leading zeros added if less than 5 digits. Finally, the 17-digit reference grid code, the 5-digit latitude grid span code, and the 5-digit longitude grid span code are sequentially concatenated to form a fixed-length location code. For example, a 2.2-meter resolution image extracted by S1 corresponds to a resolution code of 0022m; when the imaging date is April 3, 2022, the timestamp code is 20220403000000; when the sensor type is PMS and the satellite platform is GF1B, the variable-length field is PMS-GF1B, and the complete GeoSOT composite code is G001121333013110300019000230022m20220403103025-PMS-GF1B, achieving integrated identification of image space, time, resolution, and source information. This is further combined with... Figure 2 Example of a location code: Based on the spatial range information extracted from S1, the location of the lower left corner of a remote sensing image is determined to be the 16-level GeoSOT grid, code G0011321001002230. Calculations show that the number of grids spanned in the latitude direction is 17 and the number spanned in the longitude direction is 21, corresponding to 5-bit fixed-length codes 00017 and 00021 respectively. By concatenating these three codes, a 27-bit fixed-length location code for the image is obtained: {G0011321001002230}{00017}{00021}, fully reflecting the structural rule of the fixed-length location code: "baseline grid code + number of latitude spans + number of longitude spans".

[0036] S3. Encoding Binding and Multidimensional Index Feature Annotation: The GeoSOT composite code generated in S2 is used as the primary key and bound to the physical storage address of the corresponding remote sensing image and the structured metadata record built based on the metadata in S1. At the same time, the GeoSOT composite code is parsed and combined with the structured metadata to derive multidimensional features for constructing the inverted index. Furthermore, in step S3, the derived multidimensional features include: Using the 16-level GeoSOT grid encoding with fixed-length position codes in the GeoSOT composite encoding generated by S2 as a method, spatial dimension features of at least one spatial grid cell identifier and which can be further derived into upper-level region IDs are obtained. Using a fixed time stamp in the GeoSOT composite code generated by a predefined granularity mapping S2 based on seconds, days, months, and quarters, the time dimension features of the corresponding time slice identifier are obtained. By mapping the values ​​of the fixed-length resolution codes in the GeoSOT composite encoding generated according to S2 to predefined resolution ranges of sub-meter, meter, and ten-meter levels, the resolution dimension features of the corresponding resolution level identifiers are obtained. By parsing the variable-length field in the GeoSOT composite code generated by S2, the sensor dimension features of the sensor type identifier and satellite platform identifier are obtained; The derivation of multidimensional features is accomplished by combining structured metadata constructed based on metadata extracted from S1.

[0037] Specifically, the GeoSOT composite code generated by S2 is used as the primary key to establish a one-to-one binding relationship with the physical storage address of the corresponding remote sensing image and the structured metadata record built based on S1 metadata, so as to realize the association mapping of "encoding-storage address-metadata" and ensure that image data can be quickly located through encoding.

[0038] Simultaneously, by parsing the GeoSOT composite encoding and combining it with structured metadata, multidimensional features for the "keys" of the inverted indexes in various dimensions are derived, specifically including: Spatial dimensional characteristics: Spatial grid cell identifiers obtained by parsing the 16-level GeoSOT grid encoding in the fixed-length position code can be further derived into higher-level region IDs (such as township / county-level aggregated grids). Time dimension features: Time slice identifiers obtained by mapping fixed long-duration stamps to predefined granularities of seconds, days, months, and quarters; Resolution dimension features: Resolution level identifiers obtained by mapping fixed-length resolution code values ​​to predefined resolution ranges of sub-meter, meter, and ten-meter levels; Sensor dimensional characteristics: Sensor type identifier and satellite platform identifier obtained by parsing variable-length fields; The derivation of all multidimensional features is completed in conjunction with the structured metadata constructed by S1, ensuring that the feature values ​​are consistent with the actual attributes of the image. For example, the derived features of the aforementioned GF1B satellite image include: spatial grid cell IDG0011213330131103, upper region ID430124, time slice ID20220403, resolution level ID002, sensor identifier PMS, and satellite identifier GF1B, providing standardized feature dimensions for the subsequent construction of the multidimensional inverted index.

[0039] S4. Multidimensional Inverted Index Construction and Storage: In the distributed search and analysis engine, a document index is constructed using the GeoSOT composite code generated in S2 as the document identifier; corresponding inverted indexes are established for the multidimensional features derived from S3; when image data is added, updated, or deleted, each inverted index is updated synchronously using the GeoSOT composite code as the clue. Furthermore, in step S4, building the corresponding inverted index includes: A spatial inverted index with the spatial grid cell identifier in the spatial dimension feature as the key and the GeoSOT composite code set containing the spatial grid cell identifier as the value; A time inverted index with time slice identifiers in the time dimension features as keys and GeoSOT composite encoding sets containing those time slice identifiers as values; A resolution inverted index with the resolution level identifier in the resolution dimension feature as the key and the GeoSOT composite code set containing that resolution level identifier as the value; A sensor inverted index with the sensor type identifier or satellite platform identifier in the sensor dimension features as the key and the GeoSOT composite code set containing that identifier as the value; Furthermore, in step S4, when constructing the document index, GeoSOT composite encoding is used as the primary key and written into the distributed search engine. The preset length prefix of the fixed-length position code in the encoding is used for sharding routing, so that spatially adjacent grid cells fall into the same or adjacent sharding nodes, reducing the cross-sharding access overhead during spatial queries.

[0040] Specifically, in the distributed search and analysis engine, a document index is constructed using the GeoSOT composite code generated by S2 as the document identifier. The document index is associated with the physical storage address of the image and the structured metadata constructed by S1.

[0041] For the multidimensional features derived from S3, corresponding inverted indexes are created (which can be implemented using a multi-index table structure or a distributed engine multi-field inverted index mechanism), all using a "key-value" structure: Spatial inverted index: The key is the spatial grid cell identifier or the upper-level region ID, and the value is a set of GeoSOT composite codes containing that identifier; Inverted Time Index: The key is the time slice identifier, and the value is the set of GeoSOT composite codes within that time slice; Resolution Inverted Index: The key is the resolution level identifier, and the value is the set of GeoSOT composite codes that fall into that level. Sensor Inverted Index: The key is the sensor type identifier or satellite platform identifier, and the value is the associated GeoSOT composite code set.

[0042] When constructing the document index, the preset length prefix of the fixed-length position code in GeoSOT composite encoding is used as the basis for sharding routing, ensuring that spatially adjacent grid cells fall into the same or adjacent sharding nodes, reducing the cross-sharding access overhead during spatial queries. When image data is added, updated, or deleted, each inverted index is updated synchronously using GeoSOT composite encoding as the clue: when adding, the code is appended under the corresponding key; when updating, the old code is deleted first and then the new code is added; when deleting, the code is removed from the inverted list of all relevant dimensions, ensuring index and data consistency.

[0043] For example, when the aforementioned GF1B satellite imagery is added to the database, its encoding will be added to the spatial key 430124, the time key 20220403, the resolution key 002, the sensor key PMS, and the encoding set corresponding to GF1B, respectively.

[0044] S5. Fast retrieval based on multidimensional inverted index: Parse the user-submitted retrieval request containing spatial range, time interval, resolution conditions and sensor conditions, and map each retrieval condition to the corresponding dimension key; query the inverted index built in S4 based on the dimension key to obtain the candidate GeoSOT composite code set under each dimension; perform set operations, fine filtering and sorting on the candidate code set, and backtrack the document index to obtain the retrieval results.

[0045] Furthermore, in step S5, the search results obtained by backtracking the document index specifically include: Set operations are performed on the candidate GeoSOT composite encoding sets under each dimension, starting with the dimension with the smallest number of candidates, to obtain a preliminary result set; The preliminary result set is finely filtered. The actual coverage of the image is calculated by the number of grids spanned by the latitude and longitude in the fixed long location code. Images that do not completely fall within the target spatial range are removed. Images that do not meet the preset resolution range are filtered by the precise value of the fixed long resolution code. Finally, the images are sorted in chronological order or in reverse order based on the fixed long time stamp to obtain the final code set.

[0046] The document index built by S4 is traced back from the final encoded set to return the matching image data.

[0047] Specifically, the system parses the user-submitted search request, which includes spatial range, time interval, resolution conditions, and sensor conditions, and maps each search condition to a corresponding dimension key: spatial range is mapped to the GeoSOT grid cell identifier or upper-level region ID set covering the range; time interval is mapped to the time slice ID set corresponding to the time granularity; resolution condition is mapped to the resolution level identifier set, preserving precise numerical range constraints; and sensor condition is mapped to the sensor type identifier or satellite platform identifier set.

[0048] Based on the inverted index built by S4 using dimension key queries, candidate GeoSOT composite codes for each dimension are obtained. The size of each set is estimated by combining index statistics. Operations are performed starting from the dimension with the smallest number of candidates: intersection is performed for "AND" relationships, and union is performed on the corresponding dimensions before intersection for "OR" relationships, to obtain a preliminary result set.

[0049] The preliminary result set is finely filtered: the actual coverage area of ​​the image is calculated by the number of grids spanned by the latitude and longitude in the fixed long location code, and images that do not completely fall within the target spatial range are removed; images that do not meet the preset resolution range are filtered by the precise value of the fixed long resolution code; and finally, the final coded set is obtained by sorting the images in chronological order or in reverse order based on the fixed long time stamp.

[0050] By backtracking the document index built by S4 based on the final encoded set, the corresponding physical storage address and structured metadata of the images are retrieved, and a list of matching images is returned. This process transforms complex multidimensional retrieval into one-dimensional index lookup + set operations, significantly reducing retrieval overhead.

[0051] For example, if a user searches for "images of a certain county 430124, April 1-5, 2022, 1-3 meter resolution, PMS sensor, cloud cover ≤20%", the aforementioned GF1B satellite images are processed and filtered before being included in the final results and returned.

[0052] Finally, it should be noted that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A unified labeling method for remote sensing images based on GeoSOT composite coding, characterized in that, Includes the following steps: S1. Remote sensing image metadata extraction: Preprocess the original remote sensing images to extract image spatial range, image resolution, imaging time, sensor type and satellite platform information; S2, GeoSOT composite code generation: Based on the metadata extracted from S1, combined with the GeoSOT grid system, a GeoSOT composite code is generated as a globally unique identifier. S3. Encoding Binding and Multidimensional Index Feature Annotation: The GeoSOT composite code generated in S2 is used as the primary key and bound to the physical storage address of the corresponding remote sensing image and the structured metadata record built based on the metadata in S1. At the same time, the GeoSOT composite code is parsed and combined with the structured metadata to derive multidimensional features for constructing the inverted index. S4. Multidimensional Inverted Index Construction and Storage: In the distributed search and analysis engine, a document index is constructed using the GeoSOT composite code generated in S2 as the document identifier; corresponding inverted indexes are established for the multidimensional features derived from S3; when image data is added, updated, or deleted, each inverted index is updated synchronously using the GeoSOT composite code as the clue. S5. Fast retrieval based on multidimensional inverted index: Parse the user-submitted retrieval request containing spatial range, time interval, resolution conditions and sensor conditions, and map each retrieval condition to the corresponding dimension key; query the inverted index built in S4 based on the dimension key to obtain the candidate GeoSOT composite code set under each dimension; perform set operations, fine filtering and sorting on the candidate code set, and backtrack the document index to obtain the retrieval results.

2. The method for unified identification of remote sensing images based on GeoSOT composite coding according to claim 1, characterized in that, Step S1 also includes: extracting product level and cloud cover information; the image spatial range is defined by the coordinates of the lower left corner and the upper right corner of the image; all extracted metadata is recorded in the structured metadata and serves as the basic data for S2 encoding generation, S3 feature derivation and S4 index construction.

3. The method for unified identification of remote sensing images based on GeoSOT composite coding according to claim 1, characterized in that, In step S2, the generated GeoSOT composite code is combined in a preset order with a fixed-length location code, a fixed-length resolution code, a fixed-length time stamp, and a variable-length field. The fixed-length location code is a 27-bit fixed-length code, which is composed of a 17-bit 16-level GeoSOT grid code located at the lower left corner of the image, a 5-bit latitude grid span code, and a 5-bit longitude grid span code. The spatial resolution of the 16-level GeoSOT grid is 1 kilometer.

4. The method for unified identification of remote sensing images based on GeoSOT composite coding according to claim 3, characterized in that, The fixed-length resolution code is a 5-bit fixed-length code, which consists of three integer bits, one decimal bit, and one unit character. The unit character includes c, m, and k, which represent centimeters, meters, and kilometers, respectively. The three integer bits and one decimal bit together constitute the resolution value. The fixed-length resolution code can represent a resolution ranging from 1 millimeter to 999 kilometers.

5. The method for unified identification of remote sensing images based on GeoSOT composite coding according to claim 3, characterized in that, The fixed time stamp is a 14-bit fixed-length code, with the encoding order being 4 bits for year, 2 bits for month, 2 bits for day, 2 bits for hour, 2 bits for minute, and 2 bits for second; if the imaging time information in the metadata is not precise enough, it is padded with digits 0 from the lowest missing bit to 14 bits.

6. The method for unified identification of remote sensing images based on GeoSOT composite coding according to claim 3, characterized in that, The variable-length field includes a sensor type identifier and a satellite platform identifier, which are connected by a hyphen; the combination order of the GeoSOT composite code is: fixed-length location code, fixed-length resolution code, fixed-length time stamp, and variable-length field.

7. The method for unified identification of remote sensing images based on GeoSOT composite coding according to claim 1, characterized in that, In step S3, the derived multidimensional features include: By parsing the 16-level GeoSOT grid encoding with fixed-length position codes in the GeoSOT composite encoding generated by S2, spatial dimensional features of at least one spatial grid cell identifier and which can be further derived into higher-level region IDs are obtained. The time dimension features of the corresponding time slice identifier are obtained by fixing the long time stamp in the GeoSOT composite code generated by mapping S2 to a predefined granularity of seconds, days, months, and quarters. By mapping the values ​​of the fixed-length resolution codes in the GeoSOT composite encoding generated by S2 to predefined resolution ranges of sub-meter, meter, and ten-meter levels, the resolution dimension features of the corresponding resolution level identifiers are obtained. By parsing the variable-length field in the GeoSOT composite code generated by S2, the sensor dimension features of the sensor type identifier and satellite platform identifier are obtained; The derivation of the multidimensional features is accomplished by combining structured metadata constructed based on metadata extracted from S1.

8. The method for unified identification of remote sensing images based on GeoSOT composite coding according to claim 1, characterized in that, In step S4, establishing the corresponding inverted index includes: A spatial inverted index with the spatial grid cell identifier in the spatial dimension feature as the key and the GeoSOT composite code set containing the spatial grid cell identifier as the value; A time inverted index with time slice identifiers in the time dimension features as keys and GeoSOT composite encoding sets containing those time slice identifiers as values; A resolution inverted index with the resolution level identifier in the resolution dimension feature as the key and the GeoSOT composite code set containing that resolution level identifier as the value; A sensor inverted index with the sensor type identifier or satellite platform identifier from the sensor dimension features as the key and the GeoSOT composite code set containing that identifier as the value.

9. A unified identification method for remote sensing images based on GeoSOT composite coding according to claim 1, characterized in that, In step S4, when constructing the document index, GeoSOT composite encoding is used as the primary key and written into the distributed search engine. The preset length prefix of the fixed-length position code in the encoding is used for sharding routing, so that spatially adjacent grid cells fall into the same or adjacent sharding nodes, reducing the cross-sharding access overhead during spatial queries.

10. A unified identification method for remote sensing images based on GeoSOT composite coding according to claim 1, characterized in that, In step S5, the retrieval results obtained by backtracking the document index specifically include: Set operations are performed on the candidate GeoSOT composite encoding sets under each dimension, starting with the dimension with the smallest number of candidates, to obtain a preliminary result set; The preliminary result set is finely filtered. The actual coverage of the image is calculated by the number of grids spanned by the latitude and longitude in the fixed long location code. Images that do not completely fall within the target spatial range are removed. Images that do not meet the preset resolution range are filtered by the precise value of the fixed long resolution code. Finally, the images are sorted in chronological order or in reverse order based on the fixed long time stamp to obtain the final code set. The document index built by S4 is traced back from the final encoded set to return the matching image data.

Images