Map tile generation method based on FPGA technology

By designing a map tile generation algorithm on FPGA, using parallel computing and reconstructible logical resources, the problems of calculation bottlenecks and storage pressure in traditional methods are solved, and efficient map tile generation is achieved, satisfying the high-resolution marine data. Real-time rendering requirements.

CN118885554BActive Publication Date: 2024-12-03SANYA INST OF OCEANOGRAPHY OCEAN UNIV OF CHINA
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411356785.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-27
Publication Date
2024-12-03
Estimated Expiration
2044-09-27

AI Technical Summary

Technical Problem

Traditional CPU-based map tile generation methods have computing bottlenecks and storage pressures when processing large-scale, high-resolution marine data, and the problems of GPU's memory capacity limitation and waste of computing resources are difficult to meet the requirements of real-time and efficiency.

Method used

Using the map tile generation method based on FPGA technology, the tile is designed by setting up the reference coordinate matrix A, the input coordinate matrix B, the reference value matrix Ck and the output value matrix D. Generate algorithms to realize efficient conversion and calculation of data.

Benefits of technology

It improves the efficiency and performance of map tile generation, optimizes resource utilization, meets the real-time rendering needs of high-resolution marine data, and significantly improves the stability and application effect of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118885554B_ABST
    Figure CN118885554B_ABST
Patent Text Reader

Abstract

A map tile generation method based on FPGA technology includes setting four matrix reference coordinate matrices for tile images A , input coordinate matrix B , benchmark value matrix C k , output value matrix D ,when A , B , C k It is known that using A , B , C k Find D ;by A The center divides the earth's longitude and latitude into nine regions from 0 to 8. B element b (p,q) Depending on the latitude and longitude values, when the coordinates are in zone 0, 1, or 2, A There are four adjacent coordinates in b (p,q) Surrounded by C k Find the four corresponding values ​​and perform two-dimensional linear interpolation; when it is in the other 6 areas, A The boundary is found with b (p,q) Two adjacent coordinates, in C k Find the two corresponding values ​​and perform one-dimensional linear interpolation. The tile generation algorithm of the present invention proposes to achieve data conversion by setting up four matrices, and fully utilizes the parallel computing capability of FPGA to improve the execution efficiency and performance of the algorithm.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a map tile generation method, in particular to a map tile generation method based on FPGA technology, and belongs to the technical field of digital maps. Background Art

[0002] In the information age, a large amount of ocean data can be obtained through various observation methods, including satellite remote sensing, drifting buoys, and cruise observation. In order for people to better understand ocean data and perceive the information contained in the data, it is necessary to visualize the data. Ocean data visualization maps are a common form. The method of processing ocean data into maps is usually to interpolate the data into a map grid, and then render it to show the ocean information characteristics of different areas on the map in different colors.

[0003] With the widespread popularity and diversification of digital map applications, the scale and complexity of map data are increasing. Large-scale map data sets and high-resolution map layers pose many challenges to traditional map tile generation methods. The computational bottlenecks and latency problems of traditional central processing unit (CPU)-based computing methods when processing huge amounts of map data are gradually emerging, and the requirements for real-time and high efficiency are becoming increasingly difficult to meet. The resolution of ocean data is constantly increasing, and traditional direct interpolation methods face tremendous pressure in terms of computational complexity and storage overhead when processing high-resolution ocean data.

[0004] The process of generating map tiles usually involves a large number of computing and processing tasks, such as image processing, data parsing, texture compression, etc. Therefore, it is particularly important to choose a suitable computing platform, including CPU, GPU and FPGA, to generate map tiles faster. The CPU is less efficient for large-scale parallel computing tasks. Therefore, for processing large-scale, high-resolution ocean data, a large number of repetitive simple computing tasks need to be performed, and the serial computing capability of the CPU may show a computing bottleneck. The memory capacity of the GPU is usually much smaller than that of the CPU. When processing large-scale map data, the memory capacity of the GPU may become a limiting factor, resulting in the inability to load and process all the required data at the same time. At the same time, the GPU does not fully match the map tile generation algorithm, resulting in a waste of computing resources.

[0005] The map tile generation service of ocean data is inseparable from the calculation and rendering of massive data. For high-performance rendering of large-scale data, domestic and foreign scholars have conducted extensive research on data organization and spatial indexing technology, LOD technology, and GPU hardware accelerated rendering. The current research status of map tile storage structure is that the map tile storage structure usually adopts a pyramid structure and introduces LOD technology, but there are problems such as long data preprocessing time, data volume redundancy, and large scale jumps between tiles. A lot of research has been done at home and abroad on tile pyramid generation technology and rapid construction. In summary, it mainly focuses on two aspects: one is to improve the original tile pyramid generation method, and the other is to use parallel technology to speed up the construction of tile pyramid.

[0006] FPGA is a field programmable gate array that features highly parallel computing, low power consumption, and reprogrammable features. It can be wired and synthesized an unlimited number of times, and can re-describe the hardware structure at any time according to demand to generate dedicated circuits. FPGA has a high degree of parallel processing capabilities and can efficiently complete large-scale calculations. Therefore, FPGA has made a series of progress in the fields of deep learning, numerical computing, and algorithm acceleration. FPGA has demonstrated powerful computing capabilities in the fields of numerical computing and algorithm acceleration, but in the field of image coding, better tile generation methods still need to be designed to enable FPGA to play a more significant role in the field of map tile generation. Summary of the invention

[0007] The purpose of the present invention is to provide a map tile generation method based on FPGA technology, so as to realize the design of tile generation algorithm in FPGA and apply it to actual website services.

[0008] The map tile generation method based on FPGA technology is characterized by comprising the following steps:

[0009] (1) Select the image encoding scheme for map tiles.

[0010] (2) For tile images, set four matrices A , B , C k , D .in:

[0011] A is the reference coordinate matrix, each element of which a (i,j) It is the coordinates of a reference point, including two values: longitude and latitude, i.e. (longitude, latitude); A The latitude is determined by the longitude and latitude resolution, and the maximum resolution of longitude and latitude is 0.2°. A Longitude range: -180°+longitude resolution / 2,…,180°-longitude resolution / 2;A Latitude range: -90°+latitude resolution / 2,…,90°-latitude resolution / 2; A The maximum dimension is 900×1800, or 901×1801; then the above latitude and longitude area is the reference coordinate matrix A .

[0012] B is the input coordinate matrix, each element of which b (p,q) is the coordinates of an input point, which also includes two values, longitude and latitude, i.e. (longitude, latitude), but this coordinate pair is generally not the same as A One of a (i,j) overlap; about B The dimension of the tile is first introduced: each tile is 256×256 in size, and the input coordinate matrix B The horizontal and vertical dimensions of are integer multiples of a single tile, that is, B The minimum dimension is (2 2 ×256) ×(2 2 ×256), the maximum dimension is (2 10 ×256) ×(2 10 ×256); and B It must be a square array.

[0013] C k is the reference value matrix, which is A The dimensions are the same, and each element c k,(i,j) , is the coordinate of the reference point a (i,j) A corresponding ocean or meteorological physical value; k is a time series, k Take 0, 1, 2, 3, ..., 23, 24, corresponding to 0 o'clock, 1 o'clock, 2 o'clock, 3 o'clock, ..., 23 o'clock, 24 o'clock respectively. C k Elements in c k,(i,j) It's the coordinates a (i,j) The corresponding ocean or meteorological physical value, that is, for a physical value, there will be 25 C k : C 0. C 1....... C 24 Corresponding to its values ​​at different times.

[0014] Dis the output value matrix, which is B The dimensions are the same, and each element d (p,q) is the coordinate of the input point b (p,q) A physical value corresponding to C k The physical values ​​represented have the same meaning.

[0015] (3) Output value matrix D Solution

[0016] when A , B , C k It is known that using A , B , C k Find D ; D The dimensions and B The dimensions are exactly the same. D Elements in d (p,q) The method to find is as follows:

[0017] (3.1) Taking the entire longitude and latitude of the earth as the global domain, the matrix A The area is the center. A The boundary is extended along the longitude and latitude, thereby dividing the longitude and latitude of the earth into nine square grid areas, and the center, left side, right side, bottom, top, bottom left, bottom right, top left, and top right areas are recorded as area 0 to area 8 respectively;

[0018] (3.2) From the input coordinate matrix B Read out the elements one by one b (p,q) The longitude and latitude value ( x , y ), and determine b (p,q) In which area between 0 and 8?

[0019] When in zone 0, 1, or 2, A There are four adjacent coordinates in b (p,q) Surrounded by;

[0020] When located in the other 6 areas, A The boundary is found with b (p,q) Two adjacent coordinates;

[0021] (3.3) When located in region 0, 1 or 2, the four adjacent coordinates are in the reference value matrixC k Find the four corresponding values, and perform two-dimensional linear interpolation between the four values ​​and any one of the four coordinates. The result is d (p,q) ;

[0022] When located in other areas, based on the two adjacent coordinates in the reference value matrix C k Find the two corresponding values, and perform one-dimensional linear interpolation between the two values ​​and any one of the two coordinates. The result is d (p,q) .

[0023] The step (3.2) determines b (p,q) In which area between 0 and 8 is it located according to the reference coordinate matrix A First determine the following 8 parameters:

[0024] A The width is the number of columns M ;

[0025] A The height is the number of rows N ;

[0026] Longitude lower limit Lon min : -180°+longitude resolution / 2;

[0027] Upper longitude limit Lon max : 180°-longitude resolution / 2;

[0028] Lower latitude limit Lat min : -90°+latitude resolution / 2;

[0029] Upper latitude limit Lat max : 90°-latitude resolution / 2;

[0030] Longitude resolution Lon res ;

[0031] Latitude resolution Lat res ;

[0032] Then according to b (p,q) The longitude and latitude value ( x , y ) to determine which region it belongs to.

[0033] In the step (3.2),b (p,q) The regions are divided into 9 categories:

[0034] (0) When Lon min ≤ x < Lon max and Lat min ≤ y < Lat max hour, b (p,q) Falls in area 0;

[0035] Calculate the subscript:

[0036] ,

[0037] ,

[0038] According to the subscript j and i , from the reference coordinate matrix A Take out a (i,j) , a (i,j+1) , a (i+1,j) , a (i+1,j+1) ,

[0039] select A middle a (i,j) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0040] According to the subscript j and i , from the benchmark value matrix C k Take out c k,(i,j) , c k,(i,j+1) , c k,(i+1,j) , c k,(i+1,j+1) , and determine z 1. z 2. z 3. z 4:

[0041] c k,(i,j) = z 1

[0042] c k,(i,j+1) = z 2

[0043] c k,(i+1,j) = z 3

[0044] c k,(i+1,j+1) = z 4

[0045] Then go to step (3.3).

[0046] (1) When x < Lon min and Lat min ≤ y < Lat max hour, b (p,q) Falls in zone 1;

[0047] Calculate the subscript:

[0048] ,

[0049] select A middle a (i,1) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0050] Subscript by row i , from the benchmark value matrix C k Take out c k,(i,j) , c k,(i,j+1) , c k,(i+1,j) , c k,(i+1,j+1) , and determine z 1. z 2. z 3. z 4:

[0051] c k,(i,j) = c k,(i,M) = z 1

[0052] c k,(i,j+1) = c k,(i,1) =z 2

[0053] c k,(i+1,j) = c k,(i+1,M) = z 3

[0054] c k,(i+1,j+1) = c k,(i+1,1) = z 4

[0055] Then go to step (3.3).

[0056] (2) When x ≥ Lon max and Lat min ≤ y < Lat max hour, b (p,q) Falls in zone 2;

[0057] Calculate the subscript:

[0058] ,

[0059] select A middle a (i,M) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0060] Subscript by row i , from the benchmark value matrix C k Take out c k,(i,j) , c k,(i,j+1) , c k,(i+1,j) , c k,(i+1,j+1) , and determine z 1. z 2. z 3. z 4:

[0061] c k,(i,j) = c k,(i,M) = z 1

[0062] c k,(i,j+1) =c k,(i,1) = z 2

[0063] c k,(i+1,j) = c k,(i+1,M) = z 3

[0064] c k,(i+1,j+1) = c k,(i+1,1) = z 4

[0065] Then go to step (3.3).

[0066] (3) When Lon min ≤ x < Lon max and y ≤ Lat min hour, b (p,q) Falls in zone 3;

[0067] Calculate the subscript:

[0068] ,

[0069] select A middle a (1,j) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0070] According to column subscript j , from the benchmark value matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2:

[0071] c k,(i,j) = c k,(1,j) = z 1

[0072] c k,(i,j+1) = c k,(1,j+1) = z 2

[0073] Then go to step (3.3).

[0074] (4) When Lon min ≤ x < Lon max and y ≥ Lat max hour, b (p,q) Falls in zone 4;

[0075] Calculate the subscript:

[0076] ,

[0077] select A middle a (N,j) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0078] According to column subscript j , from the benchmark value matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2:

[0079] c k,(i,j) = c k,(N,j) = z 1

[0080] c k,(i,j+1) = c k,(N,j+1) = z 2

[0081] Then go to step (3.3).

[0082] (5) When x < Lon min and y ≤ Lat min hour, b (p,q) Falls in zone 5;

[0083] select A middle a (1,1) The coordinate value of the point ( x 1,y 1) Used for interpolation;

[0084] From the benchmark matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2:

[0085] c k,(i,j) = c k,(1,M) = z 1

[0086] c k,(i,j+1) = c k,(1,1) = z 2

[0087] Then go to step (3.3).

[0088] (6) When x ≥ Lon max and y ≤ Lat min hour, b (p,q) Falls in zone 6;

[0089] select A middle a (1,M) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0090] From the benchmark matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2:

[0091] c k,(i,j) = c k,(1,M) = z 1

[0092] c k,(i,j+1) = c k,(1,1) = z 2

[0093] Then go to step (3.3).

[0094] (7) When x < Lon max and y ≥ Lat max hour, b (p,q) Falls in zone 7;

[0095] select A middle a (N,1) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0096] From the benchmark matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2:

[0097] c k,(i,j) = c k,(N,M) = z 1

[0098] c k,(i,j+1) = c k,(N,1) = z 2

[0099] Then go to step (3.3).

[0100] (8) When x ≥ Lon max and y ≥ Lat max hour, b (p,q) Falls in zone 8;

[0101] select A middle a (N,M) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0102] From the benchmark matrix C k Take out c k,(i,j) , ck,(i,j+1) , and determine z 1. z 2:

[0103] c k,(i,j) = c k,(N,M) = z 1

[0104] c k,(i,j+1) = c k,(N,1) = z 2

[0105] Then go to step (3.3).

[0106] The two-dimensional linear interpolation described in step (3.3) is as follows:

[0107] ,

[0108] g That's what you want d (p,q) ,

[0109] The one-dimensional linear interpolation is as follows:

[0110] ,

[0111] e That's what you want d (p,q) .

[0112] The implementation of tile generation algorithm in FPGA needs to be redesigned and optimized. This paper studies the tile generation algorithm in FPGA and proposes to set up four matrices: reference coordinate matrix A , input coordinate matrix B , reference value matrix C k , output value matrix D , data conversion is realized, and the parallel computing capability and reconfigurable logic resources of FPGA are fully utilized to improve the execution efficiency and performance of the algorithm.

[0113] The present invention uses FPGA to generate map tiles and has achieved good results in map tile quality evaluation, generation speed and efficiency evaluation, resource utilization evaluation, system stability evaluation and actual application scenario evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0114] Figure 1 is a dimension of (2 2 ×256)×(22 ×256) input coordinate matrix B .

[0115] Figure 2 is a matrix A , B , C k , D Relationship diagram.

[0116] Figure 3 Schematic diagram of dividing the earth's longitude and latitude with the reference coordinate matrix as the center area.

[0117] Figure 4 Will Figure 3 Schematic diagram of the splitting of the area shown.

[0118] Figure 5 Schematic diagram of the corresponding relationship between coordinates and physical values ​​when belonging to area 0.

[0119] Figure 6 Schematic diagram of the corresponding relationship between coordinates and physical values ​​when belonging to area 1.

[0120] Figure 7 Schematic diagram of the corresponding relationship between coordinates and physical values ​​when belonging to area 2.

[0121] Figure 8 Schematic diagram of the corresponding relationship between coordinates and physical values ​​when belonging to area 3.

[0122] Fig. 9 Schematic diagram of the correspondence between coordinates and physical values ​​when belonging to area 4.

[0123] Fig.10 Schematic diagram of two-dimensional linear interpolation.

[0124] Fig.11 Schematic diagram of one- and two-dimensional linear interpolation. DETAILED DESCRIPTION

[0125] The map tile generation method based on FPGA technology includes the following steps:

[0126] 1) Select the image encoding scheme for map tiles, such as JPEG, PNG and other image encodings, to implement subsequent processes including image segmentation, color space conversion, data filtering, DEFLATE compression and other processes.

[0127] 2) For tile images, set four matrices A , B , C k , D ,in:

[0128] A is the reference coordinate matrix, each element of which is a (i,j) It is the coordinates of a reference point, including longitude and latitude.

[0129] B is the input coordinate matrix, each element of which b (p,q) It is the coordinates of an input point, including longitude and latitude values;

[0130] C k is the reference value matrix, which is A The dimensions are the same, and each element c k,(i,j) is the coordinate element of the reference point a (i,j) A corresponding ocean or meteorological physical value; k is a time series;

[0131] D is the output value matrix, which is B The dimensions are the same, and each element d (p,q) is the coordinate of the input point b (p,q) A physical value corresponding to C k The physical values ​​represented have the same meaning.

[0132] if A , B , C k It is known that using A , B , C k Find D .

[0133] 2.1) About the reference coordinate matrix A

[0134] 2.1.1) About A Dimensions:

[0135] Generally, the resolution of longitude and latitude is 1°, so A The dimensions are 180 × 360. For example:

[0136] The longitudes are -179.9, -178.9, ......, -0.9, 0.1, 1.1, 2.1,......, 178.1,179.1, for a total of 360 longitude values, as shown in Table 1.

[0137] Table 1

[0138]

[0139] The latitudes are -89.9, -88.9, ......, -0.9, 0.1, 1.1, 2.1,......, 88.1, 89.1, a total of 180 latitude values, as shown in Table 2.

[0140] Table 2

[0141]

[0142] The maximum resolution of longitude and latitude is 0.2°, so A The dimensions are 900×1800, or 901×1801. For example:

[0143] The longitudes are -179.9, -179.7, ......, -0.1, 0.1, 0.3, 0.5,......, 179.7,179.9, a total of 1800 longitude values;

[0144] The latitudes are -89.9, -89.7, ......, -0.1, 0.1, 0.3, 0.5,......, 89.7, 89.9, for a total of 900 latitude values.

[0145] 2.1.2) About A Elements in a (i,j) :

[0146] A Elements in a (i,j) It is a pair of coordinates, namely (longitude, latitude).

[0147] 2.2) About the input coordinate matrix B

[0148] 2.2.1) About B Dimensions:

[0149] In explaining B Before we talk about the dimensions, we first introduce the concept of "tiles": each tile is 256×256 in size.

[0150] Input coordinate matrix B The horizontal and vertical dimensions of are integer multiples of a single tile, that is,

[0151] B The minimum dimension is (2 2 ×256) ×(2 2×256), the maximum dimension is (2 10 ×256)×(2 10 ×256); and B It must be a square array.

[0152] Figure 1 is a dimension of (2 2 ×256) ×(2 2 ×256) input coordinate matrix B Example.

[0153] 2.2.2) About B Elements in b (p,q) :

[0154] B Elements in b (p,q) is also a pair of coordinates, i.e. (longitude, latitude). However, this pair of coordinates is generally not the same as A One of a (i,j) coincide.

[0155] 2.3) About the benchmark matrix C k

[0156] 2.3.1) About C k Dimensions: C k The dimensions and A The dimensions are exactly the same.

[0157] 2.3.2) About C k Elements in c k,(i,j) :

[0158] C k The corresponding physical values ​​can be various meteorological and hydrological data such as wind speed, temperature, waves, wave height, precipitation, salinity, depth, storm surge, chlorophyll, water level, dissolved oxygen, nitrate, etc. c k,(i,j) The coordinate pair a (i,j) The corresponding wind speed, temperature and other values.

[0159] k Take 0, 1, 2, 3, ..., 23, 24, which correspond to the reference physical values ​​at 0 o'clock, 1 o'clock, 2 o'clock, 3 o'clock, ..., 23 o'clock, and 24 o'clock respectively.

[0160] That is, for the same physical value, there are usually 25C k ( C 0. C 1....... C 24 ) represents the reference value matrix at different times.

[0161] 2.4) About the output value matrix D

[0162] 2.4.1) About D Dimensions: D The dimensions and B The dimensions are exactly the same.

[0163] 2.4.2) About D Elements in d (p,q) : d (p,q) The method to calculate is as follows.

[0164] 3) About the output value matrix D Solution

[0165] Assume that the resolution of the reference coordinates is 1°, that is, the reference coordinate matrix A The dimension is 180×360, so the reference value matrix C k The dimensions are also 180×360.

[0166] b (p,q) As input coordinate matrix B An element of A Four adjacent elements in a (i,j) , a (i,j+1) , a (i+1,j) , a (i+1,j+1) surrounded by.

[0167] According to the benchmark values ​​corresponding to these four elements c k,(i,j) , c k,(i,j+1) , c k,(i+1,j) , c k,(i+1,j+1) , we can find the coordinates b (p,q) Corresponding physical value d (p,q) The relationship between the four is as follows Figure 2 shown.

[0168] The specific solution steps are as follows:

[0169] 3.1) According to the reference coordinate matrix A Determine the following 8 parameters:

[0170] A The width is the number of columns M ; A The height is the number of rows N ;

[0171] Longitude lower limit Lon min : -180°+longitude resolution / 2; longitude upper limit Lon max : 180°-longitude resolution / 2;

[0172] Lower latitude limit Lat min : -90° + latitude resolution / 2; upper latitude limit Lat max : 90°-latitude resolution / 2;

[0173] Longitude resolution Lon res ; Latitude resolution Lat res .

[0174] 3.2) From the input coordinate matrix B Read out the elements one by one b (p,q) ,Sure b (p,q) From the reference coordinate matrix A Which coordinate points are surrounded by .

[0175] like Figure 3 As shown, the vast majority b (p,q) Will fall A The interior of , that is, area 0 in the figure above;

[0176] It may also fall A Areas outside the range, i.e., areas 1 to 8 in the above figure, such as when it is smaller than the longitude numbered 1 in Table 1, smaller than the latitude numbered 1 in Table 2, larger than the longitude numbered 360 in Table 1, and larger than the latitude numbered 180 in Table 2.

[0177] Will Figure 3 The nine-square grid area shown is based on Figure 4 The solid line represents a closed interval, and the dotted line represents an open interval. b (p,q The longitude values ​​included are x , the latitude value isy , the following 9 situations can be obtained.

[0178] (0) When Lon min ≤ x < Lon max and Lat min ≤ y < Lat max hour, b (p,q) falls in region 0; corresponds to Figure 4 Central region;

[0179] Calculate the subscript:

[0180] ,

[0181] ,

[0182] According to the subscript j and i , from the reference coordinate matrix A Take out a (i,j) , a (i,j+1) , a (i+1,j) , a (i+1,j+1) ,

[0183] select A middle a (i,j) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0184] According to the subscript j and i , from the benchmark value matrix C k Take out c k,(i,j) , c k,(i,j+1) , c k,(i+1,j) , c k,(i+1,j+1) , and determine z 1. z 2. z 3. z 4:

[0185] c k,(i,j) = z 1

[0186] c k,(i,j+1) = z 2

[0187] c k,(i+1,j) = z 3

[0188] c k,(i+1,j+1) = z 4

[0189] Then go to step (3.3).

[0190] (1) When x < Lon min and Lat min ≤ y < Lat max hour, b (p,q) Falls in area 1; corresponds to Figure 4 left center area;

[0191] Calculate the subscript:

[0192] ,

[0193] select A middle a (i,1) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0194] Subscript by row i , from the benchmark value matrix C k Take out c k,(i,j) , c k,(i,j+1) , c k,(i+1,j) , c k,(i+1,j+1) , and determine z 1. z 2. z 3. z 4:

[0195] c k,(i,j) = c k,(i,M) = z 1

[0196] ck,(i,j+1) = c k,(i,1) = z 2

[0197] c k,(i+1,j) = c k,(i+1,M) = z 3

[0198] c k,(i+1,j+1) = c k,(i+1,1) = z 4

[0199] For longitude less than the lower limit Lon min In the case of , since longitude has a circular nature, the highest column is mapped to the left of the lowest column, and the subsequent calculation is shown in 3.3).

[0200] (2) When x ≥ Lon max and Lat min ≤ y < Lat max hour, b (p,q) Falls in area 2; corresponds to Figure 4 right middle area;

[0201] Calculate the subscript:

[0202] ,

[0203] select A middle a (i,M) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0204] Subscript by row i , from the benchmark value matrix C k Take out c k,(i,j) , c k,(i,j+1) , c k,(i+1,j) , c k,(i+1,j+1) , and determine z 1. z 2. z 3. z 4:

[0205] c k,(i,j) = c k,(i,M) = z 1

[0206] c k,(i,j+1) = c k,(i,1) = z 2

[0207] c k,(i+1,j) = c k,(i+1,M) = z 3

[0208] c k,(i+1,j+1) = c k,(i+1,1) = z 4

[0209] For longitude greater than the upper limit Lon max In the case of , since longitude has a circular nature, map the lowest column to the right of the highest column, and then go to step (3.3);

[0210] (3) When Lon min ≤ x < Lon max and y ≤ Lat min hour, b (p,q) Falls in area 3; corresponds to Figure 4 Lower middle area;

[0211] Calculate the subscript:

[0212] ,

[0213] select A middle a (1,j) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0214] According to column subscript j , from the benchmark value matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z2:

[0215] c k,(i,j) = c k,(1,j) = z 1

[0216] c k,(i,j+1) = c k,(1,j+1) = z 2

[0217] For latitudes less than the lower latitude limit Lat min In this case, the latitude is set as the lower limit of the latitude, and then go to step (3.3).

[0218] (4) When Lon min ≤ x < Lon max and y ≥ Lat max hour, b (p,q) Falls in zone 4; corresponds to Figure 4 Upper middle area;

[0219] Calculate the subscript:

[0220] ,

[0221] select A middle a (N,j) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0222] According to column subscript j , from the benchmark value matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2:

[0223] c k,(i,j) = c k,(N,j) = z 1

[0224] c k,(i,j+1) = c k,(N,j+1) =z 2

[0225] For latitudes greater than the latitude upper limit Lat max In this case, the latitude is set as the upper limit of the latitude, and then go to step (3.3).

[0226] (5) When x < Lon min and y ≤ Lat min hour, b (p,q) Falls in area 5; corresponds to Figure 4 Lower left area;

[0227] select A middle a (1,1) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0228] From the benchmark matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2:

[0229] c k,(i,j) = c k,(1,M) = z 1

[0230] c k,(i,j+1) = c k,(1,1) = z 2

[0231] Then go to step (3.3).

[0232] (6) When x ≥ Lon max and y ≤ Lat min hour, b (p,q) Falls in area 6; corresponds to Figure 4 Lower right area

[0233] select A middle a (1,M) The coordinate value of the point ( x 1,y 1) Used for interpolation;

[0234] From the benchmark matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2:

[0235] c k,(i,j) = c k,(1,M) = z 1

[0236] c k,(i,j+1) = c k,(1,1) = z 2

[0237] Then go to step (3.3).

[0238] (7) When x < Lon max and y ≥ Lat max hour, b (p,q) Falls in zone 7; corresponds to Figure 4 Upper left area;

[0239] select A middle a (N,1) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0240] From the benchmark matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2:

[0241] c k,(i,j) = c k,(N,M) = z 1

[0242] c k,(i,j+1) = c k,(N,1) = z 2

[0243] Then go to step (3.3).

[0244] (8) When x ≥ Lon max and y ≥ Lat max hour, b (p,q) Falls in zone 8; corresponds to Figure 4 upper right area;

[0245] select A middle a (N,M) The coordinate value of the point ( x 1, y 1) Used for interpolation;

[0246] From the benchmark matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2:

[0247] c k,(i,j) = c k,(N,M) = z 1

[0248] c k,(i,j+1) = c k,(N,1) = z 2

[0249] Then go to step (3.3).

[0250] 3.3) Based on the parameters and data determined in 3.2), perform the following calculations:

[0251] Perform two-dimensional linear interpolation on the results of (0), (1), and (2) in 3.2):

[0252] ,

[0253] g It's what's required d (p,q) ,

[0254] The results of (3) to (8) in 3.2) are simplified to one-dimensional linear interpolation:

[0255] ,

[0256] e It's what's required d (p,q) .

[0257] 4) Effect display and evaluation plan

[0258] By comparing and evaluating the three solutions of generating map tiles using CPU, generating map tiles using CPU+GPU, and generating map tiles using CPU+FPGA, the solutions were evaluated, including map tile quality assessment, generation speed and efficiency assessment, resource utilization assessment, system stability assessment, and actual application scenario assessment. CPU+FPGA is the fastest and most effective.

[0259] 5) Feasibility and prospect analysis

[0260] Currently, GPU is usually used to accelerate the calculation of map tiles. This study uses FPGA to generate map tiles. This study is expected to replace GPU in the field of map tile generation calculation. Using FPGA for image encoding has reference significance for the new image coding technology studied.

[0261] At present, the data transmission between FPGA and PC has been realized, and the image information has been transmitted to PC and saved as an image. The high-speed data transmission solution is feasible; the interpolation operation in FPGA has been realized, and the tile generation algorithm design solution is feasible. At present, the generation of map tiles using CPU and GPU has been completed and actually applied to the website, and the effect display and evaluation solution is feasible.

Claims

1. A map tile generation method based on FPGA technology, characterized by: The following steps are involved: (1) Select the image encoding scheme for map tiles; (2) For tile images, set four matrices A , B , C k , D ,in: A is the reference coordinate matrix, each element of which a (i,j) It is the coordinates of a reference point, including longitude and latitude. A The latitude is determined by the longitude and latitude resolution, and the maximum resolution of longitude and latitude is 0.2°. A Longitude range: -180°+longitude resolution / 2,…,180°-longitude resolution / 2; A Latitude range: -90°+latitude resolution / 2,…,90°-latitude resolution / 2; A The maximum dimension is 900×1800, or 901×1801; then the above latitude and longitude area is the reference coordinate matrix A ; B is the input coordinate matrix, each element of which b (p,q) is the coordinates of an input point, including longitude and latitude, but the coordinate pair will not be A One of a (i,j) overlap; about B The dimension of the tile is first introduced: each tile is 256×256 in size, and the input coordinate matrix B The horizontal and vertical dimensions of are integer multiples of a single tile, that is, B The minimum dimension is (2 2 ×256) ×(2 2 ×256), the maximum dimension is (2 10 ×256) ×(2 10 ×256); and B It must be a square array; C k is the reference value matrix, which is A The dimensions are the same, and each element c k,(i,j) , is the coordinate of the reference point a (i,j) A corresponding ocean or meteorological physical value; k is a time series, k Take 0, 1, 2, 3, ..., 23, 24, corresponding to 0 o'clock, 1 o'clock, 2 o'clock, 3 o'clock, ..., 23 o'clock, 24 o'clock respectively. C k Elements in c k,(i,j) It's the coordinates a (i,j) The corresponding ocean or meteorological physical value, that is, for a physical value, there will be 25 C k : C 0. C 1....... C 24 Corresponding to its values ​​at different times; D is the output value matrix, which is B The dimensions are the same, and each element d (p,q) is the coordinate of the input point b (p,q) A physical value corresponding to C k The physical values ​​represented have the same meaning; (3) Output value matrix D Solution when A , B , C k It is known that using A , B , C k Find D ; D The dimensions and B The dimensions are exactly the same. D Elements in d (p,q) The method to find is as follows: (3.1) Taking the entire longitude and latitude of the earth as the global domain, the matrix A The area is the center. A The boundary is extended along the longitude and latitude, thereby dividing the longitude and latitude of the earth into nine square grid areas, and the center, left side, right side, bottom, top, bottom left, bottom right, top left, and top right areas are recorded as area 0 to area 8 respectively; (3.2) From the input coordinate matrix B Read out the elements one by one b (p,q) The longitude and latitude value ( x , y ), and determine b (p,q) In which area between 0 and 8? When in zone 0, 1, or 2, A There are four adjacent coordinates in b (p,q) Surrounded by; When located in the other 6 areas, A The boundary is found with b (p,q) Two adjacent coordinates; (3.3) When located in region 0, 1 or 2, the four adjacent coordinates are in the reference value matrix C k Find the four corresponding values, and perform two-dimensional linear interpolation between the four values ​​and any one of the four coordinates. The result is d (p,q) ; When located in the other 6 areas, the reference value matrix is ​​based on the two adjacent coordinates. C k Find the two corresponding values, and perform one-dimensional linear interpolation between the two values ​​and any one of the two coordinates. The result is d (p,q) .

2. The map tile generation method based on FPGA technology as claimed in claim 1, characterized in that The step (3.2) determines b (p,q) In which area between 0 and 8 is it located according to the reference coordinate matrix A First determine the following 8 parameters: A The width is the number of columns M ; A The height is the number of rows N ; Longitude lower limit Lon min : -180°+longitude resolution / 2; Upper longitude limit Lon max : 180°-longitude resolution / 2; Lower latitude limit Lat min : -90°+latitude resolution / 2; Upper latitude limit Lat max : 90°-latitude resolution / 2; Longitude resolution Lon res ; Latitude resolution Lat res ; Then according to b (p,q) The longitude and latitude value ( x , y ) to determine which region it belongs to.

3. The map tile generation method based on FPGA technology as claimed in claim 2, characterized in that In the step (3.2), b (p,q) The area is divided into 9 situations: (0) When Lon min ≤ x < Lon max and Lat min ≤ y < Lat max hour, b (p,q) Falls in area 0; Calculate the subscript: , , According to the subscript j and i , from the reference coordinate matrix A Take out a (i,j) , a (i,j+1) , a (i+1,j) , a (i+1,j+1) , select A middle a (i,j) The coordinate value of the point ( x 1, y 1) Used for interpolation; According to the subscript j and i , from the benchmark value matrix C k Take out c k,(i,j) , c k,(i,j+1) , c k,(i+1,j) , c k,(i+1,j+1) , and determine z 1. z 2. z 3. z 4: c k,(i,j) = z 1 c k,(i,j+1) = z 2 c k,(i+1,j) = z 3 c k,(i+1,j+1) = z 4 Then go to step (3.3); (1) When x < Lon min and Lat min ≤ y < Lat max hour, b (p,q) Falls in zone 1; Calculate the subscript: , select A middle a (i,1) The coordinate value of the point ( x 1, y 1) Used for interpolation; Subscript by row i , from the reference value matrix C k Take out c k,(i,j) , c k,(i,j+1) , c k,(i+1,j) , c k,(i+1,j+1) , and determine z 1. z 2. z 3. z 4: c k,(i,j) = c k,(i,M) = z 1 c k,(i,j+1) = c k,(i,1) = z 2 c k,(i+1,j) = c k,(i+1,M) = z 3 c k,(i+1,j+1) = c k,(i+1,1) = z 4 Then go to step (3.3); (2) When x ≥ Lon max and Lat min ≤ y < Lat max hour, b (p,q) Falls in zone 2; Calculate the subscript: , select A middle a (i,M) The coordinate value of the point ( x 1, y 1) Used for interpolation; Subscript by row i , from the benchmark value matrix C k Take out c k,(i,j) , c k,(i,j+1) , c k,(i+1,j) , c k,(i+1,j+1) , and determine z 1. z 2. z 3. z 4: c k,(i,j) = c k,(i,M) = z 1 c k,(i,j+1) = c k,(i,1) = z 2 c k,(i+1,j) = c k,(i+1,M) = z 3 c k,(i+1,j+1) = c k,(i+1,1) = z 4 Then go to step (3.3); (3) When Lon min ≤ x < Lon max and y ≤ Lat min hour, b (p,q) Falls in zone 3; Calculate the subscript: , select A middle a (1,j) The coordinate value of the point ( x 1, y 1) Used for interpolation; According to column subscript j , from the benchmark value matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2: c k,(i,j) = c k,(1,j) = z 1 c k,(i,j+1) = c k,(1,j+1) = z 2 Then go to step (3.3); (4) When Lon min ≤ x < Lon max and y ≥ Lat max hour, b (p,q) Falls in zone 4; Calculate the subscript: , select A middle a (N,j) The coordinate value of the point ( x 1, y 1) Used for interpolation; According to column subscript j , from the reference value matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2: c k,(i,j) = c k,(N,j) = z 1 c k,(i,j+1) = c k,(N,j+1) = z 2 Then go to step (3.3); (5) When x < Lon min and y ≤ Lat min hour, b (p,q) Falls in zone 5; select A middle a (1,1) The coordinate value of the point ( x 1, y 1) Used for interpolation; From the benchmark matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2: c k,(i,j) = c k,(1,M) = z 1 c k,(i,j+1) = c k,(1,1) = z 2 Then go to step (3.3); (6) When x ≥ Lon max and y ≤ Lat min hour, b (p,q) Falls in zone 6; select A middle a (1,M) The coordinate value of the point ( x 1, y 1) Used for interpolation; From the benchmark matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2: c k,(i,j) = c k,(1,M) = z 1 c k,(i,j+1) = c k,(1,1) = z 2 Then go to step (3.3); (7) When x < Lon max and y ≥ Lat max hour, b (p,q) Falls in zone 7; select A middle a (N,1) The coordinate value of the point ( x 1, y 1) Used for interpolation; From the benchmark matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2: c k,(i,j) = c k,(N,M) = z 1 c k,(i,j+1) = c k,(N,1) = z 2 Then go to step (3.3); (8) When x ≥ Lon max and y ≥ Lat max hour, b (p,q) Falls in zone 8; select A middle a (N,M) The coordinate value of the point ( x 1, y 1) Used for interpolation; From the benchmark matrix C k Take out c k,(i,j) , c k,(i,j+1) , and determine z 1. z 2: c k,(i,j) = c k,(N,M) = z 1 c k,(i,j+1) = c k,(N,1) = z 2 Then go to step (3.3).

4. The map tile generation method based on FPGA technology as claimed in claim 3 is characterized by: The two-dimensional linear interpolation described in step (3.3) is as follows: , g That's what you want d (p,q) , The one-dimensional linear interpolation is as follows: , e That's what you want d (p,q) .

Citation Information

Patent Citations

  • Tile map generating method and updating method in virtual map, and apparatus

    CN107016924A

  • Urban area Gaussian projection remote sensing image acquisition method based on sky map

    CN117710606A