A method for fusing multi-source heterogeneous non-uniform spatiotemporal scale data

Abstract

This invention relates to a method for fusing multi-source heterogeneous and non-uniform spatiotemporal scale data. First, the coverage area and data time of raster data are determined, and a unified spatial reference frame is established between the raster data and station data. Observation stations within the coverage area are selected, and their observation values ​​are extracted. Then, the stations and raster data are spatially registered, and the rasters corresponding to the stations in the spatial reference are extracted. Using a seed propagation algorithm, the center raster value with deviation is assigned to neighboring null value rasters, resulting in deviation raster data consistent with the coverage area. Finally, the deviation raster data and the original raster data are processed through raster calculations to obtain fused raster data consistent with the coverage area. This method can fuse station data and raster data from different time and spatial scales into a dataset of the same spatiotemporal scale. The fused data has high accuracy and wide coverage, laying a data foundation for data analysis and numerical model construction in disciplines such as hydrology and meteorology.

Description

Technical Field

[0001] This invention relates to the fields of hydrology, meteorology, computer science and their interdisciplinary fields, and in particular to a method for fusing multi-source heterogeneous non-uniform spatiotemporal scale data. Background Technology

[0002] In traditional hydrological and meteorological research, the main data sources are hydrological measurement stations and meteorological monitoring stations. For example, hydrological measurement stations distributed along water bodies such as rivers, lakes, and reservoirs can continuously observe and record hydrological data such as water level, flow velocity, and flow rate at the cross-section where the station is located. Meanwhile, distributed meteorological monitoring stations can continuously acquire meteorological data such as precipitation, temperature, evapotranspiration, wind speed, and sunshine duration. However, since hydrological and meteorological stations are usually far apart, the spatial representativeness of high temporal resolution station data is not strong, and it cannot accurately reflect the spatial distribution characteristics of each monitoring element. This leads to certain limitations when conducting large-scale numerical simulations in hydrological, meteorological, and other fields.

[0003] In recent years, with the development and popularization of remote sensing technology, large-scale raster data based on remote sensing inversion has become increasingly easier to acquire and has been widely used in scientific research in fields such as hydrology and meteorology. However, due to the influence of meteorological conditions at the time of satellite imaging, the overall accuracy of raster data is not high. At the same time, due to the different on-orbit modes of satellites, the revisit cycles of various satellites are inconsistent. For example, MODIS satellites provide observation data every 1-2 days, LandSat satellites provide observation data every 6 days, and Sentinel satellites provide observation data every 17 days. Medium and high resolution satellites generally do not have a fixed shooting cycle and are basically mission-oriented shooting with high costs. As a result, the temporal resolution of current satellite remote sensing data and its inverted raster data is far from reaching the level of continuous observation by the station.

[0004] With the rapid development of computer technology, high-temporal-resolution, large-scale, high-precision simulation based on existing station data and raster data retrieved from remote sensing is a key bottleneck encountered by hydrology, meteorology, computer science, and their interdisciplinary fields in related numerical simulations and calculations. Therefore, how to fuse various types of station data and raster data with inconsistent structures at different spatiotemporal scales to obtain large-scale, high-precision raster data that combines the advantages of high accuracy of station data and wide coverage of raster data is a key issue that needs to be addressed in data analysis and numerical model construction in hydrology, meteorology, and other fields. Summary of the Invention

[0005] The problem to be solved by this invention is to provide a multi-source heterogeneous non-uniform spatiotemporal scale data fusion method to fuse station data and raster data of different time and spatial scales to obtain large-scale, high-precision data with high temporal resolution.

[0006] This invention adopts the following technical solution: a method for fusing multi-source heterogeneous non-uniform spatiotemporal scale data, comprising the following steps:

[0007] S1. Based on the raster data to be fused, determine the spatial coverage s and data time t of the raster data, and select the observation stations located within the spatial range of the raster data according to the spatial relationship.

[0008] S2. Based on the data time t of the raster data to be fused, extract the observation station data at time t from the long-term series measured data of the selected observation stations;

[0009] S3. Spatial registration is performed between the observation station and the raster data to be fused, and the raster corresponding to the observation station in the spatial reference frame is extracted; the deviation between the observation station data and the raster value of the corresponding raster is calculated, and the seed propagation algorithm is used to generate the deviation raster based on the raster position relationship and its deviation value.

[0010] S4. Perform raster overlay calculation on the original raster data to be fused and the bias raster data to obtain the fused raster data at time t.

[0011] Furthermore, based on the spatial location relationship, the stations and raster data are spatially registered, and the rasters that correspond to the stations in spatial relationship are selected. Based on the time t of the raster data, the station observation value corresponding to time t is extracted from the long sequence of station data. The raster deviation value is obtained by subtracting the value of the corresponding raster from the station observation value. This raster is defined as a known deviation raster, and the remaining rasters that do not match the stations in spatial relationship are defined as null rasters.

[0012] Furthermore, using a grid with a known deviation value as the center grid, the deviation value of the center grid is assigned to several adjacent null-value grids around the center grid based on the seed propagation algorithm. The assignment order starts from the grid directly north of the center grid and proceeds clockwise. Priority is given to assigning the deviation value of the nearest grid to the adjacent null-value grid, that is, the deviation value of the center grid is assigned to null-value grids that are 1, 2, ..., x grids away from the center grid in sequence.

[0013] Furthermore, if a null value grid is at the same distance as the center grid corresponding to the deviation values ​​of y known observation stations, then the null value grid is assigned the average value of the deviation values ​​of the corresponding center grids at the same distance.

[0014] That is, among the N selected stations, if y stations are at the same distance from the null value grid, then the average value of the corresponding deviation values ​​of the y stations is assigned to the null value grid.

[0015] Furthermore, after all null raster cells are assigned a deviation value, deviation raster data with the same spatial range as the original raster data is obtained; raster calculations are then performed on the deviation raster data and the original raster data to obtain the fused raster data.

[0016] Furthermore, the merged multi-period raster data is combined to form a raster dataset. Using the time node of each raster data as the x-axis and the time interval as the step size, a linear fitting method is used based on the different raster values ​​at the same spatial location before and after the time interval to obtain the raster value at a certain moment within the range of raster data in any time period. This achieves encrypted interpolation processing based on the merged raster data and improves the temporal resolution of the raster dataset.

[0017] The present invention also provides: an electronic device, comprising:

[0018] One or more processors;

[0019] A storage device on which one or more programs are stored;

[0020] When the one or more programs are executed by the one or more processors, the one or more processors implement any of the above-described methods for fusing multi-source heterogeneous non-uniform spatiotemporal scale data.

[0021] The present invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps in any of the above-mentioned multi-source heterogeneous non-uniform spatiotemporal scale data fusion methods.

[0022] Compared with the prior art, the present invention, employing the above technical solution, has the following technical effects:

[0023] (1) This invention fully considers the characteristics of high station data accuracy and wide raster data coverage. Based on the time of the raster data, the corresponding station data is extracted from the long sequence of station data. The raster data is corrected by the high-precision station data, which improves the overall accuracy of the raster data. It has high application value in practical work.

[0024] (2) The core idea of ​​the method used to calculate the deviation grid in this invention is to fully consider the influence of the known deviation grid on the null value grid. If a null value grid is closest to the center grid of a known deviation value, the deviation value of the center grid is assigned to the null value grid. If a null value grid is at the same distance from the center grids of several known deviation values, the average value of the deviation values ​​of the corresponding center grids is assigned to the null value grid. Since this method considers the influence of the distance between grids on the assignment of the null value grid, it can fit the variation characteristics between grids with different deviation values.

[0025] (3) The method in this invention has good adaptability and high economy. For example, it can generate deviation raster data with the same spatial resolution according to the spatial characteristics of the raster data to be fused, which can reduce the error caused by different spatial resolutions to a certain extent. At the same time, the deviation value assignment method based on seed spread algorithm and raster distance can generate deviation matrix consistent with the spatial characteristics of the original raster data in a faster manner, which is more efficient than other interpolation algorithms. Attached Figure Description

[0026] Figure 1 This is a flowchart of the multi-source heterogeneous non-uniform spatiotemporal scale data fusion method of the present invention;

[0027] Figure 2 This is a schematic diagram of the original raster data in an embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram of grid data distance measurement according to an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram showing the spatial registration of the site with the original raster data in an embodiment of the present invention;

[0030] Figure 5 This is a schematic diagram illustrating the conversion of raster data into a matrix according to an embodiment of the present invention.

[0031] Figure 6 This is a schematic diagram of deviation grid calculation and matrix in an embodiment of the present invention;

[0032] Figure 7 This is a schematic diagram illustrating the fusion of non-consistent spatiotemporal scale station data and raster data in an embodiment of the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the application will be further described in detail below with reference to the accompanying drawings. The described embodiments are only a part of the embodiments involved in this invention. All non-innovative embodiments based on these embodiments by other researchers in the art are within the protection scope of this invention. Furthermore, the step numbers in the embodiments of this invention are only set for ease of explanation and do not limit the order of the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0034] This invention provides a method for fusing multi-source heterogeneous non-uniform spatiotemporal scale data, such as... Figure 1 As shown, the details are as follows:

[0035] First, based on the raster data retrieved from remote sensing, determine the coverage area s and data time t of the raster data; unify the spatial reference system between the raster data and the station data, select N stations located within the coverage area s based on spatial relationships, and extract the observation value at time t from the long sequence of station data according to the raster data time.

[0036] Then, the stations and raster data are spatially registered, the raster corresponding to the stations in the spatial reference is extracted, and the observation value of each station at time t is subtracted from the corresponding raster value to obtain the initial deviation raster. This deviation raster is the center raster, and the other rasters are null rasters.

[0037] Next, using the seed propagation algorithm, the center grid value with the deviation value is assigned to the surrounding null grids. The null grids to be assigned a value are equidistant from the center grid, and their assigned value is the average of the deviation values ​​of the corresponding center grids at the same distance. Finally, deviation grid data consistent with the coverage area s is obtained. The deviation grid data and the original grid data are then combined through grid calculations, and the deviation grid data is added to the original grid data to obtain the fused grid data consistent with the coverage area s.

[0038] In this process, the raster data is processed by converting the raster into a matrix with the same number of rows and columns. The calculated matrix is ​​then converted back into a raster to obtain the final merged raster data.

[0039] In one embodiment of the present invention, the observation station data consists of measured values ​​recorded by various hydrological, meteorological, and environmental observation stations set up on the Earth's surface, and the raster data to be fused consists of various physically meaningful raster datasets interpreted from satellite remote sensing images, UAV multispectral images, etc. Guided by the goal of generating high-precision, wide-coverage raster data through the fusion of high-precision observation station data and wide-coverage raster data, a multi-source heterogeneous non-uniform spatiotemporal scale data fusion method is proposed, specifically including the following steps:

[0040] S-1 acquires the original raster data to be fused and the observation station data. Since both the observation station data and the raster data have a clear geographic coordinate system or projected coordinate system, they can be directly spatially overlaid to extract overlapping target objects, or spatially overlaid after coordinate transformation to extract overlapping target objects.

[0041] S-2 determines the spatial range and data time based on the attribute characteristics of raster data, selects observation stations with the same physical attribute measured values ​​as the raster data within the spatial range of the raster data, and extracts the station measured values ​​at the corresponding time from the measured data of the selected observation stations according to the raster data time t, forming a station dataset at time t within the raster data range.

[0042] The specific implementation process is as follows: Figure 2 As shown:

[0043] Figure 2 (a) is the raster data with corresponding physical meaning interpreted by remote sensing, which is the original raster data to be fused. In this embodiment, the raster data to be fused is a 6*6 raster matrix.

[0044] Figure 2 (b) The stations selected based on the characteristics of the raster data. In this embodiment, the observation stations are D1, D2, ..., D6, all of which are located within the raster data range;

[0045] Figure 2 (c) is the station dataset at time t within the raster data range. In this embodiment, the station dataset at time t is D1(t), D2(t), ..., D6(t).

[0046] Based on the spatial relationship between the site and the grid, S-3 subtracts the corresponding grid value from the site value at time t to obtain the deviation value of the corresponding grid at time t. According to the seed propagation algorithm, the deviation value is assigned to the surrounding adjacent grids, and finally the deviation grid data of the coverage area s is obtained.

[0047] S-3-1 In this embodiment, the grid in the raster data is square, and the grid corresponding to various hydrological, meteorological, and environmental observation stations on the Earth's surface is identified as the central grid. The distance measurement method between grids is as follows: Figure 3 As shown: the distance from the center grid is 1 for the 8 grids in the first ring outside the center grid, the distance from the center grid is 2 for the 16 grids in the second ring outside the center grid, the distance from the center grid is 3 for the 24 grids in the third ring outside the center grid, and the distance from the center grid is x for the 3*x grids in the xth ring outside the center grid.

[0048] Specifically, null grid cells at a distance x from the center grid are assigned values ​​in a clockwise direction starting from the north of the center grid.

[0049] When assigning values ​​to null-value graticles according to the method in S-3-1, S-3-2 prioritizes assigning the deviation value of the center graticle to the nearest null-value graticle. The deviation value of the center graticle is the measured value Dn(t) at time t minus the corresponding graticle value a. m,n The diagram illustrating the calculation of the deviation value of the original center grid is shown below. Figure 4 As shown.

[0050] S-3-3 raster data can be converted into a matrix for numerical calculations. If the raster data has m rows and n columns, then the raster data can be converted into an m row and n column matrix.

[0051] In this embodiment, the raster data is 6 rows and 6 columns of raster data, and its conversion into a matrix is ​​illustrated in the diagram below. Figure 5 As shown.

[0052] S-3-3 When a null raster is equidistant from several known center raster cells, the average deviation value of the center raster cells at the same distance is assigned to the null raster. This assignment is performed on all null raster cells using a seed propagation algorithm, ultimately resulting in deviation raster data and a deviation matrix that are consistent with the original raster data in terms of spatial resolution and coverage. The values ​​in the deviation raster matrix are the raster values ​​at the corresponding row and column positions in the deviation raster data, specifically as follows: Figure 6 As shown.

[0053] S-4 performs raster calculations on the original raster matrix and the offset raster matrix, which has the same spatial resolution and coverage as the original raster. This involves summing the raster values ​​at corresponding locations with the offset values ​​to obtain matrix data. Finally, the matrix is ​​further converted into raster data, resulting in the fused raster data. Specifically, as follows... Figure 7 As shown.

[0054] In summary, the multi-source heterogeneous non-uniform spatiotemporal scale data fusion method of this embodiment can generate raster data that combines the advantages of high precision of station data and wide coverage of raster data in a simple, fast and efficient manner. It overcomes the shortcomings of poor spatial representativeness of existing station data and low precision of raster data, making the representation results of raster data more consistent with the actual situation and suitable for data analysis and numerical model construction.

[0055] Furthermore, the fused multi-period raster data can be combined to form a raster dataset. Using the time nodes of each raster data as the x-axis and the time interval as the step size, and based on the different raster values ​​at the same spatial location before and after, a linear fitting method is used to interpolate the raster value at a certain moment within the range of raster data for any time period. This enables encrypted interpolation processing based on the fused raster data, resulting in higher temporal resolution of the raster dataset. This meets the needs of various models in fields such as hydrology, meteorology, and ecological environment for spatiotemporal feature analysis and trend extraction based on long-term remote sensing inversion parameters in scientific research and applications.

[0056] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for fusing multi-source heterogeneous non-uniform spatiotemporal scale data, characterized in that, Includes the following steps: S1. Based on the raster data to be fused, determine the spatial coverage s and data time t of the raster data, and select the observation stations located within the spatial range of the raster data according to the spatial relationship. S2. Based on the data time t of the raster data to be fused, extract the observation station data at time t from the long-term series measured data of the selected observation stations; S3. Spatial registration is performed between the observation station and the raster data to be fused, and the raster corresponding to the observation station in the spatial reference frame is extracted; the deviation between the observation station data and the raster value of the corresponding raster is calculated, and the seed propagation algorithm is used to generate the deviation raster based on the raster position relationship and its deviation value. The seed propagation algorithm includes the following sub-steps: S3.1 Assign the deviation value of the center grid to the nearest null grid: Assign the deviation value of the center grid to several adjacent null grids around the center grid, that is, null grids 1 away from the center grid. The assignment order starts from the grid due north of the center grid and proceeds clockwise. S3.2 The raster data consists of a raster with m rows and n columns. The deviation value of the center raster is sequentially assigned to the null raster at a distance of 2, ..., x raster cells from the center raster. S3.3 When a null value grid is at the same distance as the corresponding center grid of y known observation stations, the value of the null value grid is the average of the deviation values ​​of the corresponding center grids at the same distance. S3.

4. Assign values ​​to all null raster cells to obtain deviation raster data that is consistent with the spatial range of the raster data to be merged; S4. Perform raster overlay calculation on the original raster data to be fused and the bias raster data to obtain the fused raster data at time t.

2. The multi-source heterogeneous non-uniform spatiotemporal scale data fusion method according to claim 1, characterized in that, The raster data to be fused is a physically meaningful raster dataset interpreted from satellite remote sensing images and UAV multispectral images, while the observation station data consists of measured values ​​recorded by hydrological, meteorological, and environmental observation stations set up on the ground.

3. The multi-source heterogeneous non-uniform spatiotemporal scale data fusion method according to claim 2, characterized in that, In step S1, both the raster data to be fused and the observation stations have a clear geographic coordinate system or projected coordinate system. The spatial reference system of the raster data to be fused and the observation stations is unified. The overlapping target objects are extracted by spatial overlay according to the spatial position relationship, or the overlapping target objects are extracted by spatial overlay after coordinate transformation. N observation stations with the same physical attribute measured values ​​as the raster data within the spatial coverage area s are selected.

4. The multi-source heterogeneous non-uniform spatiotemporal scale data fusion method according to claim 1, characterized in that, Step S3 involves spatially registering the observation station with the raster data to be fused, ensuring that the raster data of the observation station and the raster data to be fused correspond spatially. The raster corresponding to the observation station in the spatial reference frame is extracted, and the data value of the observation station at time t is subtracted from the corresponding raster value to obtain the initial deviation raster. This initial deviation raster is the center raster corresponding to the observation station, and other raster values ​​that do not match the spatial relationship of the observation station are null raster values.

5. The multi-source heterogeneous non-uniform spatiotemporal scale data fusion method according to claim 1, characterized in that, Step S3: Convert the raster data into a raster matrix for numerical calculation to obtain biased raster data and a biased raster matrix that are consistent with the original raster data in terms of spatial resolution and coverage; the biased raster matrix is ​​m rows and n columns, and the matrix value is the raster value at the corresponding row and column position in the biased raster data.

6. The multi-source heterogeneous non-uniform spatiotemporal scale data fusion method according to claim 5, characterized in that, Step S4: Perform raster calculations on the deviation raster data and the original raster data to be fused: Based on the raster matrix and the deviation raster matrix, sum the raster values ​​and deviation values ​​at the corresponding positions, convert the resulting matrix data into raster data, and obtain the fused raster data, thereby realizing the mutual fusion of station data and raster data based on different time scales and spatial scales.

7. The multi-source heterogeneous non-uniform spatiotemporal scale data fusion method according to claim 1, characterized in that, The merged raster data from multiple periods is combined to form a raster dataset. Encrypted interpolation is then performed to improve the temporal resolution of the raster dataset. The method includes: using the time node of each raster data as the horizontal axis and the time interval as the step size, interpolating based on different raster values ​​at the same spatial location before and after the time interval using a linear fitting method to obtain the raster value at a certain moment within the range of raster data for any time period, thus realizing encrypted interpolation processing based on the merged raster data.

8. An electronic device, characterized in that, include: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the steps of any one of the multi-source heterogeneous non-uniform spatiotemporal scale data fusion methods of claims 1 to 7.

Images