Regional equidistant grid coding method based on GeoSOT network

Abstract

The invention discloses a regional equidistant grid coding method based on a GeoSOT network, relates to the technical field of grid coding, defines the structural composition of a grid in detail according to a grid coding rule, and designs coding methods of a G domain, a P domain, a T domain, a polar region identifier and a height domain respectively, so as to ensure that a grid coding system can be compatible with expression requirements of different spatial regions, and improve the coding efficiency. Efficient data storage, retrieval and cross-scale application are achieved, a foundation is laid for application and optimization of a subsequent equidistant grid system, and meanwhile theoretical support is provided for cross-regional and cross-scale spatial data organization and management.

Description

Technical Field

[0001] This invention belongs to the field of grid coding technology, specifically relating to a regional equidistant grid coding method based on GeoSOT networks. Background Technology

[0002] The equidistant grid subdividing coding model is a model developed based on the GeoSOT (Geographic Coordinate Subdividing Grid with One Dimension Integral Coding on a 2n Tree) Earth subdivision framework. Due to its seamless, non-overlapping global latitude and longitude coverage, GeoSOT grids can achieve multi-scale representation and efficient coding computation. Currently, it has been widely used in disaster reduction and emergency response, aerospace remote sensing, land surveying, smart cities, and smart agriculture. It has been incorporated into national and military standards and has played a role in the development of international standards by organizations such as the Open Geospatial Consortium (OGC), the International Organization for Standardization (ISO), and the Institute of Electrical and Electronics Engineers (IEEE).

[0003] The distribution of spatiotemporal data extends from the equator to the poles, covering complex regions across multiple latitudes. However, due to the Earth's spherical shape, existing gridding methods struggle to achieve spatial consistency, especially in equatorial and high-latitude regions where grid cell variations are particularly pronounced. Near the equator, grid cells are relatively regular, but as latitude increases, especially in high-latitude regions, the area and shape of grid cells gradually change (see...). Figure 8 Furthermore, severe deformation or degradation can occur, even near the poles. Globally, various grid-based machine learning and deep learning algorithms require input data to maintain a consistent scale to ensure the model captures uniform and accurate spatial features. However, because multi-source spatial data often employs different coordinate systems, resolutions, and data models, their representations exhibit significant inconsistencies. This regional data disparity often leads to slow convergence and decreased prediction accuracy when AI models trained in different regions are applied across latitudinal dimensions.

[0004] In practical applications, the rigid fixed grid division used in global equidistant grid systems has limitations when processing local areas. In some areas, the lack of flexibility in grid division leads to information redundancy, storing a large amount of data that has no substantial value for the analysis of that area. This not only wastes storage space but also increases the burden of data processing.

[0005] Meanwhile, the data interfaces between various industry systems in the city are extremely complex and lack a unified grid standard. This situation leads to high adaptation costs when conducting joint data analysis. The overlap of latitude and longitude between grids of different systems reduces data accuracy, affecting the accuracy of network planning, optimization, and other business operations, and hindering the improvement of refined management efficiency. The existing local geographic grid division system lacks a unified standard. The geographic grids divided by different modules, systems, and business domains are different, and the grid divisions cannot be aligned, making it difficult to conduct joint analysis and data sharing and exchange. Cross-domain data collaboration faces high adaptation costs and accuracy loss. At present, mainstream geographic grids are mostly used for visualization, usually only storing the latitude and longitude of the grid center point, lacking effective encoding and indexing capabilities, and unable to support efficient spatial computing and analysis.

[0006] The GeoSOT network achieves unified encoding and efficient indexing of global spatial data through fixed-level recursive partitioning, which plays a positive role in large-scale spatial analysis, spatial querying, and data sharing. However, the core advantage of the globally unified model is built on fixed and uniform grid partitioning rules. This regular grid partitioning structure has high efficiency and consistency when dealing with large homogeneous areas, but problems gradually emerge when applied to urban scenarios.

[0007] First, rigid grid systems have fixed grid divisions, which can lead to the same city being placed in two or more grids when identifying administrative divisions, such as the United States National Grid (USNG). For grids with fixed hierarchical recursive partitioning, it is difficult to accurately represent the extent of a region, often resulting in redundancy. Due to its rigid quadtree structure, adjacent grids can only be merged according to the rigid structure. Using this type of grid for spatial object identification presents challenges. Figure 9 (a) Figure 9 (b) Figure 9 In the three cases shown in (c), parent grid A contains child grids A0, A1, A2, and A3, while parent grids B, C, and D are not drawn at all. Although the size of the objects remains unchanged in all three cases, only their positions change. Figure 9 In the case shown in (a), the four sub-grids occupied by the object are merged, so the object can be identified by its parent grid A. And... Figure 9 (b) Figure 9In the case shown in (c), the four sub-grids have no merging relationship, and the existing rigid grid structure and coding system has two ways to identify objects in the latter two cases: one is to use multiple grid codes to identify the same object, and the other is to use a larger grid code to identify the object; the former method will increase the number of identifiers, which is not conducive to large-scale data management and storage; the latter method, although it uses a single code to identify the object, has a low degree of matching between the grid scale and the object, high redundancy, and is not conducive to spatial indexing.

[0008] Figure 9 (b) Figure 9 The situation shown in (c) occurs in many locations around the world, and Figure 1 (a) The probability of this situation occurring is relatively small. For example... Figure 10 As shown, when using quadtree meshing globally, the yellow circles mark locations where four meshes are difficult to merge into a higher-level mesh, while the red circles mark locations where four meshes can be merged into a higher-level mesh. Many meshes can be merged into larger meshes, but cannot be merged into a higher-level non-rigid structure mesh, such as the mesh at the intersection of the equator and the prime meridian in the figure. From Figure 2 It can be observed that although many of the four grids possess the characteristics of being adjacent and at the same level, they are not intermediate nodes in a rigid quadtree structure and therefore cannot be merged. Figure 9 (b) Figure 9 The situation shown in (c) is unavoidable. Furthermore, fixed grid division may lead to large grid gaps in the city center, which is also detrimental to the organization and management of grid data.

[0009] Urban environments are highly complex and diverse, with densely distributed buildings, intricate street networks, and varied terrain. In such contexts, fixed-grid methods may result in insufficient spatial detail representation or data redundancy. Due to the scale limitations of fixed grids, it is impossible to accurately represent details such as buildings and streets, leading to the loss of spatial information. In open areas with sparse data distribution, fixed-grid methods can generate a large number of invalid grids, causing data redundancy and wasting computational resources. By establishing multi-scale spatial data representation models, appropriate scales can be selected for data analysis and processing based on different application requirements. This not only helps improve the accuracy of spatial data representation but also enhances data processing efficiency, meeting the needs of various application levels.

[0010] Therefore, given the complexity and diversity of urban scenarios, it is necessary to adopt methods of non-rigid equidistant regional division and multi-scale representation to improve the accuracy and computational efficiency of spatial data representation. Summary of the Invention

[0011] This invention aims to design a regional equidistant grid coding method (City Equal-Distance Grid, or CEDG) based on the GeoSOT grid as the data foundation, so as to achieve the unification of grid division standards, achieve globally unique coding, break down the barriers between multi-domain data of OBS (Operations Support System), and thus comprehensively improve the performance of spatial analysis.

[0012] To achieve the above goals, CEDG encoding has the following characteristics:

[0013] ① Global Coding Uniformity: This feature allows CEDG to generate unique codes globally, ensuring that every geospatial unit on Earth has a unique identifier. Furthermore, this coding is compatible with current national standards in my country and can be seamlessly integrated into existing geographic information system frameworks. Taking smart city construction as an example, the unified CEDG coding enables the integration of base station coverage data from the telecommunications industry with geographic information data from urban planning departments, providing more comprehensive data support for urban informatization.

[0014] ② Compatibility: Over the long term, different industries have developed their own business systems based on geographic raster data of specific scales, accumulating massive amounts of historical data. For example, grid scales such as 50m×50m and 20m×20m have become widely adopted de facto standards in the telecommunications industry. If a new coding system is incompatible with these scales, existing data will be unable to be effectively utilized, requiring large-scale reconstruction of business systems. This is not only costly but may also affect business continuity and stability. Therefore, to fully utilize existing data assets and reduce transformation costs, CEDG coding must be compatible.

[0015] ③ Reduced modeling and computational overhead: In critical operations such as network optimization and low-altitude flight path planning, equidistant grids are often used for extensive calculations of spatial distances and orientations. Because equidistant grids have regular geometry and consistent spatial resolution, relatively simple and efficient algorithms can be used in these calculations. Compared to non-equidistant grids, they do not require complex distance and angle corrections during computation, thus significantly reducing computational load and time.

[0016] The design of CEDG encoding fully considers the complexity and diversity of geospatial data. Its core idea is to separate the global scope and local details of geospatial objects, and describe them respectively through GeoSOT non-rigid grid encoding based on dimension separation and multi-scale local equidistant grid encoding, thereby achieving efficient spatial data organization and management.

[0017] The design goal of GeoSOT non-rigid grid coding based on dimension separation is to provide unique identifiers for administrative division data of prefecture-level cities nationwide and to address the limitations of traditional coding schemes in single-scale grid coding. Multi-scale local isometric grid coding addresses the shortcomings of traditional coding in multi-scale representation by proposing a multi-scale integer coding scheme. This scheme uses integers of a specific number of bits to uniformly encode grids of different scales, transforming the two-dimensional coding problem into a one-dimensional coding index problem while preserving spatial and scale characteristics, thereby improving the efficiency and adaptability of the coding.

[0018] This invention discloses a regional equidistant grid encoding method based on GeoSOT network, which consists of non-rigid grid encoding and equidistant grid encoding. The non-rigid grid encoding includes a non-rigid grid center domain G and a non-rigid grid scale domain P. The G domain identifies the center of the equidistant region, and the P domain describes the shape of the equidistant region. The equidistant grid encoding is a characterization code of the equidistant grid domain T, representing equidistant grid information of different granularities.

[0019] Specifically, the non-rigid mesh uses administrative divisions as its boundary contour, and the center of the smallest bounding rectangle of this boundary contour is taken as the center point of the corresponding non-rigid mesh region. G-domain encoding is then performed based on this center point. The G-domain encoding method is as follows: obtain the row coordinates I and column coordinates J of the center point of the non-rigid mesh region on the GeoSOT mesh at a specified level, and calculate the G-domain code that forms the non-rigid mesh region.

[0020] Specifically, the P-domain encoding method is as follows: obtain the network levels m and n of the non-rigid grid region in the latitude and longitude directions, and generate the directional scale code Code2 = (m-1)×11 + (n-1), where m and n are less than or equal to the specified level of the GeoSOT grid;

[0021] The network hierarchy along the latitude and longitude is determined in the following ways:

[0022] According to B X Select a grid level m that corresponds to the latitude of administrative division A, and set the latitudinal side length B of the m-level grid. m Condition B is satisfied m+1 <B X ≤B m B m This refers to the mesh size of the GeoSOT mesh at layer m;

[0023] According to L X Choose a grid level n that corresponds to the longitude direction of administrative division A, and let the longitude side length L of the n-level grid be... n Condition L is satisfied n+1 <L X ≤L n L nThis refers to the grid size of the GeoSOT mesh at the nth layer.

[0024] Specifically, the T-domain encoding method is as follows: a cross-coding method using horizontal and vertical coordinates is adopted, with the vertical coordinate encoding first and the horizontal coordinate encoding second to form the Morton cross-coding Code3; then the scale information is encoded into the code according to the multi-scale encoding method, finally forming Code4;

[0025] Additionally, the following methods are provided for converting from Code3 to Code4:

[0026] Step A1: Multiply Code3 by 2 to obtain the intermediate value Code0 = Code3 << 1;

[0027] Step A2: Calculate Code4 corresponding to the initial position in the encoding of level L, i.e. Code40 = 1 << (34 - N << 1) - 1;

[0028] Step A3: Calculate the code corresponding to field T: Code4 = (Code0 >> (35 - N << 1)) << (35 - N << 1) + Code40.

[0029] Preferably, the regional isometric grid coding also includes polar identification coding, which includes North and South Pole identification codes and multi-scale local isometric grid location codes. The polar identification codes are used to identify the geographical locations of the North and South Poles. The multi-scale local isometric grid location codes cover several levels and are used to describe the spatial characteristics and local details of the polar regions.

[0030] The method for converting polar latitude and longitude points (B, L) into polar identification codes is as follows:

[0031] Step B1: Determine the region where the latitude and longitude point is located, i.e., the North and South Poles; when B ≥ 65°, it is the Arctic region; when B ≤ -65°, it is the Antarctic region.

[0032] Step B2: Convert the coordinates (B,L) in the geographic coordinate system to coordinates (x,y) in the orthogonal azimuth projection coordinate system.

[0033] Step B3, based on the grid granularity size and grid level N level Calculate coordinate offset The local coordinates are shifted to the bottom left corner to ensure the encoding is in the first quadrant of the coordinate axis, and the decimal row and column numbers of the grid are obtained (decimal). x ,dec y ),in

[0034] Step B4, convert the decimal row and column numbers (dec) x ,dec yConvert to binary, then perform Morton cross coding with the vertical coordinate encoding first and the horizontal coordinate encoding last to form Code3;

[0035] Step B5: Convert Code3 into the corresponding multi-scale code Code4.

[0036] Preferably, the regional equidistant grid encoding also includes height domain encoding, and the encoding method is as follows: the value range of the height dimension partitioning frame is [h s ,h s +2 n d] uses binary encoding, taking 2 m d(m=n,n-1,n-2,···,1,0) is the nth level partitioning interval, that is, the mth level partitioning divides the height dimension into 2 equal parts. n-m The code length of the partitioned unit is m bits, and the encoding format is Ec1c2...c m-1 c m Where E is the height dimension identifier, c i =0,1.

[0037] Based on the equidistant grid partitioning coding model, this invention focuses on urban scenarios and conducts research around the region-oriented equidistant grid partitioning coding model. It proposes a grid coding method applicable to different geographical regions to meet the needs of multi-scale spatial data representation and efficient computation.

[0038] First, addressing the global non-rigid grid coding problem, this paper studies the structural characteristics of non-rigid grids and designs corresponding coding schemes. Second, regarding local equidistant grid partitioning, an equidistant grid construction method is proposed, and multi-scale coding rules are designed to meet multi-scale representation requirements, including hierarchical calculation rules, grid transformation rules, and hierarchical relationship operations, achieving flexible adaptation across different scales. Subsequently, considering the unique characteristics of polar regions and altitude domain data representation, polar region partitioning methods are studied, and an altitude domain partitioning model is established based on practical application needs, clarifying the partitioning method, key parameter calculations, and applicability analysis. Finally, based on the grid coding rules, the structural composition of the grid is defined in detail, and coding methods for the G domain, P domain, T domain, polar identifier, and altitude domain are designed respectively, ensuring that the grid coding system can be compatible with the representation needs of different spatial regions, achieving efficient data storage, retrieval, and cross-scale applications. The CEDG coding proposed in this invention lays the foundation for the subsequent application and optimization of equidistant grid systems, and provides theoretical support for cross-regional and cross-scale spatial data organization and management. Attached Figure Description

[0039] Figure 1 A schematic diagram of the three Earth expansions of the GeoSOT network;

[0040] Figure 2A schematic diagram of the GeoSOT network's multi-scale quadtree mesh system;

[0041] Figure 3 A schematic diagram of the quadtree grid coding scheme for the GeoSOT network;

[0042] Figure 4 This is a schematic diagram of the CEDG encoding method provided by the present invention;

[0043] Figure 5 A schematic diagram of the horizontal and vertical coordinate cross-coding method adopted for T-field encoding;

[0044] Figure 6 This is a schematic diagram of the height field encoding method;

[0045] Figure 7 A bar chart comparing the coding efficiency of different grids;

[0046] Figure 8 This is a schematic diagram illustrating the grid differences between the equatorial and high-latitude regions.

[0047] Figure 9 A schematic diagram illustrating the object association problem caused by a rigid mesh structure, in which... Figure 9 (a) indicates that objects in the region belong to a parent network. Figure 9 (b) For objects in a region that span two parent networks, Figure 9 (c) For objects in the region, they span four parent networks;

[0048] Figure 10 A schematic diagram showing the locations where rigid meshes cannot completely encompass objects globally.

[0049] Figure 11 This is a schematic diagram showing the geographical distribution of mobile traffic in the Milan and Trentino areas. Figure 11 (a) refers to Milan. Figure 11 (b) is Trentino;

[0050] Figure 12 This is a schematic diagram showing the airspace traffic distribution in the core areas of Shenzhen and Changchun. Figure 12 (a) refers to Shenzhen. Figure 12 (b) is the core area of ​​Changchun;

[0051] Figure 13 A schematic diagram of the basic framework for airspace traffic transfer learning;

[0052] Figure 14 This is a schematic diagram illustrating the height domain encoding of the loss function curve during training and validation. Detailed Implementation

[0053] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. The embodiments of the present invention are given for illustrative and descriptive purposes only, and are not intended to be exhaustive or to limit the invention to the forms disclosed. Many modifications and variations will be apparent to those skilled in the art. The embodiments were chosen and described to better illustrate the principles and practical application of the invention, and to enable those skilled in the art to understand the invention and design various embodiments with various modifications suitable for a particular purpose.

[0054] In the GeoSOT equidistant mesh subdivision coding model, the following key concepts and terms need to be defined:

[0055] ① Equidistant grid (EG): An equidistant grid is a discrete spatial unit formed by the spatial division of a local surface of the Earth, where the distance between the positioning points of the unit is equal. While not a strict geographical definition, in this paper, an equidistant grid refers to a local Cartesian coordinate system established on the Earth's projected surface, with the equator as the X-axis and the projection of the zero meridian or a custom origin meridian as the Y-axis. Within this local coordinate system, a grid with the same span along both coordinate axes is called an equidistant grid.

[0056] ② Non-rigid grid (NGR): A non-rigid grid is a non-fixed grid structure formed by separating the edges of grid cells in different dimensions under the condition that the basic grid division framework remains unchanged.

[0057] ③ Grid level: Used to describe the resolution and accuracy of the grid. Grids at different levels have different sizes and numbers.

[0058] ④ Grid code: A unique encoding rule that identifies each grid, including hierarchical encoding, position encoding, etc.

[0059] In an equidistant grid Earth partition space, the Earth's surface is divided into multi-layered discrete regions of similar area and shape, without gaps or overlaps. Each discrete region is an equidistant grid. Let the grid system... G i ={G i0 G i1 …G ij …G ini} represents all grids at level i, and for any two grids G at level i... is G it (s≠t), all have

[0060] If a geographic entity has a coverage area of ​​E and a level of l, then the entity can be described in the partitioned space as a set of grids at level l. Among them, G lj This represents a mesh with level l, n r This represents the number of grid cells at the r-th level covered by the entity.

[0061] For the isometric grid G ​​encoded as code dl equidistant grid computation model M ld (t) can be defined as

[0062]

[0063] Where d is the spatial dimension d∈{2,3}, t∈T is the time slice, and f θ Assign a computation function to the grid data, depending on the specific computational scenario and algorithm model; θ is the function parameter. It is the i-th type of data (1≤i≤m) participating in the computation in the grid where the time slice t is encoded as code.

[0064] The basic element subdivision coding model of the equidistant grid subdivision coding model has multi-scale characteristics. Let... but and The following relationship exists between them:

[0065]

[0066] Where N represents the highest level of the equidistant mesh partitioning. and These represent the subdivision codes at different subdivision levels (layer and layer+n) under the same latitude and longitude. The above formula indicates that the subdivision grid G... father With G son There is an empty containment relationship between them; and The fact that the bit encoding of the first layer is the same indicates that the partitioning coding has multi-scale properties.

[0067] The GeoSOT meshing framework uses the China Geodetic Coordinate System 2000 (CGCS2000) as the reference ellipsoid. The origin of the coordinate system is located at the intersection of the Earth's reference ellipsoid, the Prime Meridian, and the equatorial plane, providing a unified transformation basis and standard. GeoSOT uses longitude as the X-axis, with east of the Prime Meridian being positive and west being negative; and latitude as the Y-axis, with north of the equator being positive and south being negative. GeoSOT meshes the entire Earth's surface spatially along both meridians and parallels, forming a multi-level grid system covering the globe. Through three Earth extensions, the entire Earth is extended to 512°×512°, 1°×1° is extended to 64′×64′, and 1′×1′ is extended to 64″×64″. See the schematic diagrams for the three spatial extensions respectively. Figure 1 (a) Figure 1 (b) Figure 1 (c) This achieves global, integer-degree, and integer-division quadtree partitioning, forming a multi-scale quadtree grid system from Earth-level (level 0, level G) to centimeter-level surface units (level 32). See [link to relevant documentation]. Figure 2 With each increase in the mesh hierarchy, the positioning accuracy of the mesh also increases, and the accuracy information of the data is implicitly included in the encoding process. Table 1 shows the details of GeoSOT mesh hierarchy and meshing features.

[0068] Table 1

[0069]

[0070]

[0071]

[0072] GeoSOT uses a quadtree recursive partitioning of each grid, and its grid encoding rules are based on the quadtree algorithm and Z-order space-filling curves. Encoding starts with the global level 0 grid (coded as G). Each level divides the current grid into four sub-grids, using the numbers "0", "1", "2", and "3" to represent different sub-regions. As the level increases, the grid encoding becomes progressively more refined, with the encoding format being a progressively increasing string of numbers. During the encoding process, the coordinates of the sub-regions are alternately arranged in a latitude-first, longitude-last order, forming a quaternary one-dimensional Morton code sequence. The 32-level GeoSOT grid encoding format is Gddddddddd-mmmmmm-ssssss.uuuuuuuuuuu, and the quadtree grid encoding method is as follows: Figure 3 As shown.

[0073] Example 1

[0074] The CEDG encoding based on the GeoSOT network provided in this embodiment can be found in [reference needed]. Figure 4 It consists of non-rigid grid coding and equidistant grid coding.

[0075] The non-rigid mesh encoding includes a non-rigid mesh center domain G and a non-rigid mesh scale domain P. The G domain identifies the center of the equidistant region formed by the non-rigid mesh, and the P domain describes the shape of the equidistant region formed by the non-rigid mesh.

[0076] The non-rigid mesh here refers to a hierarchical mesh structure that replaces the rigid tree structure between mesh layers with a tree structure with variable hierarchical relationships, under the condition that the basic framework of mesh generation remains unchanged. Based on the grid cell, a multi-type grid cell that spans grid points and grid edges is formed, and a unified coding method is used for identification and calculation.

[0077] The advantages of non-rigid grids are twofold: First, rigid grid systems have fixed grid divisions, which can lead to the same city being placed in two or more grids (such as the MGRS grid system in the United States) when identifying administrative divisions, causing inconvenience in use, such as data conversion and analysis. Second, given the diverse nature of global administrative division data, a more flexible structure is needed. This structure should be able to use a single code to uniquely identify prefecture-level city administrative division data within a unified global grid framework. Furthermore, the code should have spatial attributes, allowing for rapid range calculations and spatial relationship calculations through code operations.

[0078] I. G-domain encoding

[0079] When identifying administrative divisions, due to the diversity of their shapes and the inherent regularity of the grid itself, duplicate encoding is inevitable when identifying their boundary contours, especially for administrative division data with nested relationships. Therefore, in this embodiment, the non-rigid grid uses the administrative division as its boundary contour, with the center of the smallest bounding rectangle of that boundary contour as the center point of the corresponding non-rigid grid region, and G-domain encoding is performed based on this center point. Simultaneously, two dimensions are designed for encoding in different dimensional directions to control the boundaries of the administrative division data.

[0080] The G-domain encoding method provided in this embodiment is as follows: obtain the row coordinates I and column coordinates J of the center point of the non-rigid mesh region on the 11th layer of the GeoSOT mesh, and calculate the G-domain encoding that forms the non-rigid mesh region. In other words, the G-domain encoding uses the subdivision result of GeoSOT level 11, which is 22 bits in total.

[0081] The calculation method for the partitioned range is as follows:

[0082] 1. Find the bounding rectangle: Calculate the minimum bounding rectangle Box of administrative division A, and calculate the length B of Box in both the latitude and longitude directions. X and L X The center of Box is (B0, L0), where B0 and L0 are the latitude and longitude of the center of Box, respectively.

[0083] 2. Determining the dimensional direction hierarchy:

[0084] According to B X Select a grid level m that corresponds to the latitude of administrative division A, and set the latitudinal side length B of the m-level grid. m Condition B is satisfied m+1 <B X ≤B m B m This refers to the mesh size of the GeoSOT mesh at layer m;

[0085] According to L X Choose a grid level n that corresponds to the longitude direction of administrative division A, and let the longitude side length L of the n-level grid be... n Condition L is satisfied n+1 <L X ≤L n L n This refers to the grid size of the GeoSOT mesh at the nth layer.

[0086] 3. Determining the center of the non-rigid grid region: In the latitudinal direction, calculate the column coordinate J of B0 in the GeoSOT specified level grid; in the longitude direction, calculate the row coordinate I of L0 in the GeoSOT specified level grid.

[0087] 4. Determining the grid level: Calculate the coverage area of ​​the non-rigid grid based on the center grid coordinates (I,J) and the corresponding grid level (m,n) of the non-rigid grid area to ensure that the grid can achieve full coverage of the administrative division.

[0088] The method for determining the latitude direction hierarchy is as follows:

[0089] Step a, calculate the latitude center point coordinates B corresponding to the grid coordinates (I,J). center ;

[0090] Step b: Calculate the initial coverage area (B) based on the center point coordinates and layer m of the grid. min B max ), where B min =B center -B m / 2, B max =B center +B m / 2;

[0091] Step c, determine the coverage relationship between the initial grid and the administrative divisions:

[0092] If B min <B0-B X / 2, and B max >B0+B X If the value is 2, then the grid covers the administrative division, and m is the final grid level; otherwise, m = m-1, and the coverage judgment continues, with the condition being the primary one, and the final grid level m is determined through iterative judgment.

[0093] The method for determining the longitude level is the same as that for the latitude level.

[0094] 5. Determination of orientation scale code: Based on the obtained grid levels m and n, generate orientation scale code code2 = (m-1)×11 + (n-1).

[0095] 6. Generation of non-rigid grid encoding: Based on the region center, the latitudinal side length B m Longitude side length L n The defined region is the range of the Box region CEDG non-rigid mesh.

[0096] The advantages of the above G-field encoding method are:

[0097] First, uniqueness. Since the central point of all administrative division data is different, a coding mapping relationship based on the central point of the administrative division data is established. This mechanism ensures that each administrative division unit receives a unique identifier, effectively preventing duplicate coding.

[0098] Secondly, regarding accuracy, thanks to the centimeter-level minimum resolution characteristic of the GeoSOT encoding system, theoretically, the center point calculated by this method can be infinitely close to the actual center point, achieving centimeter-level precision. However, considering the optimization requirements of the encoding system design, this embodiment, through the overall design of CEDG encoding, uses only 22 bits for center point encoding and identification, effectively controlling the center point error within the range of 0.087°. This error value is the maximum value at the 11th layer grid scale of GeoSOT, and is usually even smaller in practical applications.

[0099] II. P-field encoding

[0100] In the scale direction design (i.e., the P-domain), an 11-layer coding structure is adopted, generating 121 combinations corresponding to a 7-bit coding space. This has good consistency with the actual situation, ensuring both coding accuracy and optimizing coding efficiency. The P-domain coding method provided in this embodiment is as follows: Obtain the network levels m and n of the non-rigid grid region in the latitude and longitude directions, and generate the directional scale code Code2 = (m-1)×11 + (n-1), where m and n are less than or equal to the specified level of the GeoSOT grid. Since the specified level in this embodiment is 11, the values ​​of m and n are both integers between 1 and 11.

[0101] The network hierarchy along the latitude and longitude is determined in the following ways:

[0102] According to B X Select a grid level m that corresponds to the latitude of administrative division A, and set the latitudinal side length B of the m-level grid. m Condition B is satisfied m+1 <B X ≤B m B m This refers to the mesh size of the GeoSOT mesh at layer m;

[0103] According to L X Choose a grid level n that corresponds to the longitude direction of administrative division A, and let the longitude side length L of the n-level grid be... n Condition L is satisfied n+1 <L X ≤L n L n This refers to the grid size of the GeoSOT mesh at the nth layer.

[0104] III. T-field encoding

[0105] The equidistant grid is encoded as a representation code for the equidistant grid field T, representing equidistant grid information at different granularities. The T-field encoding method uses a cross-coding approach based on horizontal and vertical coordinates; see [link to relevant documentation]. Figure 5 The Morton cross-coding Code3 is formed by encoding the Y-axis first and the X-axis second, with a total of 17 layers and 34 bits. Then, the scale information is encoded into the code according to the multi-scale encoding method, and finally the Code4 is formed.

[0106] The method for converting from Code 3 to Code 4 is as follows:

[0107] 1. Multiply Code3 by 2 to get the intermediate value Code0 = Code3 << 1.

[0108] 2. Calculate the Code4 corresponding to the initial position in the L-th level encoding, i.e., Code40 = 1 << (34 - N << 1) - 1.

[0109] 3. Calculate the code corresponding to field T: Code4 = (Code0 >> (35 - N << 1)) << (35 - N << 1) + Code40.

[0110] Example 2

[0111] Since there are no administrative divisions (including the North and South Poles) in the Arctic and Antarctic regions, this embodiment, based on the isometric grid coding of administrative divisions completed in Embodiment 1, uses a unified polar grid for coding and identification. The polar identification code includes Arctic and Antarctic identification codes and multi-scale local isometric grid location codes. The polar identification code is used to identify the geographical location of the Arctic or Antarctic; the multi-scale local isometric grid location code covers several levels and is used to describe the spatial characteristics and local details of the polar regions.

[0112] The polar identification encoding provided in this embodiment uses the following rules: a total of 64 binary bits, the first 23 bits are the North and South pole identification bits, when the first 23 bits are all 1, it represents the Arctic region, when the first 23 bits are all 0, it represents the Arctic region, and the last 41 binary bits are the polar region multi-scale local equidistant grid location code.

[0113] The method for converting polar latitude and longitude points (B, L) into polar identification codes is as follows:

[0114] 1. Determine the region where the latitude and longitude point is located, i.e., the North and South Poles; when B ≥ 65°, it is the Arctic region; when B ≤ -65°, it is the Antarctic region.

[0115] 2. Convert the coordinates (B, L) in the geographic coordinate system to coordinates (x, y) in the orthogonal azimuth projection coordinate system. The conversion formula is as follows:

[0116] x = R × cos(B) × cos(L),

[0117]

[0118] Where R is the average radius of the Earth, which is taken as 6,371,000 m in this embodiment for polar calculations, B is latitude, and L is longitude.

[0119] 3. Based on the granularity of the grid size and grid level N level Calculate coordinate offset The local coordinates are shifted to the bottom left corner to ensure the encoding is in the first quadrant of the coordinate axis, and the decimal row and column numbers of the grid are obtained (decimal). x ,dec y ), as shown in the following formula:

[0120]

[0121] 4. Convert the decimal row and column numbers (dec)x ,dec y Convert to binary, according to Figure 5 Morton cross-coding, which uses the method of first encoding the vertical coordinate and then encoding the horizontal coordinate, is used to form Code3;

[0122] 5. Convert Code3 into the corresponding multi-scale code Code4, in the same way as in Example 1.

[0123] Example 3

[0124] Examples 1 and 2 complete the encoding of the planar domain, and this example provides a height domain encoding scheme.

[0125] The range of values ​​for the height-dimensional partitioning framework is [h] s ,h s +2 s d] (unit: m), using binary encoding, taking 2 t d(t=s,s-1,s-2,···,1,0) is the (st)th level partitioning interval, that is, the t-th level partitioning divides the height dimension into 2 equal parts. s-t The code length of the partitioned unit is m bits, and the encoding format is Ec1c2...c t-1 c t Where E is the height dimension identifier, c i =0,1. To align with the centimeter-level accuracy achieved after partitioning the GeoSOT framework, d = 2 is chosen. -5 =0.03125m; Since the application objects of the three-dimensional mesh proposed in this embodiment are mainly geospatial entity object information, considering the scope of human research on geographic objects, h s = -1000m, n = 18, then the height range is [-1000m, 7192m]. The grid system contains 1m grids, which facilitates the direct conversion of relative elevation coordinates into binary code.

[0126] The height-based mesh size 'd' can also be selected by the user to match the planar mesh. For the minimum unit size 'S' required by the user on the plane, in low-altitude airspace below 3000m, 'd' can be set to S, while h... s = -1000m, n≥ln(3000 / S) / ln2. If the user requires the minimum scale of the equidistant planar grid to be 5m, then d=5m, h s = -1000m, n = 10, altitude range is [-1000m, 5120m].

[0127] As the scope of height dimension research expands to deep space, the height dimension encoding of the equidistant grid also changes. The value range of the height dimension partitioning framework of the equidistant grid is [h]. s ,h s+2 s d]. If it is necessary to extend the range of altitude values ​​to deep space to d×2 n+k Simply add k zeros after the original E identifier, such as E000c1c2...c t-1 c t This is the encoding after extending the original partitioned space to deep space by 8 times its height.

[0128] The following experiments verify the encoding efficiency and decoding efficiency of the CEDG encoding provided by this invention.

[0129] I. Experimental Environment

[0130] First, the hardware and software environment configurations used in all experiments are described to ensure the reproducibility of experimental results and the efficiency of data processing. All experiments will be conducted in the same computing environment to ensure fairness and consistency. Detailed hardware and software environment configurations are shown in Table 2. This environment configuration supports large-scale data processing and the training of complex deep learning models, guaranteeing the stability and efficiency of the experiments.

[0131] Table 2

[0132] Hardware / Software Environment parameter processor Apple M2 Pro Memory 32GB harddisk 1TB operating system MacOSSequoia compiler Python 3.8, Matlab, Java Deep learning framework PyTorch 2.2.1 Visualization QGIS, Cesium 1.124

[0133] II. Experimental Data

[0134] The experimental data primarily comes from OpenStreetMap (OSM, https: / / www.openstreetmap.org / ), a platform that provides global, multi-scale, and multi-type open-source spatial data, covering vector elements such as urban road networks, building boundaries, water systems, and terrain. The data format is Shapefile or GeoJSON, and the spatial reference system is WGS84. The data obtained from the OpenStreetMap platform mainly includes the following three categories:

[0135] (1) Point elements: Urban points of interest (POIs) such as restaurants, shops, hospitals, schools, and parks; public facilities such as libraries, museums, and fire stations; and landmark buildings such as monuments and towers.

[0136] (2) Line elements: road networks such as highways, urban main roads, streets and alleys, railways, subways, light rails, etc.;

[0137] (3) Surface elements: residential buildings, office buildings and other buildings, lakes, wetlands, farmland and other land plots, and the scope of national, provincial, municipal and district administrative divisions.

[0138] III. Efficiency Analysis of Equidistant Grid Encoding and Decoding

[0139] The efficiency of the spatial coding algorithm was evaluated experimentally, focusing on the time consumption and accuracy of the encoding and decoding processes. For any point P (116.7089779°E, 40.53906045°N) on the Earth's surface, a 20m equidistant grid scale was selected to generate a grid CEDG code: 1101101110000110010110100011111000000111010000011010011001001111. The coding efficiency test aimed to evaluate the time consumption of converting latitude and longitude data into CEDG grid codes of different scales, while the decoding efficiency test evaluated the time consumption of decoding CEDG grid codes of different scales into latitude and longitude data. In the experiment, the encoding and decoding performance were tested under different data sizes (10,000, 50,000, 100,000, 200,000, and 500,000 latitude and longitude data) and different grid sizes (10m, 20m, 50m, and 100m). The encoding and decoding times are shown in Table 3 and Table 4, respectively.

[0140] Table 3

[0141]

[0142] Table 4

[0143]

[0144] The results in Tables 3 and 4 show that the time consumption for equidistant grid encoding and decoding exhibits a significant variation trend under different grid scales and data rows.

[0145] First, the encoding and decoding times increase with the increase in data volume and the decrease in grid granularity. During encoding, smaller grid scales (e.g., 5m) require longer encoding times, resulting in greater memory usage and more computational resources. For example, when processing 10,000 data entries, the encoding time for a 50m grid is 16.123ms, while for a 5m grid it is 24.215ms, an increase of approximately 50%. When the number of data entries reaches 500,000, the encoding time increases from 16.123ms to 495.190ms, an increase of nearly 30 times. Similarly, the decoding time also increases with the increase in data volume. With a 50ms grid, the decoding time is 26.545ms, and as the number of data entries increases to 500,000, the decoding time increases to 419.746ms, a relatively gradual increase compared to the encoding time.

[0146] Overall, although the encoding and decoding time increases with the amount of data, the equidistant grid system still exhibits relatively high performance in processing large-scale data, especially when choosing an appropriate grid scale. Comparing the encoding and decoding times of grids with different scales reveals that with smaller data volumes, 5m or 10m grids offer a good balance between accuracy and processing speed, while in large-scale scenarios, a 50m grid effectively reduces processing time. Encoding 500,000 data points at a 5m scale takes less than 1 second, and the overall processing time remains within an acceptable range, demonstrating the algorithm's efficiency in large-scale data processing and spatial data transformation. Future research could focus on optimizing algorithm performance, such as introducing parallel processing or distributed computing, to further improve processing speed, especially in complex scenarios involving massive amounts of data and complex spatial objects.

[0147] To evaluate the decoding accuracy of equidistant grid coding, this experiment randomly generated 10,000 points globally and performed encoding and decoding operations on these points at grid scales of 5m, 10m, 20m, and 50m, respectively. During decoding, the generated CEDG codes were converted back to latitude and longitude coordinates, and the Euclidean distance between the decoded coordinates and the original coordinates was calculated as the decoding error. By analyzing the maximum error, average error, and 95% confidence interval, a comprehensive understanding of the decoding accuracy performance of CEDG coding at different grid scales can be obtained, as shown in Table 5.

[0148] Table 5

[0149] Grid Scale Maximum error average error 95% confidence interval 5 3.510 1.902 (1.888，1.916) 10 7.034 3.817 (3.789，3.845) 20 14.006 7.708 (7.652，7.764) 50 35.023 19.287 (19.148，19.425)

[0150] Theoretically, the coordinate information in the equidistant grid encoding process contains a certain quantization error, causing the decoded coordinate points to deviate from the original coordinate points. In a two-dimensional plane, the theoretical maximum error should be half the grid scale. The experimental results show that the maximum error measured in the experiment is basically consistent with the theoretical calculation value, and the decoding accuracy is closely related to the grid scale. As the grid scale increases, the average error shows a linear growth trend. The average error at different grid scales is much smaller than the maximum error. At a 5m grid scale, the maximum error is 3.510m, the average error is 1.902m, and 95% of the error values ​​fall between 1.888m and 1.916m. At a 50m grid scale, the maximum error increases to 35.023m, the average error is 19.287m, and the 95% confidence interval is (19.148, 19.425)m. The 95% confidence intervals for all experimental groups are relatively narrow, indicating a stable error distribution and a high confidence level for the mean error.

[0151] Furthermore, by comparing the performance of different grid coding methods (GeoSOT-P, GeoSOT, and CEDG) under different data volumes, the efficiency of coding in large-scale data processing is evaluated, and the results are as follows: Figure 7 As shown, CEDG encoding is more efficient than other grid coding methods when processing large-scale data.

[0152] Compared to the GeoSOT-P grid, CEDG encoding is significantly faster, especially when processing large datasets. For example, with 500,000 data points, GeoSOT-P encoding took 10,147.3 ms, while CEDG took only 542.2 ms, a difference of nearly 19 times. This demonstrates the advantage of CEDG encoding in large-scale data processing.

[0153] Compared to the GeoSOT grid, CEDG encoding takes slightly longer, but the difference is relatively small, especially when processing large amounts of data, where CEDG's encoding efficiency remains relatively high. Overall, the CEDG grid demonstrates significant efficiency in large-scale data processing, particularly compared to the GeoSOT-P grid, where its encoding time is significantly reduced, fully demonstrating its potential in large-scale data applications.

[0154] IV. Spatial Object Encoding Efficiency Analysis

[0155] This experiment aims to evaluate the effectiveness of the CEDG grid spatial object encoding method in real-world urban environments, particularly its encoding efficiency and accuracy when handling line and polygon objects. Data is sourced from OpenStreetMap (https: / / www.openstreetmap.org / ). By encoding this spatial data, the performance of the CEDG grid encoding method can be evaluated under different regions and data scales.

[0156] To ensure the reliability of the encoding results, we performed necessary data cleaning and preprocessing to ensure uniform format and accurate precision. Encoding efficiency tests were conducted in low-latitude (Singapore), mid-latitude (Guangdong Province, Shanghai, Beijing, and China as a whole), and high-latitude (Iceland) regions. By selecting data from different regions, we tested the encoding efficiency of the CEDG grid at different regional scales, verifying its processing capabilities under different geographical ranges and data volumes. The results are shown in Table 6.

[0157] Table 6

[0158]

[0159] This experiment, through a systematic analysis of the encoding time of spatial objects in low, mid, and high latitude regions, reveals the impact of data volume, spatial object type, grid scale, and regional differences on encoding efficiency. Data volume is the primary factor determining encoding time. At the same grid scale, the encoding time for equidistant grids in mid-latitude regions (such as Guangdong Province, Shanghai, and Beijing in China) increases with increasing data volume. As the encoding granularity increases, the number of nodes and boundaries the algorithm needs to process decreases, thus significantly improving computational efficiency. Nationwide, encoding a 5m grid takes 733.859 seconds, while a 50m grid takes only 119.428 seconds, a reduction of nearly 84%. Encoding time for polygonal objects is generally lower than that for linear objects. Taking Beijing as an example, encoding a 5m precision linear object takes 13.324 seconds, while encoding a polygonal object takes only 6.463 seconds. This is because linear data involves more complex spatial calculations and topological relationship processing during gridding, while urban polygonal objects typically have more regular geometric structures, reducing the complexity of the algorithm when processing boundaries and nodes. In low-data-density regions (such as Singapore and Iceland), changes in grid granularity have a relatively small impact on processing time. The choice of encoding granularity should be optimized based on the specific application scenario and data distribution to balance accuracy with computational resource consumption.

[0160] The following provides an application experiment of CEDG encoding.

[0161] I. Experimental Objective

[0162] With the widespread application of artificial intelligence technology, sufficient data support has become a key factor in improving model performance. However, differences in economic development levels and infrastructure construction among cities have led to a severe imbalance in the spatial distribution of data. For example, consider the distribution of mobile traffic in Italy (see...). Figure 11 Developed cities like Milan possess well-developed mobile communication infrastructure and abundant traffic flow data, while less developed regions like Trentino face data scarcity due to insufficient data collection equipment and low user density. This uneven distribution of data has led current research to focus primarily on data-rich developed cities, neglecting the actual needs of data-scarce regions, thus hindering the application of artificial intelligence technology across a wider geographical area.

[0163] Transfer learning, by transferring general knowledge learned in a source city (data-rich region) to a target city (data-scarce region), can significantly reduce the dependence on local data in the target city. Theoretically, there are certain similarities in traffic flow patterns between different cities; for example, traffic demand in commercial and residential areas exhibits cyclical changes. This consistency in spatial characteristics provides a theoretical basis for cross-regional transfer learning. However, in cross-latitudinal transfer learning, non-equidistant data inputs can cause convergence problems due to grid deformation, affecting the model's generalization performance and prediction accuracy. Equidistant grids, by maintaining consistent distances between grid cells, can maintain spatial scale uniformity globally, effectively mitigating the convergence problem caused by grid deformation.

[0164] The core objective of this experiment is to evaluate the feasibility of applying equidistant grids in cross-regional and cross-latitudinal transfer learning. Therefore, this experiment focuses on the impact of grid design on transfer learning tasks, rather than the optimization of the transfer learning algorithm itself.

[0165] II. Experiment Content

[0166] The experiment employed a rigorous ablation experimental design, systematically evaluating the applicability and effectiveness of equidistant grids in cross-regional and cross-latitude transfer learning by comparing three different input methods: no grid, GeoSOT grid, and CEDG equidistant grid. Low-altitude airspace traffic data from Shenzhen (around 22°N, a low-latitude region) was used for model training, and the data was then transferred to Changchun (around 44°N, a high-latitude region) for testing and fine-tuning. Based on the distribution characteristics of urban hotspots, this experiment generated the necessary data through simulation.

[0167] First, the airspace traffic data is processed in three ways:

[0168] 1. Gridless input: Directly input the raw airspace flow point data into the model.

[0169] 2. GeoSOT grid input: Based on the GeoSOT grid, the data is mapped to regularized latitude and longitude grid cells, and the flow data is statistically analyzed for each grid cell.

[0170] 3. CEDG equidistant grid input: The Shenzhen area is uniformly divided using a CEDG grid, and the flow characteristics of each grid cell are statistically analyzed (see...). Figure 12 The design of the equidistant grid aims to solve the deformation problem of traditional latitude and longitude grids in high-latitude regions caused by the shortening of latitude circles.

[0171] To evaluate the effectiveness of equidistant grids in low-altitude airspace traffic forecasting, this experiment used gridless, GeoSOT, and CEDG equidistant grids as inputs to conduct the following tasks (see [link]). Figure 13 ):

[0172] Source City Model Training: A regression model based on a convolutional neural network is trained on the source domain (low-altitude airspace data of Shenzhen) to learn the features of the source domain data. Features are extracted from the model trained on the source domain and used for prediction in the target domain (data of Changchun).

[0173] Target City Model Transfer: The model trained in the source city is transferred to the target domain. Fine-tuning is performed using a small amount of data from the target city. This can be done by freezing only some layers of the model and training only the last few layers, or by fine-tuning the entire model. The model is then trained in the target domain, parameters are adjusted, and the transfer performance is tested. Mean Squared Error (MSE) is chosen as the loss function.

[0174] III. Results Analysis

[0175] The low-altitude airspace traffic model from Shenzhen was transferred to Changchun for fine-tuning training. The performance of the transferred model is shown in Table 7. For input without a grid, the prediction performance of the transferred model in the target domain decreased significantly due to the lack of spatial partitioning support, resulting in large RMSE and MAE values. When GeoSOT grid was used as input, the performance improved, but the model's adaptability in the target domain was limited due to the inconsistency in cross-latitude accuracy of the grid partitioning. However, the performance after transfer based on CEDG grid was excellent. Thanks to the cross-latitude consistency of the equidistant grid, the model can effectively learn and transfer the spatial distribution patterns of low-altitude airspace traffic.

[0176] Table 7

[0177]

[0178] Figure 14 This is the decline curve of the loss function during training and validation. In the early stages of training, the model loss decreases rapidly, indicating that the equidistant grid significantly improves learning efficiency and helps the model quickly capture the characteristics of low-altitude airspace traffic. In the middle stage, the rate of decrease in the loss value gradually decreases, and the model begins to stably learn the characteristics of regional traffic patterns. In the later stage (30 training epochs), the loss value tends to converge, with the training loss and test loss stabilizing at 0.17 and 0.23, respectively, indicating that the model achieves good generalization performance in the target domain without significant overfitting. Meanwhile, although fluctuations in the test set loss occur occasionally, the overall error remains at a low level, further validating the spatial consistency and transfer performance of the equidistant grid. Compared to traditional grid partitioning methods, the equidistant grid avoids the introduction of accuracy errors in cross-latitude scenarios, providing a solid spatial foundation for efficient learning and transfer of traffic patterns.

[0179] The CEDG grid has demonstrated significant advantages in many aspects of cross-latitude applications, and its characteristics and performance have been fully verified.

[0180] First, spatial consistency is a major highlight of the CEDG grid. It maintains consistent grid division across latitudinal applications, effectively avoiding the accuracy errors introduced by other grids during latitudinal transformations. Traditional grid systems may exhibit shape or area deviations across different latitudinal regions due to geographical span, while the CEDG grid ensures spatial consistency through a finely defined equidistant division strategy, thus accurately reflecting actual spatial characteristics in both high and low latitude regions.

[0181] Secondly, the CEDG grid significantly improves the efficiency of model learning traffic patterns by providing a unified spatial partition. Experiments show that models based on the CEDG grid can more effectively capture traffic characteristics between regions, enabling models trained in source cities to maintain strong generalization ability after being transferred to target cities. The spatial consistency of the grid reduces the complexity of the model during training and transfer, and reduces learning bias caused by scale inconsistencies, ultimately achieving a significant performance improvement.

[0182] Furthermore, the CEDG grid exhibits excellent transfer performance. Experimental results show that the transfer learning model based on the CEDG grid significantly outperforms other grids in the target domain. The transferred model demonstrates lower errors (such as lower RMSE and MAE) and higher stability in the target city, indicating that the CEDG grid can more effectively retain and transfer knowledge from the source domain, thereby achieving more accurate traffic prediction for the target domain.

[0183] Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art and related fields based on the embodiments of the present invention without inventive effort should fall within the scope of protection of the present invention.

Claims

1. A regional equidistant grid coding method based on GeoSOT networks, characterized in that, It consists of non-rigid grid coding and equidistant grid coding; the non-rigid grid coding includes a non-rigid grid center domain G and a non-rigid grid scale domain P, the G domain identifies the center of the equidistant region, and the P domain describes the shape of the equidistant region. The equidistant grid is encoded as a characterization code for the equidistant grid domain T, representing equidistant grid information at different granularities.

2. The regional equidistant grid coding method based on GeoSOT network according to claim 1, characterized in that, The non-rigid mesh uses administrative divisions as its boundary contour, and the center of the smallest outer rectangle of the boundary contour is taken as the center point of the corresponding non-rigid mesh region. G-domain encoding is performed based on this center point.

3. The regional equidistant grid coding method based on GeoSOT network according to claim 2, characterized in that, G-domain encoding method: Obtain the row coordinates I and column coordinates J of the center point of the non-rigid mesh region on the specified level of the GeoSOT mesh, and calculate the G-domain encoding that forms the non-rigid mesh region.

4. The regional equidistant grid coding method based on GeoSOT network according to claim 1, characterized in that, P-domain encoding method: Obtain the network level m and n of the non-rigid grid region in the latitude and longitude directions, and generate the directional scale code Code2 = (m-1)×11+(n-1), where m and n are less than or equal to the specified level of the GeoSOT grid; The network hierarchy along the latitude and longitude is determined in the following ways: According to B X Select a grid level m that corresponds to the latitude of administrative division A, and set the latitudinal side length B of the m-level grid. m Condition B is satisfied m+1 <B X ≤B m B m This refers to the mesh size of the GeoSOT mesh at layer m; According to L X Choose a grid level n that corresponds to the longitude direction of administrative division A, and let the longitude side length L of the n-level grid be... n Condition L is satisfied n+1 <L X ≤L n L n This refers to the grid size of the GeoSOT mesh at the nth layer.

5. The regional equidistant grid coding method based on GeoSOT network according to claim 1, characterized in that, T-domain encoding method: The horizontal and vertical coordinates are cross-coded, with the vertical coordinates encoded first and the horizontal coordinates encoded later to form Morton cross-coding Code3; then the scale information is encoded into the code according to the multi-scale encoding method, and finally Code4 is formed.

6. The regional equidistant grid coding method based on GeoSOT network according to claim 5, characterized in that, The method for converting from Code 3 to Code 4 is as follows: Step A1: Multiply Code3 by 2 to obtain the intermediate value Code0 = Code3 << 1; Step A2: Calculate Code4 corresponding to the initial position in the encoding of level L, i.e. Code40 = 1 << (34 - N << 1) - 1; Step A3: Calculate the code corresponding to field T: Code4 = (Code0 >> (35 - N << 1)) << (35 - N << 1) + Code40.

7. The regional equidistant grid coding method based on GeoSOT network according to claim 1, characterized in that, It also includes polar identification coding, which includes North and South Pole identification codes and multi-scale local isometric grid location codes. The polar identification coding is used to identify the geographical location of the North or South Pole; the multi-scale local isometric grid location coding covers several levels and is used to describe the spatial characteristics and local details of the polar regions.

8. The regional equidistant grid coding method based on GeoSOT network according to claim 7, characterized in that, The method for converting polar latitude and longitude points (B, L) into polar identification codes is as follows: Step B1: Determine the region where the latitude and longitude point is located, i.e., the North and South Poles; when B ≥ 65°, it is the Arctic region; when B ≤ -65°, it is the Antarctic region. Step B2: Convert the coordinates (B,L) in the geographic coordinate system to coordinates (x,y) in the orthogonal azimuth projection coordinate system. Step B3, based on the grid granularity size and grid level N level Calculate coordinate offset The local coordinates are shifted to the bottom left corner to ensure the encoding is in the first quadrant of the coordinate axis, and the decimal row and column numbers of the grid are obtained (decimal). x ,dec y ),in Step B4, convert the decimal row and column numbers (dec) x ,dec y Convert to binary, then perform Morton cross coding with the vertical coordinate encoding first and the horizontal coordinate encoding last to form Code3; Step B5: Convert Code3 into the corresponding multi-scale code Code4.

9. The regional equidistant grid coding method based on GeoSOT network according to claim 1, characterized in that, It also includes height field encoding, the encoding method being: taking the value range of the height dimension partitioning frame as [h s ,h s +2 s d] uses binary encoding, taking 2 t d(t=s,s-1,s-2,···,1,0) is the (st)th level partitioning interval, that is, the t-th level partitioning divides the height dimension into 2 equal parts. s -t The code length of the partitioned unit is m bits, and the encoding format is Ec1c2...c t-1 c t Where E is the height dimension identifier, c i =0,1.

10. The application of the GeoSOT network-based regional isometric grid coding method according to any one of claims 1-9 in cross-regional transfer learning.

Images