Hexagonal pixel remote sensing image storage method with compatibility
By mapping spherical hexagonal pixel data to the icosahedral surface and chunked storage, the problem of hexagonal pixel data storage and sharing is solved, and efficient data processing and analysis is realized, suitable for standard GIS/RS software.
Patent Information
- Application Number
- CN202311145448.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-06
- Publication Date
- 2024-11-05
AI Technical Summary
The prior art is difficult to effectively store and share hexagonal pixel remote sensing image data, and the spherical surface is not easy to directly establish a geometric mapping relationship between hexagons and rectangles, resulting in limited application and promotion of data.
Icosahedral hexagonal global discrete grid is used as the bearing frame for remote sensing image data, and equally product of spherical hexagonal pixels are mapped to the surface of the icosahedral, and the spherical hexagonal and rectangle are created by unfolding it into a flat rhombus surface. The equal element mapping relationship between them, and the hexagonal pixel data is stored in blocks in the form of rectangular pixels, and is stored in the GeoTIFF file format.
It realizes efficient storage and sharing of hexagonal pixel data, is compatible with standard GIS/RS software, maintains the advantages of hexagonal data, can quickly obtain spherical hexagonal pixel images of different resolutions, and improves data processing and analysis. efficiency.
Smart Images

Figure CN118897900A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of geographic information technology, and in particular relates to a compatible hexagonal pixel remote sensing image storage method. Background Art
[0002] With the rapid development of remote sensing earth observation technology, the growth of data volume and the diversification of data types have brought challenges to their application. The integrated analysis of a large amount of multi-source, multi-scale and multi-dimensional remote sensing data has become a widespread application requirement. Although data cubes and spatiotemporal databases (GeoMesah or MoblityDB), such as the Open Data Cube (ODC) of CEOS (Committee on Earth Observation Satellites) and the Australian Geoscience Data Cube (AGDC), support the storage and analysis of time and space dimensions, the current remote sensing data covers a wide area or the world. The traditional two-dimensional (2D) plane grid system based on local area projection is difficult to support global scale analysis, and the three-dimensional (3D) latitude and longitude grid will also have area degradation at the poles. After exploration, the Australian Science and Technology Agency found that the Discrete Global Grid System (Discrete Global Grid System) can be used as the underlying framework for managing geospatial data and a potential data model to support infrastructure computing. It uses a specific partitioning method to discretize the earth's surface into isomorphic forms, forming a seamless and non-overlapping multi-resolution discrete hierarchical structure, which helps to establish a new data association model with spatial location as the primary key, and is expected to become a new generation of digital earth reference model that supports the fusion processing of earth space data. Compared with traditional spatial data organization and application models, the global discrete grid system is more suitable for solving large-scale problems, and its structure supports efficient processing of multi-resolution data, which is expected to provide solutions for new digital earth data processing and analysis.
[0003] Common DGGS subdivision shapes include triangles, quadrilaterals, and hexagons. Compared with triangles and quadrilaterals, hexagons have more potential in geospatial data processing because of their good geometric properties such as consistent connectivity, isotropy of local neighborhoods, higher sampling rate, and higher angular resolution. In addition, hexagonal grids with different subdivision apertures (3, 4, 7) can be flexibly combined, so that when sampling geospatial data to the grid, the grid unit closest to its resolution can be obtained, which not only improves the sampling accuracy of the data but also reduces the amount of data.
[0004] In order to give full play to the advantages of hexagonal DGGS in organizing remote sensing image data, all the information contained in the image must be accurately converted to hexagonal grid representation, which is a prerequisite for subsequent grid data organization, unit aggregation and spatial relationship query. Although OGC (Open Geospatial Consortium) does not restrict the method of data conversion to DGGS, the adopted quantitative strategy plays an important role in precision control. Ma et al. designed the accuracy evaluation index of remote sensing image interpolation to hexagon, and proved that the bilinear interpolation algorithm retains more image information and structural features (compared with the nearest neighbor interpolation method and bicubic interpolation method), which is closest to the original image. However, the bilinear interpolation algorithm will cause null values in cells close to the boundary of the study area due to the lack of effective interpolation, which will affect the integrity of global data on the spherical grid.
[0005] Some scholars have conducted relevant research on hexagonal DGGS. For example, Mahdavi-Amiri used a three-hole hexagonal grid system and introduced technologies such as wavelet transform to manage multi-scale geospatial data, achieving efficient network transmission of grid data. Although the three-hole hexagonal grid system can achieve a smooth transition between levels, the directions of adjacent level units rotate, and the encoding index is more complicated. Seven-hole subdivision can provide a simpler hierarchical relationship under continuous resolution, but its adjacent level units will also rotate, and the aperture is too large to be suitable for organizing multi-scale remote sensing data. The direction of the four-hole hexagonal grid unit remains unchanged, which is conducive to the retrieval and aggregation of units between levels. This scheme is easy to combine with quadtree coding, binary operations and pyramid structures, which is conducive to a wider range of expansion operations on DGGS. In terms of acquisition and organization of earth observation data, the European Space Agency (ESA) collects soil moisture and ocean salinity information on a four-hole hexagonal grid, and provides data to users in the form of dozens of square pixels simulating an approximate hexagon. Although this has high visualization efficiency, it not only increases the amount of data, but also causes serious unit distortion as the latitude increases.
[0006] Some scholars organize hexagonal data by using GeoJSON, XML or text (ASCII) formats, which must explicitly store the topological structure of the unit so that the value associated with a single hexagon increases from 1 to 13 (increasing the coordinates of each vertex). This scheme is particularly convenient for time series streaming, but it does not seem to be suitable for hexagonal grids with specific geographic locations and ranges. At present, there is no open file format management standard for hexagonal pixel data, and because hexagons cannot directly follow the arrangement of square pixels, existing organization schemes cannot be used directly. In view of this, some researchers or teams use private file formats to record data, but their non-standard formats cannot be parsed by commercial software, making it difficult to distribute and share data sets.
[0007] The global discrete grid system evenly covers the earth's surface with units of equal area. It generally constructs a regular hexagonal grid on the surface of a polyhedron, and then uses the inverse polyhedron projection to map it to the sphere, ensuring that the unit area is equal at the cost of unit shape deformation. However, remote sensing images can only be stored as a planar matrix structure, and it is not easy to directly establish a geometric mapping relationship between spherical hexagons and regular rectangles on a sphere, resulting in the inability to achieve effective and reliable storage of remote sensing image data.
[0008] Combined with the above analysis, there are still several problems in the above research: there is a lack of high-precision hexagonal pixel remote sensing image modeling methods, a lack of file formats that are easy to store and share hexagonal pixel data, the application of hexagonal pixel remote sensing data cannot be well promoted, and it is not easy to directly establish the geometric mapping relationship between spherical hexagons and regular rectangles on a sphere. Summary of the invention
[0009] The purpose of the present invention is to provide a compatible hexagonal pixel remote sensing image storage method to solve the problem that hexagonal pixel data is difficult to store and share.
[0010] In order to solve the above technical problems, the present invention provides a compatible hexagonal pixel remote sensing image storage method, comprising the following steps:
[0011] 1) Using the icosahedral hexagonal global discrete grid as the carrier framework of remote sensing image data, a mathematical model of hexagonal pixels of remote sensing images is established;
[0012] 2) Mapping the spherical hexagonal pixels to the surface of an icosahedron composed of triangular faces with equal areas; unfolding the icosahedron into a plane, splicing two triangular faces in the icosahedron into a rhombus face in the unfolded plane, and establishing an equal-area mapping relationship between the spherical hexagons and rectangles on the surface of the icosahedron based on the established mathematical model;
[0013] 3) Divide the earth's surface into ten regions using rhombuses as units, and combine the equal-area mapping relationship obtained in step 2) to store the spherical hexagonal remote sensing image data in blocks in the form of rectangular pixels in a file, where each rectangle represents a hexagon, and store corresponding projection parameters to obtain spherical hexagonal pixel images of different resolutions based on the rectangular pixel units in the file.
[0014] The beneficial effect of the above technical solution is as follows: in order to utilize the existing standard file format storage standard to be compatible with commonly used GIS / RS software while maintaining the advantages of hexagons, the present invention finds that spherical hexagonal pixel data is not easy to be directly stored in a standard file, that is, it is not easy to directly establish the relationship between hexagons and rectangles on the sphere. The present invention first maps the spherical hexagonal equal area to the surface of an icosahedron composed of triangular faces, and then unfolds it to a plane that is easy to operate and store, and establishes an equal area correspondence between regular hexagons and regular rectangles located on the surface of the polyhedron. Based on this, the earth's surface is divided into units of rhombuses, and the hexagons are stored in the file in the form of rectangular pixels, each rectangle represents a hexagon, and the transformation projection parameters for reconstructing the hexagon are allocated. Combined with the unique correspondence between rectangles and hexagons, the hexagons are equivalently processed when operating rectangles, which facilitates the storage of data in a standard format file, and spherical hexagonal pixel images of different resolutions can be quickly obtained from the rectangular structure images in the file.
[0015] Furthermore, the process of establishing a mathematical model of hexagonal pixels of remote sensing images in step 1) includes: determining a corresponding grid level according to the spatial resolution of the remote sensing image data, spatially aligning the remote sensing image data with the global discrete grid, and after alignment, gridding modeling of hexagonal pixels of the remote sensing map image according to the determined grid level to achieve the establishment of a mathematical model of hexagonal pixels of the remote sensing image.
[0016] Furthermore, the grid level is determined according to the following formula:
[0017] Min|S h -S q |AND(S h <S q )
[0018] In the formula, S h Indicates the hexagonal pixel area when the grid level is n, S q Represents the area of a quadrilateral pixel.
[0019] The beneficial effect of the above technical solution is that the grid level determined according to the above formula can be as close to the original image as possible and minimize information loss.
[0020] Furthermore, in the process of modeling the hexagonal pixels of the remote sensing map image, the overlapping area between the hexagonal and quadrilateral pixels is calculated on the spherical surface, and the calculated overlapping area is used as a weight to determine the hexagonal pixel value. The hexagonal pixel value A is:
[0021]
[0022] Where A is the hexagonal pixel value, q iis the ratio of the area of the i-th intersection unit to the area of the quadrilateral, Value i is the pixel value of the ith quadrilateral, and M is the number of intersections between hexagonal and quadrilateral pixels.
[0023] The beneficial effects of the above technical solution are: the attribute value is calculated based on the area overlap percentage between quadrilateral and hexagonal pixels, the calculation method is more reasonable, and the problem of inaccurate hexagonal pixel values is solved. Moreover, in spherical calculation, the deformation error of hexagonal pixel units caused by projection is avoided, the unit shape is correctly retained, and the modeling accuracy is higher.
[0024] Furthermore, the process of block storage in step 2) includes:
[0025] The geographic coordinates (lat i ,lon i ) is mapped to the icosahedron surface through icosahedral equal-area projection, and the projection coordinates (x i ,y i ), and calculate its diamond Cartesian coordinates (X i , Y i );
[0026] Determine the integer coded coordinates (i, j) of the smallest circumscribed quadrilateral ABCD of the study area; the integer coded coordinates are represented as oij, defined on a diamond face composed of two triangular faces, and the unit is H[d i , (i, j) n ] logo, d i is the diamond face index, (i, j) is the grid center code, n is the grid level, i, j∈Z + ; If the image spans multiple faces, the integer coded coordinates of each point on each rhombus face are calculated separately;
[0027] Divide the spatial range according to the rhombus surface, traverse the circumscribed quadrilateral ABCD of each rhombus surface according to the integer code (i, j), and determine the rhombus Cartesian coordinates of the minimum circumscribed rectangle abcd of the study area; if the image spans multiple surfaces, calculate the rhombus Cartesian coordinates of the corresponding rhombus surfaces respectively;
[0028] According to the basis vectors b1 and b2 in the orthogonal coordinate system, the resolution, image width imageWidth and height imageHeight in the IJ direction of the image are determined, and the row and column coordinates (I, J) of the grid center in the image coordinate system are calculated, and the hexagonal pixel values are stored in the corresponding rectangular pixel units.
[0029] Furthermore, the file is a GeoTIFF file.
[0030] The beneficial effects of the above technical solution are: GeoTIFF format files are standard, open format files, and encapsulating hexagonal pixel image data in the GeoTIFF file format is convenient for users to load and process using existing GIS software, and has good scalability.
[0031] Furthermore, in step 3), the process of obtaining spherical hexagonal pixel images of different resolutions according to the rectangular pixel units in the file includes: reading the integer coordinate indexes of the even rows and odd columns of the image to obtain the attribute values of the rectangular pixel image; obtaining the projection coordinates of the center of the hexagonal grid unit according to the projection parameters encapsulated in the GeoTIFF file; calculating the projection coordinates of the six grid points according to the direction vectors in the six directions of the hexagon, and mapping them to the spherical surface using inverse projection; and iterating the above steps in a loop until the required resolution data is obtained.
[0032] Furthermore, the formula for calculating the row and column coordinates of the grid center in the image coordinate system is:
[0033]
[0034] Where (I, J) represents the row and column coordinates, (Transform[0], Transform[3]) is the coordinate of the upper left corner of the image in the OXY coordinate system; (Transform[1], Transform[5]) respectively represent the resolution of the image in the IJ direction in the OXY coordinate system; (Transform[2], Transform[4]) is the rotation angle of the image; Indicates rounding down; (X, Y) represents the diamond Cartesian coordinates of the cell center. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] FIG. 1( a ) is a diagram showing the intersection relationship between hexagonal and quadrilateral pixels in a low-latitude region according to the present invention;
[0036] FIG1( b ) is a diagram showing the intersection relationship between hexagonal and quadrilateral pixels in a high latitude region according to the present invention;
[0037] Figure 2(a) is a schematic diagram of the intersection of a spherical hexagon and a quadrilateral;
[0038] FIG2( b ) is a schematic diagram of a spherical hexagon contained by a quadrilateral;
[0039] FIG. 3( a ), FIG. 3( b ) and FIG. 3( c ) are three equal-area correspondence diagrams of hexagons and quadrilaterals established by the present invention;
[0040] FIG. 4( a ) and FIG. 4( b ) are equal-area correspondences between rectangular and hexagonal pixels in two cases established by the present invention;
[0041] Figure 5 It is the present invention that maps the equal area of the icosahedral plane hexagonal grid to a rectangular schematic diagram;
[0042] Figure 6 These are the three coordinate systems established by the present invention to describe the process of packing hexagonal pixels into GeoTIFF;
[0043] FIG. 7( a ) is a schematic diagram of the affiliation relationship between parent and child units when constructing multi-resolution data in the present invention;
[0044] FIG7( b ) is a schematic diagram showing the result of designing a downsampling scheme of the present invention when constructing multi-resolution data;
[0045] FIG8( a ) is a diagram showing the result of hexagonal pixel modeling of a Sentinel-2 multi-band image using the method of the present invention;
[0046] Figure 8(b) is the result of hexagonal pixel modeling of Sentinel-2 multi-band image using bilinear interpolation method;
[0047] Figure 8(c) is the result of hexagonal pixel modeling of Sentinel-2 multi-band image using the nearest neighbor pixel method;
[0048] FIG9( a) is a diagram showing the result of hexagonal pixel modeling of a LandSat single-band image using the present invention;
[0049] Figure 9(b) is a diagram showing the result of hexagonal pixel modeling of a LandSat single-band image using bilinear interpolation;
[0050] Figure 9(c) is the result of hexagonal pixel modeling of LandSat single-band image using the nearest neighbor pixel method.
[0051] Fig.10 is the GeoTIFF image corresponding to the local hexagonal pixel remote sensing data of the present invention;
[0052] Figure 11(a) is a detailed diagram of the SMOS data organization scheme;
[0053] FIG11( b ) is a detailed diagram of the organizational scheme of the present invention;
[0054] FIG. 12( a ), FIG. 12( b ) and FIG. 12( c ) are rectangular pixel sea salt data of different resolutions (8, 9, and 10 levels, respectively) stored in GeoTIFF according to the present invention;
[0055] Figures 13(a), 13(b) and 13(c) are hexagonal pixel sea salt data of different resolutions (8, 9 and 10 levels respectively). DETAILED DESCRIPTION
[0056] The present invention mainly discusses the mathematical model for expressing hexagonal pixel information of remote sensing images using the global discrete grid system, and the key process of organizing hexagonal pixel image data using the existing open and standard file format. The overall idea of the present invention is as follows:
[0057] Various existing remote sensing image data sets are mainly expressed in longitude and latitude or Cartesian coordinates. However, the longitude and latitude grids are severely deformed from the equator to the poles, and the resolutions of data sets in different surface areas are often inconsistent, resulting in redundancy in the data expression of the longitude and latitude grids. Therefore, the present invention uses the global discrete grid system as the carrier framework of remote sensing data, and adopts a spherical sampling method that takes into account pixel weights to establish a hexagonal pixel mathematical model for remote sensing images. The intersection area between the hexagon and the rectangular pixels involved is used as the weight, and the attribute value of the hexagonal pixel is equal to the weighted sum of the attribute values of multiple rectangular pixel units. The spherical hexagonal pixel remote sensing data can be obtained using the scheme.
[0058] Although research has shown that hexagons are the best units for data sampling, most current graphics hardware does not support the direct transmission and rendering of hexagonal pixel images. In contrast, rectangular pixel images are easy to transmit and share across networks. Remote sensing images organized in rectangles record geographic phenomena through directly located location information and attribute values that describe properties, and are organized in a simple, regular arrangement. This scheme can avoid explicit storage of topological connections between units or vertices. The value of a certain position can be obtained through a simple and fast calculation method. Combined with data compression technology, it will save a lot of storage space. In addition, rectangular rasters can be stored with the help of GeoTIFF files, which are the most widely used file format in the field of remote sensing and geographic information. It has strong scalability, interoperability, and user-friendliness.
[0059] However, it is not easy to directly establish the relationship between hexagons and rectangles on the sphere, so the present invention first uses the icosahedron Snyder equal-area polyhedron projection to map the equal-area of spherical hexagons to the surface of the polyhedron, then unfolds it to a plane that is easy to operate and store, and establishes the equal-area correspondence between the regular hexagons and regular rectangles on the surface of the polyhedron. With the scheme, the earth's surface is divided into 10 regions in units of diamond blocks, and the hexagons are stored in the GeoTIFF file in the form of rectangular pixels, each rectangle represents a hexagon, and the transformation parameters for reconstructing the hexagons are allocated.
[0060] In order to more efficiently organize and manage the global multi-scale hexagonal pixel remote sensing image data set, the present invention determines the parent-child attribution principle of units between levels based on the hexagonal four-hole subdivision structure, and designs a "hexagonal multi-scale pyramid" scheme. Combined with the unique correspondence between rectangles and hexagons, hexagons are equivalently processed when operating rectangles, and spherical hexagonal pixel images of different resolutions can be quickly obtained from GeoTIFF rectangular structure images. It should be noted that when introducing the entire process below, the selected file is a GeoTIFF file, but in the specific implementation process, the file can be selected according to the actual needs of the user, as long as it is a standard file that can be recognized by commonly used GIS or remote sensing image processing software, such as HDF5 and other files with specific file storage standards.
[0061] The following is a detailed introduction to the implementation process of the entire remote sensing image data storage method.
[0062] Step 1: Using the hexagonal global discrete grid as the carrier framework for remote sensing data, a mathematical model of hexagonal pixels in remote sensing images was established. The idea is: determine the appropriate grid level based on the spatial resolution of the remote sensing map image; align the remote sensing map image data with the global discrete grid spatial benchmark, and after alignment, perform grid modeling of hexagonal pixels in the remote sensing map image according to the determined grid level to achieve the establishment of a mathematical model of hexagonal pixels in remote sensing images. Specifically:
[0063] The sizes of remote sensing image pixel units and four-hole hexagonal grid units at different levels are fixed, and it is difficult to ensure that the two are completely matched. In the modeling process, in order to be as close to the original image as possible and minimize information loss, it can be determined according to formula (1), that is, to select a level with a slightly higher resolution than the original data.
[0064] Min|S h -S q |AND(S h <S q ) (1)
[0065] In the formula, S h Indicates the hexagonal pixel area when the grid level is n, S a Represents the area of a quadrilateral pixel.
[0066] At present, there are related studies on the gridding of remote sensing images using the global discrete grid system. The main idea is to first project the spherical unit to the projection plane where the image data is located, and then use the linear interpolation method to assign the attribute value to the pixel unit. However, since multi-source images have different projection coordinate systems, mapping the unit to different projection planes will not only bring tedious calculations, but any projection will cause it to deform to varying degrees, which can easily lead to inaccurate final sampling results.
[0067] Considering that the essence of the global discrete grid system is the subdivision of the earth's surface rather than a plane, the hexagonal pixel modeling of the remote sensing image should be completed on the spherical surface as much as possible. In addition, in the modeling process, since rectangular pixels and hexagonal pixels are spatial relationships between different types of pixels, a single hexagon will correspond to multiple different quadrilateral pixels, as shown in Figures 1(a) and 1(b). Each quadrilateral will have different degrees of influence on the hexagonal pixel value, so it is not sufficient to describe it only by the relationship between the grid points, and the spatial relationship between the two pixels needs to be carefully considered. Therefore, the present invention chooses to calculate the overlapping area between hexagonal and quadrilateral pixels on the spherical surface and use this as a weight to determine the hexagonal pixel value.
[0068] Due to the constraints of working habits, product compatibility and other factors, most remote sensing images currently still use the traditional mode based on map projection. The spatial reference of these data is different from the global discrete grid system, so they must be "aligned" according to the spatial position in a unified coordinate system before modeling. The positional relationship between the two can be obtained by restoring the plane remote sensing image to the unified spatial reference of the grid, as shown in Figure 2(a) and Figure 2(b). Assume that the vertex sequences of the hexagon and quadrilateral are h1h2h3h respectively. a h5h6,q1q2q a q4.
[0069] When a hexagon is contained by a quadrilateral, the two cover the same area on the sphere, and the hexagonal pixel value is equal to the rectangular pixel value. When a hexagon intersects with a quadrilateral, the attribute value of the hexagonal pixel is equal to the weighted sum of the values of multiple rectangular pixels. If a hexagon intersects with M quadrilateral pixels, the hexagonal pixel value A is calculated as:
[0070]
[0071] In the formula, q i is the ratio of the area of the i-th intersection unit to the area of the quadrilateral, Value i is the pixel value of the i-th quadrilateral.
[0072] Step 2: Store the hexagonal pixel image in a GeoTIFF file. The global discrete grid system uniformly covers the earth's surface with units of equal area. It generally constructs a regular hexagonal grid on the surface of a polyhedron, and then uses an inverse polyhedron projection to map it to a sphere, ensuring that the unit area is equal at the cost of unit shape deformation. However, raster data in GIS is mainly organized in a planar regular rectangular grid, and the distance between adjacent rows and columns is constant. Especially in the field of remote sensing, the grid composed of sampling points in geographic space is usually mapped to a two-dimensional plane for processing, and the remote sensing image is stored as a planar raster matrix structure. If the hexagonal pixel remote sensing image data is stored using the existing GeoTIFF file format, it is urgent to establish a mapping relationship between spherical hexagons and planar rectangular units. Therefore, the main steps to store hexagonal pixel images in GeoTIFF are as follows:
[0073] 1. The spherical hexagonal pixel remote sensing data is mapped to the surface of the polyhedron using equal-area polyhedron projection, the polyhedron surface data is unfolded to a plane that is easy to store and operate, and two triangular faces are spliced into a diamond face; then an equal-area mapping relationship between hexagons and rectangles is established on the surface of the polyhedron.
[0074] Traditional remote sensing image data is mostly expressed in rectangular pixels arranged in longitude and latitude directions, with the horizontal and vertical pixel center lines parallel to the coordinate axes. Hexagons can be used to describe the position of any point in space by establishing two different inclined coordinate systems and an orthogonal coordinate system, such as Figure 3(a) to Figure 3(c) The angles between the coordinate axes of the two inclined coordinate systems are and However, the essence of describing any point in two-dimensional space by the base coordinate system established by both is the same. Suppose two linearly independent unit vectors are υ1 and υ2, and the space grid center point set is L, then:
[0075] L = {mv1 + nv2; m, n∈Z} (3)
[0076] Unlike the inclined coordinate system, when indexing through the orthogonal coordinate system, it is necessary to consider the staggered arrangement of odd and even rows, and because the ratio of the vertical and horizontal spacing between adjacent cells is In order to use integers to quickly index any point like the oblique coordinate system mentioned above, the size of the basis vector needs to be specially set. Figure 3(a) to Figure 3(c) The horizontal spacing between the grid centers is l, and the basis vectors in the horizontal and vertical directions are The spatial grid center point set is L, then:
[0077]
[0078] In the formula, k and h both represent arbitrary integers.
[0079] Each heart can be Figure 3(a) to Figure 3(c) The three methods shown implement integer indexing according to a two-dimensional regular matrix, which can associate the hexagon with the resulting quadrilateral, where the sides of the quadrilateral are parallel to the direction of the basis vectors. With this mapping scheme, a unique and equal-area correspondence can be established between the two, S h =S q , S h and S q are the areas of a plane hexagon and a plane quadrilateral respectively.
[0080] Research Figure 3(a) to Figure 3(c) Among the three corresponding relationships, I and II produce a rhombus, which is convenient for binding the hexagon to the rhombus grid. III produces a rectangle. The present invention focuses on the regular organization of hexagonal pixel remote sensing images, so the focus is on relationship III. It is not difficult to find from Figure 4(a) that the corner point of the rectangular pixel is exactly the center of the hexagonal grid. This point can be selected to represent the rectangular pixel, which corresponds to the "Pixel in corner" in the literature (Baumann, P. (2021). A general conceptual framework for multi-dimensional spatio-temporal data sets. Environmental Modelling & Software, 143, 105096. https: / / doi.org / 10.1016 / j.envsoft.2021.105096). The rectangle can also be offset, as shown in Figure 4(b), so that the grid center is exactly the center of the rectangular pixel, and the rectangular pixel can be represented by the grid center, which corresponds to "Pixel in corner". This information is usually described by metadata, such as a tag in the TIFF format indicating a center pixel or a corner pixel.
[0081] The present invention intends to organize the spatial projection, coordinate transformation, position reference and other information of existing hexagonal pixel data with the help of open file format standards, aiming to realize data transmission sharing and efficient organization. At present, all open file format standards store pixel attribute information based on square pixel row and column numbers, so this mapping scheme provides a mathematical basis for storing hexagons into existing file formats.
[0082] 2. Encapsulate the hexagonal pixel values and related projection parameters into GeoTIFF (i.e., remote sensing image hexagonal pixel data organization).
[0083] At present, in the field of remote sensing, the grid composed of sampling points in geographic space is usually mapped to a two-dimensional plane for processing, and the remote sensing image is stored as a plane grid matrix structure. If the hexagonal pixel remote sensing image data is stored using the existing GeoTIFF file format, it is necessary to establish a mapping relationship between the spherical hexagon and the plane rectangular unit. The conversion from hexagon to rectangle can be achieved with the help of the mathematical model of equal-area mapping between hexagon and rectangular pixels in step 1 included in step 2, such as Figure 5 shown.
[0084] In order to describe the algorithm flow of organizing hexagonal pixel remote sensing images into GeoTIFF blocks, the present invention establishes a Figure 6 The three coordinate systems shown in the figure have the following specific functions: ① Projected Cartesian coordinate system oxy, whose coordinate axes are in the same direction as the triangular faces of the icosahedron, is used to establish the mapping between plane points and spherical surfaces, facilitating the mutual conversion between spherical coordinates and positions on the surface of the polyhedron. ② Integer coded coordinate system oij, defined on a rhombus surface composed of two triangular faces, with units in H[d i , (i, j) n ] logo, where d i is the diamond face index, (i, j) is the grid center code, n is the grid level, i, j∈Z + . This encoding method is consistent with the idea of using integer IJ indexes in regular rectangular grids. Its advantages are that the row and column indexes are regular and neat, which is efficient for neighborhood processing and easy to store and access. ③ The diamond Cartesian coordinate system OXY has two functions: first, it establishes the transition conversion between projection coordinates and integer coded coordinates; second, since the diamond is composed of two triangular faces, when the study area is located on the upper and lower triangular faces, the coordinate system will be inconsistent. This coordinate system can unify the coordinates of the two triangular faces, which is convenient for processing with the diamond face as the basic unit.
[0085] The process of storing spherical hexagonal pixel remote sensing images in GeoTIFF blocks is as follows.
[0086] 1) The geographic coordinates (lat i ,lon i ) is mapped to the surface of the polyhedron through the twenty-sided Snyder equal-area projection to obtain the projection coordinates (x i ,y i ), and calculate its diamond Cartesian coordinates (X i , Y i ).
[0087] 2) Determine the integer coded coordinates (i, j) of the smallest circumscribed quadrilateral ABCD of the study area. Let the integer codes of A and C be A(i A , j A )、C(i C , jC ), the index range of the circumscribed quadrilateral matrix is A(0,0), B(0,j), C(i,j), D(i,0); where i=i C -i A , j = j C -j A If the image spans multiple faces, calculate A in each rhombus face separately. k (0,0)B k (i,0)C k (i,j)D k (0, j) k∈[1, 10], the actual ij encoding needs to increase the offset A(i A , j A ).
[0088] 3) Divide the spatial range according to the rhombus surface, traverse the circumscribed quadrilateral ABCD of each rhombus surface according to the integer code (i, j), and determine the rhombus Cartesian coordinates of the minimum circumscribed rectangle abcd of the study area: a(X1, Y1), b(X2, Y1), c(X2, Y2), d(X1, Y2); Similarly, if the image spans multiple surfaces, calculate the corresponding rhombus surfaces separately.
[0089] 4) According to the basis vectors b1 and b2 in the orthogonal coordinate system, determine the resolution, image width imageWidth, and image height imageHeight in the IJ direction of the image, calculate the row and column coordinates (I, J) of the grid center in the image coordinate system according to formula (5), store the hexagonal pixel values in the corresponding rectangular pixel units, and encapsulate metadata information such as projection.
[0090]
[0091] Where (Transform[0], Transform[3]) are the coordinates of the upper left corner of the image in the OXY coordinate system; (Transform[1], Transform[5]) represent the resolution of the image in the IJ direction in the OXY coordinate system; (Transform[2], Transform[4]) are the rotation angles of the image, which is 0 if the north of the image is facing upwards; Indicates rounding down.
[0092] Step 3: Generate multi-resolution spherical hexagonal grid data.
[0093] According to the above algorithm, the connectivity between hexagonal pixel remote sensing image units can be organized by regularly arranged rectangles. GeoTIFF can be used as a container for storing spherical hexagonal pixel data and a standard format for transmission, sharing, and visualization, but at this time its internal minimum basic unit is still a single-resolution rectangular structure. In practical applications, when using hexagonal DGGS, users may still need to implement multi-resolution analysis or take advantage of the application advantages of hexagons. Visualizing or processing all data at the original resolution on a global scale will result in extremely slow calculations and consume a lot of computing resources. Multi-resolution data can be used to offset the adverse effects of the increase in data volume on transmission and visualization time. One way to deal with it is to parse the GeoTIFF rectangular image structure and convert it into a corresponding level of fine spherical hexagons, and then obtain a low-resolution grid from a high-resolution spherical grid by unit aggregation.
[0094] However, although the above method can realize the conversion of quadrilateral pixel data to hexagonal data, a large number of floating-point operations will be caused in the above calculation process ①, which consumes a lot of time. The four-hole hexagonal subdivision method adopted in this paper is consistent with the traditional pyramid structure. Therefore, the present invention designs a hexagonal pixel image downsampling scheme based on the "one-to-one" mapping model of hexagonal pixels and rectangular pixels. The basic steps are as follows: 1) Read the integer coordinate index of the even rows and odd columns of the image to obtain the attribute value of the rectangular pixel image; 2) According to the metadata parameters encapsulated in the GeoTIFF file, obtain the Snyder projection coordinates of the center of the hexagonal grid unit; 3) According to the direction vectors in the six directions of the hexagon, calculate the projection coordinates of the six grid points, and map them to the sphere using the inverse Snyder projection; 4) Iterate the above steps in a loop, which is similar to the acquisition method of the traditional quadrilateral pyramid, until the required resolution data is obtained. Specifically:
[0095] The four-hole hexagonal partitioning scheme adopted by the present invention is consistent with the traditional pyramid structure. Therefore, a hexagonal pixel image downsampling scheme is designed based on the "one-to-one" mapping model of hexagonal pixels and rectangular pixels. First, the attribution rules of the parent-child units are determined. The parent-child unit affiliation relationship is shown in Figure 7 (a). Each fine hexagon at level n has exactly one parent unit at level n-1. As can be seen from the above, hexagons can be organized with the help of regular rectangles. Operations such as convolution, downsampling, and filtering can be performed in memory using optimized matrix operations, and the unique mapping relationship established between the two enables equivalent processing of hexagons when operating rectangles. In summary, the designed downsampling scheme is as follows: first filter the image, and then extract the even rows and odd columns of the image, as shown in Figure 7 (b). It is similar to the acquisition process of the quadrilateral image pyramid, and the retained units are all children of the parent unit, which is called the "hexagonal multi-resolution pyramid". With the help of the scheme, spherical hexagonal grid data of different scales are managed with rhombus faces as units.
[0096] The resolution is n = N max GeoTIFF images are calculated to level n = N max The specific algorithm of the spherical grid under -1 is as follows: read the integer coordinate index (I, J) of the even rows and odd columns of the image, obtain the attribute value of the corresponding rectangular pixel, and use formula (6) to calculate the diamond Cartesian coordinates (X, Y) of the cell center.
[0097]
[0098] Convert the cell Cartesian coordinates (X, Y) to the integer code (i, j) of the corresponding cell N max , according to the parent-child relationship in Figure 7(a) and formula (7), determine the parent unit code (i, j) n .
[0099]
[0100] According to the projection coordinates (x, y) of the center of the parent cell n Calculate six grid points v i (i=0,1,2,3...) projection coordinates. If the unit is completely inside the triangle, directly use formula (8) to calculate. If the unit is on the boundary, first assign the unit to the corresponding rhombus and then calculate. v0, v1, v2 are the same as formula (8). Other vertices need to be recalculated using formula (9). v I , v J , v I ′,v J ′ is the step vector in the directions of different rhombus faces i, j.
[0101]
[0102]
[0103] The inverse Snyder projection is used to map the plane grid to the sphere, and the location information and attribute values of each spherical grid cell are recorded. The above process can be repeated to generate spherical hexagonal grids of different resolutions from GeoTIFF. Importantly, the general process of this method can theoretically be extended to any DGGS implementation as long as two basic functions are provided: a method to convert geographic data into grid cells and a method to extract the coordinate information of each vertex.
[0104] At this point, the implementation process of the remote sensing image data storage method of the present invention has been introduced.
[0105] The present invention selects remote sensing images of different resolutions and types as experimental objects to demonstrate the advantages of the modeling method of the present invention. Figure 8(a) to Figure 8(c) , Figure 9(a) to Figure 9(c)The modeling results of remote sensing images of Sentinel and Landsat are shown respectively. Among them, Figure 8(a) and Figure 9(a) are the modeling results of the present invention, Figure 8(b) and Figure 9(b) are the modeling results of the bilinear interpolation method, and Figure 8(c) and Figure 9(c) are the modeling results of the nearest neighbor pixel method. It can be clearly found that compared with the traditional method, the modeling results obtained by the present invention are richer in details in locations such as rivers. The present invention calculates the attribute value based on the area overlap percentage between quadrilateral and hexagonal pixels. The calculation method is more reasonable and can solve the problem of inaccurate hexagonal pixel values. In addition, the deformation error of the hexagonal pixel unit caused by projection is avoided in the spherical calculation, the unit shape is correctly retained, and the modeling accuracy is higher.
[0106] The present invention selects a global hexagonal pixel remote sensing image as an experimental object, and stores the hexagonal pixel data into the GeoTIFF file format according to the above process. Fig.10 The storage result of the local spherical grid. Rectangular display devices can only display rectangular or square pixel images. Common remote sensing and GIS software do not support the reading of hexagonal pixel images and cannot parse them directly. This section uses the open file format standard to manage hexagonal pixel data, so that it can be seamlessly integrated with other map software and has good scalability. Before obtaining data for a certain study area, users can use GeoTIFF images to preview, or combine diamond block images with hexagonal quadtrees, and use Web technology to directly transmit and share tile-level data, thereby quickly completing the transmission scheduling of data for a certain area. For example, if you need to obtain hexagonal pixel remote sensing images of the entire territory of China, you only need to parse area 2 to complete it quickly, without loading global grid data. Some areas are divided into multiple partitions, such as Africa, which is stored in areas 1 and 10, which is related to the positioning parameters between the polyhedron and the sphere.
[0107] The present invention realizes the regular organization of hexagonal pixels by establishing a "one-to-one" mapping relationship between hexagons and rectangles. Due to the arrangement of the hexagons themselves, the rectangular pixels inside the image are also presented to the user in the form of an odd-numbered and even-numbered row staggered arrangement, which is different from the arrangement structure of the traditional square pixel image, and is called a "simulated" hexagonal data set. This structure not only correctly retains the logical position relationship between the hexagonal units in the memory, but also can convert the related operations of the hexagon into more conventional rectangular grid processing and analysis. On this basis, it is easy to use standard quadrilateral image processing algorithms, which is convenient for a wider range of operations. By using the open GeoTIFF file format management standard to organize the attribute values, spatial projections, coordinate transformations and other information of the hexagonal pixel image, it can be seamlessly integrated with existing GIS software, libraries and tools, and further data processing and analysis can be performed using these software to improve the application effect of the hexagonal pixel image in practice.
[0108] In order to further highlight the superiority of the data organization scheme of the present invention, the SMOS data released by the European Space Agency was selected as the comparison object. In the Soil Moisture and Ocean Salinity Mission (SMOS) of the Living Planet Programme of the European Space Agency, the data were sampled to a four-hole hexagonal global discrete grid of the Snyder equal-area polyhedron projection of the icosahedron.
[0109] The SNAP software provided by ESA supports quick preview of hexagonal pixel images, which simulates an approximate hexagon through multiple square pixels. Although this makes it convenient to use existing processing algorithms based on rectangular pixels, it greatly increases the amount of data and the expression accuracy is inaccurate. In addition, the number of square pixels that make up a single hexagon is not fixed (about 40 on average), as shown in Figure 11(a). This makes the average area of a single square pixel unit about 5.63km 2 Therefore, in order to make the scheme comparison more reasonable, the present invention resamples the sea salt data to the 12th grid (with an average area of about 3.04km 2 ) and store it in a GeoTIFF image using the organizational strategy of the present invention, as shown in FIG11( b ).
[0110] It can be seen from Figure 11(a) and Figure 11(b) that both of the above two solutions can achieve approximate simulation of hexagons, but when the hexagonal SMOS data is simulated and displayed in the SNAP tool, the shape of the hexagonal pixels is not fixed, as shown in the right of Figure 11(a), making it impossible to accurately determine the position of the hexagon in the rectangular pixel image, which is not conducive to further processing and calculation. In the solution of the present invention, the shape of the hexagonal pixels at any level is fixed, which is convenient for direct processing on the simulated image.
[0111] The present invention directly starts from a GeoTIFF image with a resolution of n, applies a hexagonal downsampling kernel, and extracts the odd rows and even columns of the image to form a pyramid structure with data volume from large to small until the required data resolution is reached, and then it can be quickly restored to the sphere after a small amount of calculation. The present invention has the following advantages: (1) The aperture 4 hexagonal subdivision structure and the "one-to-one" mapping scheme proposed in the present invention can be well combined together, and the hexagonal pixels are processed in the form of an array in the computer, which reduces complex floating-point calculations and improves calculation efficiency. (2) The parent-child addressing relationship constructed in this paper based on the hexagonal DGGS can be quickly calculated through binary operations, reducing the complexity of parent-child queries. (3) The encapsulation of spherical unit recovery parameters reduces the computing resources required for coordinate system conversion during data processing.
[0112] The present invention also uses SMOS data as an experimental comparison object. Currently, SMOS can only meet the data requirements of a single resolution. If you want to obtain multi-resolution hexagonal pixel data, you need to use a resampling algorithm based on the existing solution, but this will destroy the correspondence between hexagons and quadrilaterals, as shown in Figure 11 (a), making it difficult for squares and real hexagonal pixels to be converted into each other. The present invention can first use a hexagonal pyramid structure to iteratively obtain the quadrilateral pixel GeoTIFF images corresponding to different levels of grids, and then use the algorithm for converting GeoTIFF to a spherical grid to convert it into real hexagonal pixel data. Figure 12(a) to Figure 12(c) , Figure 13(a) to Figure 13(c) The SMOS sea salt data of rectangular pixels and hexagonal pixels at three consecutive resolutions of 8, 9, and 10 layers are displayed respectively. It can be found that the hexagonal pixels and rectangular pixels maintain the correct position correspondence in the same area without generating data redundancy.
[0113] The present invention ensures the uniqueness of the position correspondence and the geometrical equi-area when converting the hexagon to a rectangle. The uniqueness enables the equivalent processing of the hexagon when operating the rectangle in the computer memory, avoiding the complex operation of converting the rectangle to the hexagon first, and improving the data processing efficiency in practice. The unit equi-area is conducive to further statistical analysis and processing of SMOS data. In summary, the present solution is more flexible in various aspects and can provide a feasible solution for the unified organization and management of hexagonal SMOS data and the actual processing and application.
[0114] In addition, the present invention has the following advantages: encapsulating the hexagonal pixel image data in the GeoTIFF file format is convenient for users to load and process using existing GIS software. When hexagonal grid data is desired, professional users can write their own code to parse quadrilateral data into hexagonal data, and use existing open source software to load the data onto the spherical surface.
Claims
1. A compatible hexagonal pixel remote sensing image storage method, characterized in that: The steps include: 1) Using the icosahedral hexagonal global discrete grid as the carrier framework of remote sensing image data, a mathematical model of hexagonal pixels of remote sensing images is established; 2) Equal-area mapping of spherical hexagonal pixels to the surface of an icosahedron composed of triangular faces; unfolding the icosahedron into a plane, splicing two triangular faces in the icosahedron into a rhombus face in the unfolded plane, and establishing a unique and equal-area mapping mathematical relationship between hexagons and rectangles on the surface of the icosahedron; 3) Divide the earth's surface into ten regions using rhombuses as units, and combine the equal-area mapping relationship obtained in step 2) to store the spherical hexagonal pixel remote sensing image data in blocks in the form of rectangular pixels in a file, where each rectangle represents a hexagon, and store corresponding parameters including coordinate transformation and projection metadata to restore spherical hexagonal pixel images of different resolutions based on the rectangular pixel units in the file.
2. The method for storing remote sensing images with compatible hexagonal pixels according to claim 1, characterized in that: The process of establishing the mathematical model of the hexagonal pixels of the remote sensing image in step 1) includes: determining the corresponding grid level according to the spatial resolution of the remote sensing image data, aligning the remote sensing image data with the global discrete grid for spatial reference, and after alignment, gridding the hexagonal pixels of the remote sensing image according to the determined grid level to achieve the establishment of the mathematical model of the hexagonal pixels of the remote sensing image.
3. The method for storing remote sensing images with compatible hexagonal pixels according to claim 2, characterized in that: The grid level is determined according to the following formula: Min|S h -S q |AND(S h <S q ) In the formula, S h Indicates the hexagonal pixel area when the grid level is n, S q Represents the area of a quadrilateral pixel.
4. The method for storing remote sensing images with compatible hexagonal pixels according to claim 2, characterized in that: In the process of modeling hexagonal pixels of remote sensing images, the overlapping area between hexagonal and quadrilateral pixels is calculated on the spherical surface, and the calculated overlapping area is used as a weight to determine the hexagonal pixel value. The hexagonal pixel value A is: Where A is the hexagonal pixel value of the remote sensing image, q i is the ratio of the area of the i-th intersection unit to the area of the quadrilateral, Value i is the pixel value of the ith quadrilateral, and M is the number of intersections between hexagonal and quadrilateral pixels.
5. The method for storing remote sensing images with compatible hexagonal pixels according to claim 1, characterized in that: The process of block storage in step 2) includes: The geographic coordinates (lat i ,lon i ) is mapped to the icosahedron surface through the icosahedral Snder equal-area projection, and the projection coordinates (x i ,y i ), and calculate its diamond Cartesian coordinates (X i , Y i ); Determine the integer coded coordinates (i, j) of the smallest circumscribed quadrilateral ABCD of the study area; the integer coded coordinates are represented as oij, defined on a diamond face composed of two triangular faces, and the unit is H[d i , (i, j) n ] logo, d i is the diamond face index, (i, j) is the grid center code, n is the grid level, i, j∈Z + ; If the image spans multiple faces, the integer coded coordinates of each point on each rhombus face are calculated separately; Divide the spatial range according to the rhombus surface, traverse the circumscribed quadrilateral ABCD of each rhombus surface according to the integer code (i, j), and determine the rhombus Cartesian coordinates of the minimum circumscribed rectangle abcd of the study area; if the image spans multiple surfaces, calculate the rhombus Cartesian coordinates of the corresponding rhombus surfaces respectively; According to the basis vectors b1 and b2 in the orthogonal coordinate system, the resolution, image width imageWidth and height imageHeight in the IJ direction of the image are determined, and the row and column coordinates (I, J) of the grid center in the image coordinate system are calculated, and the hexagonal pixel values are stored in the corresponding rectangular pixel units.
6. The method for storing remote sensing images with compatible hexagonal pixels according to claim 1, characterized in that: The file used is a GeoTIFF file.
7. The method for storing remote sensing images with compatible hexagonal pixels according to claim 5, characterized in that: The process of obtaining spherical hexagonal pixel images of different resolutions according to the rectangular pixel units in the file in step 3) includes: reading the integer coordinate index of the even rows and odd columns of the image to obtain the attribute value of the rectangular pixel image; obtaining the projection coordinates of the center of the hexagonal grid unit according to the projection parameters encapsulated in the GeoTIFF file; calculating the projection coordinates of the six grid points according to the direction vectors in the six directions of the hexagon, and mapping them to the spherical surface by inverse projection; and iterating the above steps in a loop until the required resolution data is obtained.
8. The method for storing remote sensing images with compatible hexagonal pixels according to claim 5, characterized in that: The formula for calculating the row and column coordinates of the grid center in the image coordinate system is: Where (I, J) represents the row and column coordinates, (Transform[0], Transform[3]) is the coordinate of the upper left corner of the image in the OXY coordinate system; (Transform[1], Transform[5]) represent the resolution of the image in the IJ direction in the OXY coordinate system; (Transform[2], Transform[4]) is the rotation angle of the image; Indicates rounding down; (X, Y) represents the diamond Cartesian coordinates of the cell center.
Citation Information
Patent Citations
Vector data drawing method on full-sphere hexagon discrete grid
CN102096946A
Remote sensing big data slicing method, device and equipment based on hexagon
CN114092656A
Digital earth framework and indexing schema to extend any "surface" discrete global grid system to construct interoperable "volumetric" discrete global grid systems
WO2018136996A1
