A multi-source data fusion inversion PM10 system

Abstract

The application discloses a kind of multi-source data fusion inversion PM10 systems, its characteristics are: the system includes the multi-source data fusion unit for monitoring site machine model training, the multi-source data fusion unit for target physical area inversion PM10;The multi-source data fusion unit for target physical area inversion PM10 utilizes the machine model trained to invert the PM10 data of target physical area.The application solves the problem that the monitoring ability of traditional PM10 ground monitoring station is unbalanced, and the influence range of monitoring station data cannot be improved by traditional PM10 ground monitoring station;Balanced development of the monitoring ability of PM10 in the whole country is realized, and the equipment cost and labor cost of constructing ground monitoring station are effectively saved.The PM10 inversion technology of the application can not only monitor the present situation of the pollution distribution characteristics of "point", but also monitor the present situation of the pollution distribution characteristics of regional "surface".

Description

Technical Field

[0001] This invention belongs to the field of remote sensing data processing and technology application, and specifically relates to a PM10 inversion system based on multi-source data fusion. Background Technology

[0002] PM10 refers to inhalable particulate matter with an aerodynamic equivalent diameter of 10 micrometers or less, encompassing both solid and liquid particles suspended in the air. The harmful effects of this inhalable particulate matter on the human body include: Impact on respiratory diseases: Increased PM10 concentrations in the atmosphere can easily lead to upper respiratory tract infections, rhinitis, chronic pharyngitis, chronic bronchitis, bronchial asthma, emphysema, and pneumoconiosis. Impact on cardiovascular diseases: It increases blood viscosity and certain albumins in the blood, potentially causing thrombosis. Carcinogenic, mutagenic, and disabling effects: Incomplete combustion of fossil fuels such as petroleum and coal, as well as organic matter such as wood and tobacco, produces polycyclic aromatic hydrocarbons (PAHs). Because PAHs have carcinogenic, mutagenic, and disabling effects, they pose a significant threat to human health. Benzo(a)pyrene (BaP), a representative example, is the most carcinogenic substance, capable of inducing skin cancer, lung cancer, and stomach cancer.

[0003] The depth and breadth of existing PM10 monitoring technologies are far behind the evolving needs of human disease control, as evidenced by the following:

[0004] First, the development of PM10 monitoring capabilities is uneven. There are significant differences in basic PM10 monitoring capabilities between regions, levels, and urban and rural areas. Monitoring equipment in some central and western regions is outdated, and experimental conditions are rudimentary. County-level monitoring capabilities are insufficient to meet the requirements of law enforcement and emergency monitoring tasks, and rural environmental monitoring is just beginning. National and key regional monitoring technology experimental capabilities are inadequate, and development space is limited. There is a shortage of application infrastructure for remote sensing satellites, a lack of unified planning for the informatization of the national monitoring system, and data barriers have not been substantially broken down. Effective collection and intelligent analysis of massive amounts of monitoring data urgently need to be strengthened.

[0005] Second, traditional PM10 ground monitoring stations cannot expand the scope of influence of monitoring station data: ① They cannot obtain the current status of regional pollution distribution characteristics; ② They can conduct all-weather, large-scale observations, but the accuracy is not high and they cannot provide data products; ③ They cannot provide an important source of information for comprehensive and three-dimensional monitoring of air pollution. Summary of the Invention

[0006] This invention addresses the problems existing in the prior art by proposing a multi-source data fusion-based PM10 inversion system. The aim is to solve the problem of uneven development of monitoring capabilities of traditional PM10 ground monitoring stations and the inability of traditional PM10 ground monitoring stations to expand the influence range of monitoring station data.

[0007] To solve its technical problems, the present invention proposes the following technical solutions:

[0008] A multi-source data fusion PM10 inversion system is characterized in that the system includes a multi-source data fusion unit for training a machine model for a specified site and a multi-source data fusion unit for inverting PM10 in a target physical area; the multi-source data fusion unit for inverting PM10 in a target physical area uses the trained machine model to invert PM10 data in the target physical area.

[0009] Furthermore, the multi-source data fusion unit for training the machine learning model at the monitoring station includes a monitoring station latitude and longitude module, a satellite remote sensing data module, a monitoring station raw data module, a monitoring station raw data imputation and interpolation module, a two-dimensional data to "point" data conversion module, a monitoring station PM10 detection module, and a machine learning model training module. The monitoring station raw data module obtains the raw meteorological data and raw AOD aerosol data of the monitoring station from the satellite remote sensing data module based on the monitoring station's latitude and longitude data. The monitoring station raw data imputation and interpolation module interpolates and / or imputes the raw meteorological data of the monitoring station, transforming the low-resolution raw data into high-resolution two-dimensional data. The two-dimensional data to "point" data conversion module converts the high-resolution two-dimensional data of the monitoring station... The system refines latitude and longitude, and searches for high-resolution meteorological data and high-resolution AOD data of the monitoring station "points" based on their latitude and longitude. The PM10 detection module of the monitoring station detects the PM10 data of the monitoring station "points." The machine learning model training module uses the high-resolution meteorological data and high-resolution AOD data of the monitoring station "points" as input values ​​for machine learning model training, and uses the PM10 data of the monitoring station "points" as output values ​​for machine learning model training. The original air data of the monitoring station includes the original meteorological data and the original AOD data of the monitoring station. The high-resolution two-dimensional data refers to the high-resolution two-dimensional meteorological data and the high-resolution two-dimensional AOD data of the monitoring station.

[0010] Furthermore, the monitoring station's raw data module includes a satellite remote sensing data acquisition module, a monitoring station latitude and longitude acquisition module, a monitoring station raw 3D meteorological data module, a monitoring station raw 3D AOD data module, a monitoring station raw 3D meteorological data to 2D conversion module, and a monitoring station raw 3D AOD data to 2D conversion module. The monitoring station raw 3D meteorological data module and the monitoring station raw 3D AOD data module acquire the monitoring station's raw 3D meteorological data and raw 3D AOD data from satellite remote sensing data based on the monitoring station's latitude and longitude data, respectively. The 3D meteorological data format is date, longitude, and latitude; the 3D AOD data format is date and hour, longitude, and latitude; the monitoring station raw 3D meteorological data to 2D conversion module converts the 3D meteorological data to 2D meteorological data according to date; the monitoring station raw 3D AOD data to 2D conversion module converts the 3D AOD data to 2D AOD data according to date and hour.

[0011] Furthermore, the monitoring station raw data filling and interpolation module includes a module for reading raw meteorological data from the monitoring station, a module for reading raw AOD data from the monitoring station, an interpolation algorithm module, a mean algorithm module, a high-resolution two-dimensional meteorological data module for the monitoring station, a low-resolution AOD data mean module, and a high-resolution two-dimensional AOD data module for the monitoring station. The low-resolution AOD data mean module performs mean processing on the raw AOD data according to the mean algorithm to obtain the low-resolution AOD data mean module. The high-resolution two-dimensional AOD data module for the monitoring station interpolates the low-resolution AOD data mean data to obtain a high-resolution two-dimensional AOD matrix for the monitoring station. The high-resolution two-dimensional meteorological data module for the monitoring station interpolates the raw meteorological data according to the interpolation algorithm to obtain high-resolution two-dimensional meteorological data for the monitoring station.

[0012] Furthermore, the two-dimensional data conversion to "point" data module includes a high-resolution meteorological data subdivision latitude and longitude module for monitoring stations, a high-resolution AOD data subdivision latitude and longitude module for monitoring stations, a high-resolution meteorological data module for "points" of monitoring stations, and a high-resolution AOD data module for "points" of monitoring stations. The high-resolution meteorological data module for latitude and longitude of monitoring stations and the high-resolution AOD data module for latitude and longitude of monitoring stations search for the data corresponding to the latitude and longitude of the monitoring stations in the high-resolution meteorological data subdivision latitude and longitude module and the high-resolution AOD data subdivision latitude and longitude module for monitoring stations, respectively, to obtain the high-resolution meteorological data and the high-resolution AOD data of "points" of the monitoring stations.

[0013] Furthermore, the mean algorithm module includes a module for reading AOD two-dimensional raw data, a low-resolution counting matrix module, a low-resolution total matrix module, and a low-resolution AOD mean module. The low-resolution counting matrix module divides the AOD two-dimensional raw data into raster data and counts the number of valid AOD data in each raster at each latitude and longitude according to the date and hour. The low-resolution total matrix module divides the AOD two-dimensional raw data into raster data and counts the data value of valid AOD in each raster at each latitude and longitude according to the date and hour. The low-resolution AOD mean module divides the value of each raster in the low-resolution counting matrix by the value of each raster in the low-resolution counting matrix to obtain the value of each raster in the low-resolution mean two-dimensional matrix, thereby obtaining the low-resolution AOD mean two-dimensional matrix.

[0014] Furthermore, the multi-source data fusion unit for PM10 inversion in the target physical region includes a target physical region shapefile module, a satellite remote sensing data module, a target physical region raw data module, a target physical region raw data filling and interpolation module, a target physical region high-resolution data 2D to 1D conversion module, a machine model PM10 inversion module, and a target physical region PM10 graphic generation module. The target physical region raw data module obtains the target physical region's 3D raw meteorological data and 3D raw AOD data from the satellite remote sensing data module based on the target physical region shapefile, and converts them into 2D raw meteorological data and 2D raw AOD aerosol data. The target physical region raw data filling and interpolation module fills and interpolates the target physical region raw data, making the low-resolution raw 2D meteorological data and low-resolution AOD aerosol data more accurate and concise. The original two-dimensional AOD data of the target physical area is transformed into high-resolution two-dimensional meteorological data and high-resolution two-dimensional AOD data; the target physical area high-resolution data two-dimensional to one-dimensional module converts the high-resolution two-dimensional meteorological data and high-resolution two-dimensional AOD data into high-resolution one-dimensional meteorological data and high-resolution one-dimensional AOD data, respectively; the machine model inversion PM10 module obtains high-resolution one-dimensional meteorological data and high-resolution one-dimensional AOD data from the target physical area high-resolution data two-dimensional to one-dimensional module as input data for the machine model application stage, the machine model inverts the PM10 one-dimensional array, and converts the inverted PM10 one-dimensional array into an inverted PM10 two-dimensional array; the target area PM10 graphics generation module generates the target physical area PM10 graphics based on the inverted PM10 two-dimensional array data and the data in the cropping method file.

[0015] Furthermore, the target physical area high-resolution data 2D to 1D conversion module includes: a target physical area high-resolution 2D meteorological data module for the date, a target physical area high-resolution 2D AOD data module for the date, a high-resolution 1D meteorological data module, and a high-resolution 1D AOD data module; the high-resolution 1D meteorological data module converts the target physical area high-resolution 2D meteorological data for the date into 1D data; the high-resolution 1D AOD data module converts the target physical area high-resolution 2D AOD array for the date into 1D data.

[0016] Furthermore, the machine model inversion PM10 module includes a machine model, a PM10 one-dimensional array inversion submodule, and a PM10 two-dimensional array inversion submodule. The machine model inversion PM10 module takes the high-resolution one-dimensional meteorological data and high-resolution one-dimensional AOD data as input data for the machine model and outputs the PM10 one-dimensional array. The PM10 two-dimensional array inversion submodule converts the PM10 one-dimensional array into a two-dimensional array.

[0017] Furthermore, the module for generating the PM10 graphic of the target physical region includes a gdal clipping method module, a PM10 two-dimensional array inversion module, a clipping module, and a module for generating the PM10 graphic file of the target physical region inversion. The clipping module obtains data from the gdal clipping method module and the PM10 two-dimensional array inversion module to generate the PM10 graphic file of the target physical region inversion.

[0018] Advantages and effects of the present invention

[0019] 1. This invention achieves balanced development of PM10 monitoring capabilities nationwide: This invention organically combines ground monitoring station technology, Kriging interpolation technology, missing data filling technology, fusion technology, "two-dimensional matrix" and "one-dimensional matrix" mutual conversion technology, and machine model technology. It successfully uses multi-source data fusion and machine model inversion to retrieve PM10 data with good results, achieving balanced development of PM10 monitoring capabilities nationwide. For areas with significant differences in basic PM10 monitoring capabilities between regions, levels, and urban and rural areas, and for some central and western regions with aging monitoring equipment and rudimentary experimental conditions lacking ground monitoring stations, the PM10 retrieval technology of this invention can still monitor local PM10 levels. This greatly solves the urgent need for PM10 monitoring in remote mountainous areas and areas with poor infrastructure, and also effectively saves on equipment and labor costs for building ground monitoring stations.

[0020] 2. This invention expands the reach of ground-based monitoring station data: Applying the PM10 retrieval technology of this invention, it is possible not only to monitor the current state of pollution distribution characteristics at specific "points" but also at the overall pollution distribution characteristics across a region. Specifically, using a machine learning model based on the monitoring station, and acquiring the daily raw meteorological data and hourly raw AOD data for the province, city, district, and county where PM10 data is to be retrieved, a PM10 distribution map for that province, city, district, and county can be obtained through multi-source data fusion. This allows data from a single station to be used nationwide, thus expanding the reach of ground-based monitoring station data. Attached Figure Description

[0021] Figure 1 This invention provides a framework for a multi-source data fusion-based PM10 inversion system.

[0022] Figure 2 This is a schematic diagram of the multi-source data fusion unit used for training machine models at monitoring sites according to the present invention;

[0023] Figure 3 This is a schematic diagram of the original data module of the monitoring station of the present invention;

[0024] Figure 4 This is a schematic diagram of the monitoring site filling and interpolation module of the present invention;

[0025] Figure 5 This is a schematic diagram of the mean algorithm module of the present invention;

[0026] Figure 6 This is a schematic diagram of the multi-source data fusion unit used for PM10 inversion in the target physical region according to the present invention;

[0027] Figure 7 This is a schematic diagram of the original data module of the target physical region of the present invention;

[0028] Figure 8 This is a schematic diagram of the target physical region filling and interpolation module of the present invention;

[0029] Figure 9 This is a schematic diagram of the high-precision two-dimensional to one-dimensional data conversion module for the target physical region of this invention;

[0030] Figure 10 This is a flowchart of a method for retrieving PM10 from multi-source data fusion according to the present invention;

[0031] Figure 11 shows a schematic diagram of the original low-resolution meteorological data three-dimensional data format. picture;

[0032] Figure 12 illustrates the image fusion process of low-resolution AOD data according to the present invention. picture;

[0033] Figure 13 illustrates how the fused low-resolution AOD data is interpolated to high-resolution data according to the present invention. picture;

[0034] Figure 14 illustrates the dimensionality reduction of high-resolution meteorological data and high-resolution AOD data according to the present invention. picture;

[0035] Figure 15 is a schematic diagram of PM10 output from the machine model of the present invention being converted into a two-dimensional array of PM10. picture;

[0036] Figure 16 is a schematic diagram of the PM10 graphic file of the target physical region generated by the present invention. Detailed Implementation

[0037] Design principle of the invention

[0038] 1. The innovation of this invention lies in the organic combination of "effective data" obtained through multi-source fusion and machine learning, achieving a new effect of improving the depth and breadth of PM10 monitoring. The key point lies in the design of the "effective data." Machine learning is merely a tool; during the training phase of the machine learning model, if the data input is not effective, the output will also not be "effective." Only when the data fed to the machine model is accurate and effective can the training result of the machine model be effective.

[0039] 2. Design Challenges of this Invention: 1) Challenge 1: The contradiction between low-resolution raw "area" data and high-precision "point" data; The "point" data refers to the data detected by the monitoring station. Since the data range that a PM10 monitoring station can detect is only a small local area, it is called "point" data. "High precision" means that the monitoring station detects valid data. This high-precision "point" data is used as the output data for the machine learning training stage, and the input values ​​for the machine learning training stage are also required to be high-precision "point" data from the monitoring station that match the output values. Since the input data comes from the raw meteorological data and raw AOD data around the monitoring station, both of which are low-resolution "area" data, a contradiction arises between raw "area" data and high-precision "point" data: it is difficult to capture high-precision "point" meteorological data and AOD data in the low-resolution raw "area" data; 2) Challenge 2: The contradiction between low-resolution missing data and high-precision "point" data; The raw AOD data not only has very low resolution but also missing data (blank, no data), and the degree of missing data varies in the same physical area throughout the 24 hours of the day. For the same physical area, some time points throughout the day are severely missing, some are only partially missing, and some are missing zero. Therefore, it is difficult to find an AOD value that can represent the current physical area. If the physical area happens to be a monitoring station, it is difficult to provide an accurate input value for machine learning. 3) The third difficulty is the contradiction between high-precision "point" data and high-precision "area" data. Point data only represents the PM10 value of a small local area of ​​the monitoring station. This small local area is only a fraction of the area data of the entire province or city. However, we ultimately hope to obtain PM10 area data that can represent the entire province or city through machine learning models. This leads to the contradiction between "point" data and area data.

[0040] 3. The solution of this invention:

[0041] First, the contradiction between low-resolution raw data and high-precision "point" data, as well as the contradiction between low-resolution missing data and high-precision "point" data, needs to be resolved. 1) For raw meteorological data, the Kriging interpolation algorithm is used to increase the number of rows and columns of the two-dimensional matrix, thereby achieving the conversion from low resolution to high resolution; 2) For raw AOD data, it needs to be divided into two steps: First, fill in the missing values ​​and then perform interpolation calculation. The principle of filling in missing values ​​is that because the degree of missing data in the same physical area varies throughout the day, the main goal is to calculate the average value: count the number of times valid data appears in the same physical area for each hour of the 24 hours of the day, then accumulate the AOD values ​​of that physical area, and then calculate the average value for each hour. The second step is to fill in the missing values, but the data is still at a low resolution. It is necessary to use the Kriging interpolation algorithm to convert the low resolution to high resolution. 3) After the original meteorological data has completed the conversion from low resolution to high resolution and the original AOD data has completed the conversion from AOD missing values ​​to low resolution to high resolution, the latitude and longitude in the raster should also be refined accordingly. As for the degree of refinement of the raster latitude and longitude, the standard is that the latitude and longitude of the monitoring station can be found in the high resolution meteorological data raster data and the high resolution AOD raster data. The degree of refinement of the raster is the same as the degree of refinement of the latitude and longitude.

[0042] Secondly, this invention addresses the contradiction between high-precision "point" data and high-precision "area" data. The ultimate goal of this invention is to train a machine learning model at a monitoring station to obtain the PM10 inversion image for any province, city, district, or county nationwide. The solution from "point" to "area" is to transform the "area" into "lines," and then transform the "lines" back into "areas." 1) Obtain the original area (raster) data of meteorological and AOD data for each city; 2) Obtain the high-precision area (raster) data of meteorological and AOD data for each city; 3) Convert the two-dimensional matrix of each high-precision area data into a one-dimensional matrix of "line" data. This "line" data can be "column" or "row" data; this invention uses "column" data. The method for transforming the "area" into "lines" is as follows: Figure 6 As shown, similarly, Figure 6 It is also possible to convert "lines" into "areas". Methods for obtaining PM10 inversion area data from other provinces, cities, and counties include... Figure 5 As shown, the six polygonal or raster data on the left are transformed into six lines or six one-dimensional "columns" of multiple rows and columns. These six "columns" are then input into a pre-trained machine learning model. The machine learning model outputs a column of inverted PM10 data, which is shown below. Figure 5 The rightmost column of data, which represents the inverted PM10 data, is then converted into a two-dimensional matrix of the inverted PM10, as shown in the image. Figure 7As shown in the upper right corner. The key to transforming a "surface" into a "line" and then back into a "surface" is to consider a "line" as being composed of multiple "points." Since machine learning models are trained based on "points," when "lines" and "points" are linked, the training results of the "point-based" machine learning model can be utilized. In practice, the data of multiple points on the "line" are read one by one into the machine learning model as input data. The machine learning model also outputs a series of points one by one, which together form the "line" output by the machine learning model.

[0043] 4. Processing flow of this invention: ① Select a site, which is a site with facilities and technical capabilities to detect local PM10, and convert the low-resolution "area" data around the "site" into high-resolution "area" data; ② Search for the "site's" data in the high-resolution "area" data; ③ Train the machine model using the "site's" data; ④ Select the target province / city for PM10 inversion, obtain the low-resolution "area" data of the target province / city, and convert it into high-resolution "area" data; ⑤ Reduce the dimensionality of the "area" data of the target province / city for PM10 inversion to a one-dimensional array, which is N*N rows and 1 column of one-dimensional data; * The N x 1 column one-dimensional data includes multiple N x N x 1 column one-dimensional arrays of meteorological data and an N x N x 1 column one-dimensional array of AOD data. ⑥ Use all the above one-dimensional arrays as input data for the machine model. The machine model outputs an N x N x 1 column array of inverted PM10. ⑦ Restore the N x N x 1 column array of inverted PM10 output by the machine model to an N x N array. ⑧ Input the N x N array of PM10 inverted by the machine model and the target physical area shp file into the "gdal cropping method file". The gdal cropping method file will crop out the PM10 graphic file of the target physical area.

[0044] Based on the above-mentioned inventive principles, this invention designs a multi-source data fusion inversion PM10 system, such as... Figure 1-9 As shown, the system is characterized by including a multi-source data fusion unit for training machine models at monitoring stations and a multi-source data fusion unit for retrieving PM10 from the target physical area; the multi-source data fusion unit for retrieving PM10 from the target physical area uses the trained machine model to retrieve PM10 data from the target physical area.

[0045] Furthermore, the multi-source data fusion unit used for training the machine model at the monitoring site, such as... Figure 2As shown, the system includes a monitoring station latitude and longitude module, a satellite remote sensing data module, a monitoring station raw data module, a monitoring station raw data imputation and interpolation module, a 2D data to "point" data conversion module, a monitoring station PM10 detection module, and a machine model training module. The monitoring station raw data module obtains the raw meteorological data and raw AOD aerosol data of the monitoring station from the satellite remote sensing data module based on the monitoring station's latitude and longitude data. The monitoring station raw data imputation and interpolation module interpolates and / or imputes the raw meteorological data of the monitoring station, transforming the low-resolution raw data into high-resolution 2D data. The 2D data to "point" data conversion module subdivides the high-resolution 2D data of the monitoring station into latitude and longitude, and based on the monitoring station... The system searches for high-resolution meteorological data and high-resolution AOD data of a monitoring station at a latitude and longitude coordinate. A PM10 detection module at the monitoring station detects the PM10 data of that monitoring station. A machine learning model training module uses the high-resolution meteorological data and high-resolution AOD data of the monitoring station as input values ​​for machine learning model training, and uses the PM10 data of the monitoring station as the output value for machine learning model training. The original air quality data of the monitoring station includes the original meteorological data and the original AOD data of the monitoring station. The high-resolution two-dimensional data refers to the high-resolution two-dimensional meteorological data and the high-resolution AOD two-dimensional data of the monitoring station.

[0046] Furthermore, the raw data module of the monitoring station is as follows: Figure 3 As shown, the system includes modules for acquiring satellite remote sensing data, acquiring the latitude and longitude of monitoring stations, acquiring raw 3D meteorological data of monitoring stations, acquiring raw 3D AOD data of monitoring stations, converting raw 3D meteorological data to 2D data, and converting raw 3D AOD data to 2D data. The raw 3D meteorological data module and the raw 3D AOD data module acquire the raw 3D meteorological data and raw 3D AOD data of the monitoring station from satellite remote sensing data based on the latitude and longitude data of the monitoring station, respectively. The 3D meteorological data format is date, longitude, and latitude; the 3D AOD data format is date and hour, longitude, and latitude. The raw 3D meteorological data to 2D data module converts the 3D meteorological data to 2D meteorological data according to date; the raw 3D AOD data to 2D data module converts the 3D AOD data to 2D AOD data according to date and hour.

[0047] Furthermore, the monitoring station's raw data filling and interpolation module, as follows: Figure 4As shown, the system includes a module for reading raw meteorological data from monitoring stations, a module for reading raw AOD data from monitoring stations, an interpolation algorithm module, a mean algorithm module, a high-resolution two-dimensional meteorological data module for monitoring stations, a low-resolution AOD data mean module, and a high-resolution two-dimensional AOD data module for monitoring stations. The low-resolution AOD data mean module performs mean processing on the raw AOD data according to the mean algorithm to obtain the low-resolution AOD data mean module. The high-resolution two-dimensional AOD data module for monitoring stations interpolates the low-resolution AOD data mean data to obtain the high-resolution AOD two-dimensional matrix of monitoring stations. The high-resolution two-dimensional meteorological data module for monitoring stations interpolates the raw meteorological data according to the interpolation algorithm to obtain high-resolution two-dimensional meteorological data of the monitoring stations.

[0048] Furthermore, the module for converting the two-dimensional data into "point" data is as follows: Figure 2 As shown, the system includes a high-resolution meteorological data subdivision latitude and longitude module for monitoring stations, a high-resolution AOD data subdivision latitude and longitude module for monitoring stations, a high-resolution meteorological data module for "points" of monitoring stations, and a high-resolution AOD data module for "points" of monitoring stations. The high-resolution meteorological data module for latitude and longitude and the high-resolution AOD data module for latitude and longitude of monitoring stations search for the data corresponding to the latitude and longitude of the monitoring stations in the high-resolution meteorological data subdivision latitude and longitude module and the high-resolution AOD data subdivision latitude and longitude module for monitoring stations, respectively, to obtain the high-resolution meteorological data and the high-resolution AOD data of the "points" of the monitoring stations.

[0049] Furthermore, the mean algorithm module is as follows: Figure 5 As shown, the system includes a module for reading AOD two-dimensional raw data, a low-resolution counting matrix module, a low-resolution total matrix module, and a low-resolution AOD mean module. The low-resolution counting matrix module divides the AOD two-dimensional raw data into raster data and counts the number of valid AOD data in each raster at each latitude and longitude according to the date and hour. The low-resolution total matrix module divides the AOD two-dimensional raw data into raster data and counts the data value of valid AOD in each raster at each latitude and longitude according to the date and hour. The low-resolution AOD mean module divides the value of each raster in the low-resolution counting matrix by the value of each raster in the low-resolution counting matrix to obtain the value of each raster in the low-resolution mean two-dimensional matrix, thereby obtaining the low-resolution AOD mean two-dimensional matrix.

[0050] Furthermore, the multi-source data fusion unit for retrieving PM10 from the target physical region, as shown in the example... Figure 6As shown, the system includes a target physical region shapefile module, a satellite remote sensing data module, a target physical region raw data module, a target physical region raw data imputation and interpolation module, a target physical region high-resolution data 2D to 1D conversion module, a machine model inversion PM10 module, and a target physical region PM10 graphic generation module. The target physical region raw data module obtains 3D raw meteorological data and 3D raw AOD data of the target physical region from the satellite remote sensing data module based on the target physical region shapefile, and converts them into 2D raw meteorological data and 2D raw AOD aerosol data. The target physical region raw data imputation and interpolation module imputs and interpolates the target physical region raw data, transforming the low-resolution raw 2D meteorological data and low-resolution raw 2D AOD data into... The system consists of high-resolution two-dimensional meteorological data and high-resolution two-dimensional AOD data. The target physical area high-resolution data two-dimensional to one-dimensional module converts the high-resolution two-dimensional meteorological data and high-resolution two-dimensional AOD data into high-resolution one-dimensional meteorological data and high-resolution one-dimensional AOD data, respectively. The machine model inversion PM10 module obtains high-resolution one-dimensional meteorological data and high-resolution one-dimensional AOD data from the target physical area high-resolution data two-dimensional to one-dimensional module as input data for the machine model application stage. The machine model inverts a one-dimensional PM10 array and converts the inverted one-dimensional PM10 array into a two-dimensional PM10 array. The target area PM10 graphics generation module generates a PM10 graphic of the target physical area based on the inverted two-dimensional PM10 array data and the data from the cropping method file.

[0051] Supplementary Note 1:

[0052] The target physical region raw data module is as follows Figure 7 As shown, the system includes a target physical area shapefile, a satellite remote sensing data acquisition module, a target physical area raw 3D meteorological data module, a target physical area raw 3D AOD data module, a 2D raw meteorological data module, and a 2D raw AOD data module. The target physical area raw 3D meteorological data module and the target physical area raw 3D AOD data module acquire the target physical area's raw 3D meteorological data and raw 3D AOD data from satellite remote sensing data based on the target physical area shapefile, respectively. The 3D meteorological data format is date, longitude, and latitude; the 3D AOD data format is date, hour, longitude, and latitude. The 2D raw meteorological data module converts the 3D meteorological data into 2D meteorological data according to date; the 2D raw AOD data module converts the 3D AOD data into 2D AOD data according to date and hour.

[0053] Supplementary Note 2:

[0054] The target physical region original data filling and interpolation module is as follows: Figure 8 As shown, the system includes a module for reading raw meteorological data of the target physical area, a module for reading raw AOD data of the target physical area, an interpolation algorithm module, a averaging algorithm module, a high-resolution two-dimensional meteorological data module for the target physical area, a low-resolution AOD data averaging module, and a high-resolution two-dimensional AOD data module for the target physical area. The low-resolution AOD data averaging module performs averaging processing on the raw AOD data according to the averaging algorithm to obtain the low-resolution AOD data averaging module. The high-resolution AOD two-dimensional data module interpolates the low-resolution AOD data averaging data to obtain the high-resolution AOD two-dimensional matrix of the target physical area. The high-resolution meteorological two-dimensional data module interpolates the raw meteorological data according to the interpolation algorithm to obtain the high-resolution meteorological two-dimensional data of the target physical area.

[0055] Furthermore, the target physical region high-resolution data two-dimensional to one-dimensional conversion module is as follows: Figure 9 As shown, the system includes: a high-resolution two-dimensional meteorological data module for the target physical area on the given date, a high-resolution two-dimensional AOD data module for the target physical area on the given date, a high-resolution one-dimensional meteorological data module, and a high-resolution one-dimensional AOD data module; the high-resolution one-dimensional meteorological data module converts the high-resolution two-dimensional meteorological data for the target physical area on the given date into one-dimensional data; the high-resolution one-dimensional AOD data module converts the high-resolution two-dimensional AOD array for the target physical area on the given date into one-dimensional data.

[0056] Furthermore, the machine model inversion PM10 module is as follows: Figure 6 As shown, the system includes a machine model, a one-dimensional PM10 array inversion submodule, and a two-dimensional PM10 array inversion submodule. The machine model PM10 inversion submodule takes the high-resolution one-dimensional meteorological data and high-resolution one-dimensional AOD data as input data to the machine model and outputs a one-dimensional PM10 array. The two-dimensional PM10 array inversion submodule converts the one-dimensional PM10 array into a two-dimensional array.

[0057] Furthermore, the module for generating the PM10 graphics of the target physical region is as follows: Figure 6 As shown in Figure 16, the system includes a gdal clipping method module, a module for obtaining and retrieving a two-dimensional PM10 array, a clipping module, and a module for generating a target physical region inverted PM10 graphic file. The clipping module obtains data from the gdal clipping method module and the PM10 array module to generate a target physical region inverted PM10 graphic file.

[0058] Based on the above-mentioned multi-source data fusion-based PM10 inversion system, this invention also designs a multi-source data fusion-based PM10 inversion method, as follows: Figure 10As shown, its characteristics include the following steps:

[0059] Step 1: Input data into the PM10 inversion system and convert the formats of different remote sensing image data;

[0060] Step 2: Perform interpolation calculations on the input data of the monitoring stations, fill in the missing values ​​of the original images, and obtain the PM10 mean of the monitoring stations, the high-resolution meteorological data two-dimensional matrix of the monitoring stations, and the high-resolution AOD data two-dimensional matrix of the monitoring stations. The training phase of machine learning is also completed.

[0061] Step 3: Perform inversion based on the interpolated and filled data to generate the final inversion result.

[0062] Furthermore, in step one, data is input into the PM10 inversion system, and format conversion is performed for different remote sensing image data; the specific process is as follows:

[0063] 1) Input the latitude and longitude data of the monitoring stations into the inversion system;

[0064] 2) Acquire raw 3D meteorological data and raw 3D AOD data from the monitoring stations;

[0065] 3) Convert the original 3D meteorological data and original 3D AOD data of the monitoring stations into different remote sensing image formats;

[0066] The meteorological data includes air pressure, humidity, wind speed, temperature, and wind direction data, and the AOD data is aerosol data.

[0067] Furthermore, step two involves interpolating the input data from the monitoring stations, filling in missing values ​​in the original images, and obtaining the PM10 mean, a high-resolution two-dimensional matrix of meteorological data for each monitoring station, and a high-resolution two-dimensional matrix of AOD data for each monitoring station. This completes the machine learning training phase. The specific process is as follows:

[0068] 1) As shown in Figure 11, the original three-dimensional meteorological data and original three-dimensional AOD data of the monitoring station are converted into original two-dimensional meteorological data and original two-dimensional AOD data. The meteorological data in this three-dimensional data format is composed of three-dimensional data according to date, longitude, and latitude. The AOD data in this three-dimensional data format is composed of three-dimensional data according to date and hour, longitude, and latitude. When they are divided into units based on their respective specific dates or specific date and hour, the low-resolution original two-dimensional meteorological data of the station on that date and the low-resolution original two-dimensional AOD data of the station on that date and hour can be obtained.

[0069] 2) Interpolation calculations were performed on the low-resolution raw two-dimensional meteorological data of the station, as shown in Figure 13, to obtain a high-resolution two-dimensional meteorological data matrix of the "points" of the monitoring station.

[0070] 3) Perform fill and interpolation calculations on the low-resolution raw two-dimensional AOD data of the monitoring station to obtain a high-resolution two-dimensional matrix of AOD data for the monitoring station "point". Here, the two-dimensional matrix of AOD data refers to the data for a specific date.

[0071] 4) Obtain the input and output data for the machine learning training phase and complete the machine learning training phase.

[0072] Further, in step two, process 2), the low-resolution raw two-dimensional meteorological data of the station are interpolated to obtain a high-resolution two-dimensional meteorological data matrix of the monitoring station's "points," as shown in Figures 12 and 13. Specifically:

[0073] a. Kriging interpolation calculation: Kriging interpolation calculation is performed on low-resolution raw two-dimensional meteorological data; the Kriging interpolation calculation is to transform the low-resolution raw two-dimensional meteorological data into high-resolution M-row M-column two-dimensional meteorological data;

[0074] b. Subdivide the latitude and longitude of each grid cell in the high-resolution M-row M-column two-dimensional meteorological data to obtain a high-resolution M-row M-column two-dimensional meteorological data matrix after subdivision of latitude and longitude.

[0075] c. Search for the high-precision original meteorological data corresponding to the latitude and longitude of the station in the high-resolution M-row M-column meteorological data two-dimensional matrix after subdividing latitude and longitude, and obtain the high-resolution meteorological data two-dimensional matrix of the "points" of the monitoring station.

[0076] Further, step two, process 3), involves filling in the low-resolution original two-dimensional AOD data of the site, as shown in Figure 12. The specific steps are as follows:

[0077] ① Divide the low-resolution original AOD two-dimensional matrix of the site into N rows and N columns of raster, and the number of rows and columns of the N rows and N columns of raster corresponds to the number of rows and columns of the low-resolution original AOD two-dimensional matrix;

[0078] ② Generate an N-row N-column counting AOD all-zero two-dimensional matrix for counting and an N-row N-column summing AOD all-zero two-dimensional matrix for summing; the counting AOD all-zero two-dimensional matrix is ​​used to count the number of times valid AOD data appears in each grid at different times; the summing AOD all-zero two-dimensional matrix is ​​used to count the aerosol data of each grid at different times: the different times refer to different hours of the same date.

[0079] ③ Count the number of valid AOD data appearing in each grid cell of the counted AOD all-zero two-dimensional matrix at different times, and generate the counted AOD two-dimensional matrix; and count the aerosol data of each grid cell of the total AOD all-zero two-dimensional matrix at each time time, and generate the total AOD two-dimensional matrix.

[0080] ④ Divide the value of each grid cell in the total AOD two-dimensional matrix by the value of each grid cell in the count AOD two-dimensional matrix to generate a low-resolution AOD mean two-dimensional matrix.

[0081] 4) Further, in step two, process 3), the low-resolution original two-dimensional AOD data of the station is filled and interpolated to obtain a high-resolution two-dimensional matrix of AOD data for the monitoring station "point," as shown in Figures 12 and 13. The specific steps are as follows:

[0082] ① Interpolate the low-resolution AOD mean two-dimensional matrix to obtain a high-resolution M-row M-column AOD mean two-dimensional matrix. Specifically, after interpolation, the matrix is ​​expanded from N rows and N columns to M rows and M columns, where M = N * K and K is the expansion factor.

[0083] ③ Subdivide the latitude and longitude of each grid cell in the high-resolution M-row M-column AOD mean two-dimensional matrix to obtain a high-resolution M-row M-column AOD mean two-dimensional matrix after subdividing latitude and longitude.

[0084] ④ After separating latitude and longitude, search the high-resolution M-row M-column AOD mean two-dimensional matrix to obtain the AOD mean value corresponding to the latitude and longitude of the monitoring station, and obtain the high-resolution AOD data two-dimensional matrix of the monitoring station "point".

[0085] Further, in step two, process 4), the input and output data for the machine learning training phase are obtained, and the machine learning training phase is completed. The specific steps are as follows:

[0086] a. The date on which monitoring data was obtained from the monitoring site; this date must be the day on which monitoring data is available.

[0087] b. Calculate the average values ​​of air pressure, humidity, wind speed, temperature, and wind direction at the monitoring stations on that date based on meteorological data;

[0088] c. Calculate the average value of aerosol data from the monitoring stations on that date based on the aerosol data;

[0089] d. Obtain the average PM10 value actually measured at the monitoring station on that day;

[0090] e. Convert the above data into a format that machine learning can recognize;

[0091] f. Use the high-resolution meteorological data two-dimensional matrix of the monitoring station “point” and the high-resolution AOD data two-dimensional matrix of the monitoring station “point” as the input data for machine learning, and use the PM10 mean of the monitoring station as the output value for machine learning to achieve model training.

[0092] Furthermore, step three involves inversion based on the interpolated and filled data, and the specific process is as follows:

[0093] a. Enter the target physical region shapefile and date for the PM10 data to be retrieved; the target physical region includes a province, city, district, or county;

[0094] b. Obtain low-resolution raw three-dimensional meteorological data and raw three-dimensional AOD data of the target physical area on the date, and then convert the raw three-dimensional meteorological data and raw three-dimensional AOD data into a low-resolution raw meteorological data two-dimensional matrix and a low-resolution raw AOD data two-dimensional matrix of the station on the date and hour, respectively.

[0095] c. Perform interpolation calculations on the low-resolution original meteorological data two-dimensional matrix of the target physical area on the date to obtain multiple high-resolution M-row M-column meteorological data two-dimensional matrices of the target physical area on the date; perform filling and interpolation calculations on the low-resolution original AOD data two-dimensional matrix to obtain a high-resolution M-row M-column AOD mean two-dimensional matrix of the target physical area on the date.

[0096] d. As shown in Figure 14, multiple high-resolution M-row M-column meteorological data two-dimensional matrices for the date and the target physical area, as well as a high-resolution M-row M-column AOD mean two-dimensional matrix, are dimensionalized. After dimensionality reduction, multiple one-dimensional matrices of multi-row single-column high-resolution meteorological data and a one-dimensional matrix of multi-row single-column high-resolution AOD mean data are obtained. The multiple one-dimensional matrices of multi-row single-column high-resolution meteorological data include one-dimensional matrices of multi-row single-column high-resolution meteorological data, one-dimensional matrices of relative humidity, one-dimensional matrices of wind speed, one-dimensional matrices of temperature, and one-dimensional matrices of wind direction.

[0097] e. Use the one-dimensional matrix of multiple multi-row, single-column high-resolution meteorological data and the one-dimensional matrix of high-resolution AOD mean data as input data for the machine learning application stage.

[0098] f. The machine model outputs an inverted PM10 multi-row, single-column, one-dimensional matrix.

[0099] Furthermore, the specific process for generating the final inversion result in step three is as follows:

[0100] 1) As shown in Figures 15 and 16, the multi-row, single-column one-dimensional matrix of inverted PM10 output by the machine model is converted into a multi-row, multi-column two-dimensional matrix of inverted PM10.

[0101] 2) Input the multi-row, multi-column two-dimensional matrix of inverted PM10 and the target physical region shapefile into the gdal clipping method file;

[0102] 3) The gdal cropping method file outputs the PM10 graphic file of the target physical region.

[0103] The process 1) converts the multi-row, single-column, one-dimensional matrix of inverted PM10 output by the machine model into a multi-row, multi-column two-dimensional matrix of inverted PM10. The specific process is as follows:

[0104] a. Based on the definition of the number of rows and columns before dimensionality reduction, convert the one-dimensional array output by the machine model into the number of rows and columns of a two-dimensional array;

[0105] b. Based on the defined number of rows and columns, convert the single-column, multi-row, one-dimensional array output by the machine model into a multi-row, multi-column, two-dimensional array.

[0106] It should be emphasized that the above specific embodiments are merely explanations of the present invention and are not intended to limit the present invention. After reading this specification, those skilled in the art can make modifications to the above embodiments without contributing any inventive step, but as long as they are within the scope of the claims of the present invention, they are protected by patent law.

Claims

1. A PM10 inversion system based on multi-source data fusion, characterized in that, The system includes a multi-source data fusion unit for training machine models at monitoring sites and a multi-source data fusion unit for retrieving PM10 from target physical areas; the multi-source data fusion unit for retrieving PM10 from target physical areas uses the trained machine model to retrieve PM10 data from the target physical area. The multi-source data fusion unit used for training machine models at monitoring stations includes a monitoring station latitude and longitude module, a satellite remote sensing data module, a monitoring station raw data module, a monitoring station raw data filling and interpolation module, a two-dimensional data to "point" data conversion module, a monitoring station PM10 detection module, and a machine model training module. The original data filling and interpolation module for the monitoring station includes a module for reading the original meteorological data of the monitoring station, a module for reading the original AOD data of the monitoring station, an interpolation algorithm module, a mean algorithm module, a high-resolution two-dimensional meteorological data module for the monitoring station, a low-resolution AOD data mean module, and a high-resolution two-dimensional AOD data module for the monitoring station. The low-resolution AOD data mean module performs mean processing on the original AOD data according to the mean algorithm to obtain the low-resolution AOD data mean. The high-resolution two-dimensional AOD data module interpolates the low-resolution AOD data mean to obtain the high-resolution AOD two-dimensional matrix of the monitoring station. The high-resolution two-dimensional meteorological data module interpolates the original meteorological data according to the interpolation algorithm to obtain the high-resolution two-dimensional meteorological data of the monitoring station. The low-resolution AOD data averaging module performs averaging on the original AOD data according to the averaging algorithm: ① Divide the low-resolution original AOD two-dimensional matrix of the site into N rows and N columns of raster, and the number of rows and columns of the N rows and N columns of raster corresponds to the number of rows and columns of the low-resolution original AOD two-dimensional matrix; ② Generate an N-row N-column counting AOD all-zero two-dimensional matrix for counting and an N-row N-column summing AOD all-zero two-dimensional matrix for summing; the counting AOD all-zero two-dimensional matrix is ​​used to count the number of times valid AOD data appears in each grid at different times; The total AOD zero two-dimensional matrix is ​​used to statistically analyze the aerosol data of each grid at different times: the different times refer to different hours of the same date; ③ Count the number of valid AOD data appearing in each grid cell of the all-zero AOD counting matrix at different times, and generate the AOD counting matrix; The aerosol data for each grid cell in the total AOD all-zero two-dimensional matrix are calculated for each time period, and the total AOD two-dimensional matrix is ​​generated. ④ Divide the value of each cell in the total AOD two-dimensional matrix by the value of each cell in the count AOD two-dimensional matrix to generate a low-resolution AOD mean two-dimensional matrix; The high-resolution meteorological two-dimensional data module and the high-resolution AOD two-dimensional data module of the monitoring station perform Kriging interpolation on the mean data of the original meteorological data and the original AOD data of the monitoring station, respectively. The multi-source data fusion unit for PM10 inversion in the target physical area includes a target physical area shapefile module, a target physical area PM10 graphic generation module, and a target physical area raw data module. The target physical area raw data module obtains the target physical area's three-dimensional raw meteorological data and three-dimensional raw AOD data from the satellite remote sensing data module based on the target physical area shapefile, and converts them into two-dimensional raw meteorological data and two-dimensional raw AOD aerosol data. The target physical area PM10 graphic generation module includes a gdal clipping method module, a PM10 two-dimensional array inversion module, a clipping module, and a target physical area PM10 inversion graphic file generation module. The clipping module obtains data from the gdal clipping method module and the PM10 two-dimensional array inversion module to generate the target physical area PM10 inversion graphic file.

2. The PM10 inversion system based on multi-source data fusion according to claim 1, characterized in that: The monitoring station's raw data module acquires the raw meteorological data and raw AOD aerosol data of the monitoring station from the satellite remote sensing data module based on the latitude and longitude data of the monitoring station; the monitoring station's raw data filling and interpolation module interpolates and / or fills the raw meteorological data of the monitoring station, transforming the low-resolution raw data into high-resolution two-dimensional data; the two-dimensional data to "point" data conversion module subdivides the high-resolution two-dimensional data of the monitoring station into latitude and longitude, and searches for high-resolution meteorological data and high-resolution AOD data of the "points" of the monitoring station based on the latitude and longitude of the monitoring station; the monitoring station's PM10 detection module detects the PM10 data of the "points" of the monitoring station. The machine model training module uses high-resolution meteorological data and high-resolution AOD data of the monitoring stations as input values ​​for machine model training, and PM10 data of the monitoring stations as output values ​​for machine model training. The raw data from the monitoring stations includes raw meteorological data and raw AOD data; the high-resolution two-dimensional data refers to high-resolution meteorological two-dimensional data and high-resolution AOD two-dimensional data from the monitoring stations.

3. The PM10 inversion system based on multi-source data fusion according to claim 2, characterized in that: The monitoring station raw data module includes a satellite remote sensing data acquisition module, a monitoring station latitude and longitude acquisition module, a monitoring station raw three-dimensional meteorological data module, a monitoring station raw three-dimensional AOD data module, a monitoring station raw three-dimensional meteorological data to two-dimensional conversion module, and a monitoring station raw three-dimensional AOD data to two-dimensional conversion module. The original 3D meteorological data module and the original 3D AOD data module of the monitoring station obtain the original 3D meteorological data and the original 3D AOD data of the monitoring station from the satellite remote sensing data module based on the latitude and longitude data of the monitoring station, respectively; the format of the 3D meteorological data is date, longitude, and latitude; The three-dimensional AOD data format is date, hour, longitude, and latitude; The original three-dimensional meteorological data to two-dimensional conversion module of the monitoring station converts the three-dimensional meteorological data into two-dimensional meteorological data according to the date; the original three-dimensional AOD data to two-dimensional conversion module of the monitoring station converts the three-dimensional AOD data into two-dimensional AOD data according to the date and hour.

4. The PM10 inversion system based on multi-source data fusion according to claim 2, characterized in that: The module for converting two-dimensional data into "point" data includes a module for subdividing high-resolution meteorological data of monitoring stations by latitude and longitude, a module for subdividing high-resolution AOD data of monitoring stations by latitude and longitude, a module for high-resolution meteorological data of "points" of monitoring stations, and a module for high-resolution AOD data of "points" of monitoring stations. The high-resolution meteorological data module for latitude and longitude and the high-resolution AOD data module for latitude and longitude of monitoring stations search for the data corresponding to the latitude and longitude of the monitoring stations in the module for subdividing high-resolution meteorological data of monitoring stations by latitude and longitude and the module for subdividing high-resolution AOD data of monitoring stations, respectively, to obtain the high-resolution meteorological data of "points" of monitoring stations and the high-resolution AOD data of "points" of monitoring stations.

5. The PM10 inversion system based on multi-source data fusion according to claim 1, characterized in that: The mean algorithm module includes a module for reading AOD two-dimensional raw data, a low-resolution counting matrix module, a low-resolution total matrix module, and a low-resolution AOD data mean module. The low-resolution counting matrix module divides the AOD two-dimensional raw data into raster data and counts the number of valid AOD data appearing in each raster of latitude and longitude according to the date and hour. The low-resolution total matrix module divides the original AOD two-dimensional data into raster data and counts the data values ​​of valid AODs in each raster at each latitude and longitude according to the date and hour. The low-resolution AOD data mean module divides the value of each raster in the low-resolution counting matrix by the value of each raster in the low-resolution total matrix to obtain the value of each raster in the low-resolution mean two-dimensional matrix, thereby obtaining the low-resolution AOD mean two-dimensional matrix.

6. The PM10 inversion system based on multi-source data fusion according to claim 1, characterized in that: The multi-source data fusion unit for PM10 inversion in the target physical area includes a satellite remote sensing data module, a target physical area raw data filling and interpolation module, a target physical area high-resolution data 2D to 1D conversion module, and a machine model PM10 inversion module. The target physical area raw data filling and interpolation module fills and interpolates the target physical area raw data, transforming low-resolution raw 2D meteorological data and low-resolution raw 2D AOD data into high-resolution 2D meteorological data and high-resolution 2D AOD data, respectively. The target physical area high-resolution data 2D to 1D conversion module converts the high-resolution 2D meteorological data and high-resolution 2D AOD data into high-resolution 1D meteorological data and high-resolution 1D AOD data according to the date. The machine model PM10 inversion module obtains high-resolution one-dimensional meteorological data and high-resolution one-dimensional AOD data from the target physical area high-resolution data two-dimensional to one-dimensional module as input data for the machine model application stage. The machine model obtains an inverted one-dimensional array of PM10 and converts the inverted one-dimensional array of PM10 into an inverted two-dimensional array of PM10. The target physical area PM10 graphics generation module generates PM10 graphics of the target physical area based on the inverted two-dimensional array of PM10 data and the data in the cropping method file.

7. The PM10 inversion system based on multi-source data fusion according to claim 6, characterized in that: The target physical area high-resolution data 2D to 1D conversion module includes: a target physical area high-resolution 2D meteorological data module for the date, a target physical area high-resolution 2D AOD data module for the date, a high-resolution 1D meteorological data module, and a high-resolution 1D AOD data module; the high-resolution 1D meteorological data module converts the target physical area high-resolution 2D meteorological data for the date into 1D data; the high-resolution 1D AOD data module converts the target physical area high-resolution 2D AOD data for the date into 1D data.

8. The PM10 inversion system based on multi-source data fusion according to claim 6, characterized in that: The machine model PM10 inversion module includes a machine model, a PM10 one-dimensional array inversion submodule, and a PM10 two-dimensional array inversion submodule. The machine model PM10 inversion module takes the high-resolution one-dimensional meteorological data and high-resolution one-dimensional AOD data as input data for the machine model and outputs the PM10 one-dimensional array. The PM10 two-dimensional array inversion submodule converts the PM10 one-dimensional array into a two-dimensional array.

Images