Monitoring methods, model training methods, equipment, and media for the distribution of contaminants.

Abstract

The present invention discloses a monitoring method, a model training method, a device and a medium for pollution distribution, including: taking the aerosol optical thickness data, near-infrared index data, reanalysis meteorological data and daily average resolution surface data obtained in the target training study area as a data sample, and inputting them into a regression tree model to obtain the output predicted salt density value; calculating the error value according to the measured salt density value and the predicted salt density value of the target training study area, and adjusting the parameters of the regression tree model according to the error value, and repeating this process until the preset accuracy requirements are verified. Then, based on the trained regression tree model, the equivalent salt density distribution of the target study area is predicted, and a pollution spatial distribution map is drawn, so as to grasp the pollution distribution of the target study area. It can be seen that the present invention can quickly and efficiently obtain high-resolution pollution conditions in the study area, reduce the consumption of manpower, material resources and funds, and provide reference and indicative significance for the relevant part of the pollution cleaning.

Description

Technical Field

[0001] The invention relates to the technical field of atmospheric remote sensing and ecological remote sensing, and in particular to a monitoring method, a model training method, equipment and a medium for pollution distribution. Background Art

[0002] At present, there are obvious pollution problems in the substation lines and transmission lines. When the pollution of the insulators accumulates to a certain extent, pollution flashover will occur, causing regional power outages or serious economic losses. The traditional method of monitoring the accumulation of insulator pollution is manual measurement, which mainly involves manually cleaning the dirt on the insulators and estimating the pollution of all insulators in the study area based on experience. This method consumes a lot of manpower and material resources, and the results are affected by human subjective factors. It cannot accurately and effectively reflect the distribution of pollution in the study area. Summary of the invention

[0003] Based on this, it is necessary to provide monitoring methods, model training methods, equipment and media for the pollution distribution in order to solve the problem of not being able to accurately and effectively reflect the pollution distribution in the study area.

[0004] A regression tree model training method, the method comprising:

[0005] In a plurality of preset training study areas, obtaining aerosol optical thickness data and infrared data in a target training study area, and calculating near infrared index data according to the infrared data; wherein the target training study area is any one of the plurality of training study areas;

[0006] Obtain reanalysis meteorological data within the target training study area; wherein the reanalysis meteorological data includes monthly average temperature, monthly maximum temperature, precipitation, and solar downward radiation;

[0007] Obtain daily average resolution ground station data within the target training study area, and upscale each daily average resolution ground station data to daily average resolution surface data covering the target training study area through a transfer learning algorithm; wherein the daily average resolution ground station data includes daily average PM2.5 data, daily average PM10 data, daily average SO2 data, daily average NO2 data, and daily average NO3 data covering the ground station;

[0008] The aerosol optical thickness data, near infrared index data, reanalysis meteorological data and daily average resolution surface data in the target training study area are taken as a data sample, and a data sample is input into the regression tree model to obtain the output predicted salt density value; wherein, the data sample input each time is different;

[0009] An error value is calculated based on the measured salt density value and the predicted salt density value of the target training study area, and the parameters of the regression tree model are adjusted according to the error value. The step of inputting a data sample into the regression tree model to obtain the output predicted salt density value and subsequent steps are returned to execute until the preset accuracy requirements are verified and the trained regression tree model is obtained.

[0010] In one embodiment, the step of obtaining aerosol optical thickness data and infrared data in a target training study area, and calculating near infrared index data based on the infrared data, includes:

[0011] Acquire the daily average resolution aerosol optical thickness data of multiple days in the target training study area, and calculate the monthly maximum value of each pixel value of the daily average resolution aerosol optical thickness data of the preset days by the maximum value synthesis method to obtain the monthly average resolution aerosol optical thickness data, and use the daily average resolution aerosol optical thickness data and the monthly average resolution aerosol optical thickness data as the aerosol optical thickness data;

[0012] The reflectance value of the near-infrared band and the reflectance value of the red light band in the target training study area are obtained, the normalized vegetation index is calculated according to the reflectance value of the near-infrared band and the reflectance value of the red light band, and the product of the normalized vegetation index and the near-infrared band reflectance is calculated, and the obtained product value is used as the near-infrared index data.

[0013] In one embodiment, after obtaining aerosol optical thickness data and infrared data in the target training study area and calculating near infrared index data based on the infrared data, the method further includes:

[0014] The near-infrared index data are resampled by the nearest neighbor sampling method so that the spatial resolution of the resampled aerosol optical thickness data and the near-infrared index data remain consistent.

[0015] In one embodiment, after obtaining the reanalysis meteorological data in the target training study area, the method further includes:

[0016] The reanalysis meteorological data are cropped, and the cropped reanalysis meteorological data are resampled by Kriging interpolation method, so that the spatial resolution of the resampled reanalysis meteorological data is consistent with the resampled aerosol optical thickness data and near infrared index data.

[0017] In one embodiment, the method further comprises:

[0018] The daily average resolution surface data is synthesized into monthly average resolution surface data by a maximum value synthesis method, and the monthly average resolution surface data is added to the data sample.

[0019] In one embodiment, the aerosol optical thickness data, near infrared index data, reanalysis meteorological data and daily average resolution surface data in the target training study area are taken as a data sample, and a data sample is input into the regression tree model to obtain the output predicted salt density value, including:

[0020] Creating multiple rules based on data samples of the target training study area; wherein each rule corresponds to a type of input data and is associated with a related multiple linear regression model and judgment condition, and all multiple linear regression models and judgment conditions constitute the regression tree model;

[0021] At the terminal node of the regression tree model, the regression values ​​output by all the multivariate linear regression models are summed and averaged, and the obtained average value is used as the predicted salt density value.

[0022] In one embodiment, the method further comprises:

[0023] The data samples are divided into a training set and a validation set by the leave-one-out method, and the correlation coefficient between the measured salt density value and the predicted salt density value of the data samples in the validation set is calculated. When the correlation coefficient is greater than the preset coefficient threshold, it is determined that the validation passes the accuracy requirement.

[0024] A method for monitoring pollution distribution, the method comprising:

[0025] Obtain aerosol optical thickness data, near infrared index data, reanalysis meteorological data and daily average resolution surface data of the target study area in a preset year, and input them into a regression tree model to obtain the output equivalent salt density distribution of the target study area; wherein the regression tree model is trained by the above-mentioned regression tree model training method;

[0026] A pollution space distribution map is drawn based on the equal salt density distribution of the target study area; wherein the pollution space distribution map is used to indicate the pollution distribution situation of the target study area.

[0027] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the processor executes the steps of a regression tree model training method and a pollution distribution monitoring method.

[0028] A device for monitoring pollution distribution includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes the steps of a regression tree model training method and a pollution distribution monitoring method.

[0029] The present invention provides a monitoring method, a model training method, a device and a medium for pollution distribution, including taking aerosol optical thickness data, near infrared index data, reanalysis meteorological data and daily average resolution surface data obtained in the target training study area as a data sample, inputting a data sample into a regression tree model, and obtaining an output predicted salt density value; calculating an error value according to the measured salt density value and the predicted salt density value of the target training study area, and adjusting the parameters of the regression tree model according to the error value, and repeating this process until the verification passes the preset accuracy requirements, and obtaining a trained regression tree model. Then, based on the trained regression tree model, the equivalent salt density distribution of the target study area is predicted, and a pollution spatial distribution map is drawn, so as to grasp the pollution distribution of the target study area. It can be seen that the present invention can quickly and efficiently obtain high-resolution pollution conditions in the study area, reduce the consumption of manpower, material resources and funds, and provide reference and indicative significance for the relevant part of the pollution cleaning. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0031] in:

[0032] Figure 1 A schematic diagram of a process of a regression tree model training method in one embodiment;

[0033] Figure 2 A schematic diagram of a flow chart of a method for monitoring pollution distribution in one embodiment;

[0034] Figure 3 FIG. 4 is a structural block diagram of a terminal device in an embodiment. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] like Figure 1 As shown, Figure 1 : is a flow chart of a regression tree model training method in one embodiment. The steps provided by the regression tree model training method in this embodiment include:

[0037] Step 102, in a plurality of preset training study areas, obtain aerosol optical thickness data and infrared data in a target training study area, and calculate near infrared index data based on the infrared data.

[0038] Among them, the preset multiple training study areas are used to train the Cubist regression tree model, and the target training study area is any one of the multiple training study areas.

[0039] Aerosol optical depth (AOD) refers to the vertical integral of the aerosol extinction coefficient in a vertical column of clear sky atmosphere. It is a dimensionless quantity that describes the light attenuation effect of aerosols and can describe the degree of regional atmospheric pollution to a certain extent. The smaller the AOD value, the cleaner the atmosphere; otherwise, the more turbid it is.

[0040] Specifically, under the premise of scientific Internet access, log in to the Google Data Cloud Platform (GEE), and download the daily average resolution AOD data of multiple days in the target training study area in the developer interface, and download it in TIFF format. And use the maximum value synthesis method to calculate the monthly maximum value of each pixel of the daily average resolution AOD data of the preset number of days (for example, the daily average resolution AOD data of eight days resolution) to obtain the monthly average resolution AOD data, thereby avoiding data errors. Finally, the daily average resolution AOD data and the monthly average resolution AOD data are used as the AOD data required for subsequent training.

[0041] Then download the MOD09 data, including the reflectance value NIR of the near infrared band and the reflectance value R of the red light band in the target training study area. And calculate the normalized vegetation index NDVI and near infrared index data NIRv according to the following formula:

[0042]

[0043] NIRv=NDVI*NIR

[0044] The median value of each month of MOD09 data was used to calculate the NIRv data. When calculating NDVI, all NDVI values ​​were subtracted by 0.05 to avoid the influence of background noise of bare soil.

[0045] The value range of NIRv data is generally between 0 and 0.2. The nearest neighbor sampling method is used to resample the 500m resolution NIRv data to 0.05 degrees, so that the spatial resolution of the resampled AOD data and NIRv data is consistent. At the same time, the shapefile file of the study area is used to crop the required AOD data and NIRv data.

[0046] Step 104, obtaining reanalysis meteorological data within the target training study area.

[0047] The reanalysis meteorological data include monthly mean temperature, monthly maximum temperature, precipitation, and solar downward radiation. The downloaded dataset is in nc format. After defining the geographic coordinate system, it is converted to TIFF format for subsequent processing. All reanalysis meteorological data are clipped using the shipfile data of the study area.

[0048] Furthermore, in order to maintain a uniform spatial resolution for all data input into the Cubist model, the cropped reanalysis meteorological data are resampled using the Kriging interpolation method, so that the 0.5-degree reanalysis meteorological data are converted to a spatial resolution of 0.05 degrees, so that the spatial resolution of the resampled reanalysis meteorological data is consistent with the resampled AOD data and NIRv data.

[0049] Among them, the principle of Kriging interpolation is:

[0050] (1) Construct semivariogram training samples.

[0051] (2) Construct a scatter plot based on the semivariogram function.

[0052] (3) Fit the empirical semivariogram based on the Circular function.

[0053] (4) Obtain the values ​​of Nugget, Sill and Range based on the fitted semivariogram function.

[0054] (5) Predicting the value of an unknown point: performing weighted summation on the values ​​of measurement points within a specific range of the unknown point or a specific number of adjacent measurement points to obtain the value of the unknown point and thus achieve prediction of the unknown point.

[0055] Step 106, obtaining daily average resolution ground station data within the target training study area, and upscaling each daily average resolution ground station data to daily average resolution surface data covering the target training study area through a transfer learning algorithm.

[0056] Among them, the daily average resolution ground station data includes daily average PM2.5 data, daily average PM10 data, daily average SO2 data, daily average NO2 data and daily average NO3 data covering the ground stations.

[0057] The principle of transfer learning is:

[0058] (1) Select a source model. Select a relevant prediction source model with rich data. There is a specific relationship between the input data, output data, and mapping between the input data and output data of the source model and the target model.

[0059] (2) Develop source model. Develop the model based on the downloaded daily average PM2.5 data, daily average PM10 data, daily average SO2 data, daily average NO2 data and daily average NO3 data.

[0060] (3) Reuse model: The source model serves as the starting point for learning the target model.

[0061] (4) Model tuning. The model can be selectively fine-tuned on input-output pairs in the target dataset to generate upscaled daily-average resolution surface data.

[0062] Furthermore, the daily average resolution surface data is synthesized into monthly average resolution surface data through the maximum value synthesis method, thereby avoiding data errors.

[0063] Step 108, taking the aerosol optical thickness data, near infrared index data, reanalysis meteorological data and daily average resolution surface data in the target training study area as a data sample.

[0064] That is, the AOD data, NIRv data, monthly average temperature, monthly maximum temperature, precipitation, solar downward radiation, upscaled daily average PM10 data, daily average S02 data, daily average NO2 data and daily average NO3 data in the target training study area are taken as a data sample.

[0065] Step 110: input a data sample into the regression tree model to obtain an output predicted salt density value.

[0066] The data samples input each time are different, that is, data from different training study areas are obtained each time and input into the Cubist model for training.

[0067] The output of the Cubist model after machine learning is multiple regression rules. Each regression rule corresponds to a type of input data and is associated with a related multivariate linear regression model and judgment condition. All multivariate linear regression models and judgment conditions constitute the Cubist model, which is expressed as:

[0068] if: X1 in range[α1], X2 in range[α2],...X n in range[α n ]

[0069] Then: Y=a+a1×X1+a2×X2+…+a n ×X n

[0070] Among them, X1, X2...Xn are the independent variables in the data sample, Y is the dependent variable, α1, α2...α nis the judgment condition for each independent variable; a is the constant term of the Cubist model; a1, a2...a n is the regression coefficient.

[0071] Finally, in the terminal node of the Cubist model, the regression values ​​output by all multivariate linear regression models are summed and averaged, and the average value is used as the predicted salt density value.

[0072] Step 112, calculate the error value based on the measured salt density value and the predicted salt density value of the target training study area, and adjust the parameters of the regression tree model according to the error value, return to execute step 110 and subsequent steps until the verification passes the preset accuracy requirements, execute step 114, and obtain the trained regression tree model.

[0073] Specifically, the data samples are divided into training set and validation set by the leave-one-out method. The leave-one-out method is to leave only one data sample as the validation set each time, and the other samples as the training set. If there are k samples, the validation set is 1 and the training set is k-1. At most, train k times and validate k times.

[0074] During training, the error value is calculated based on the measured salt density value and the predicted salt density value of the data samples in the training set, and the parameters of the Cubist model are adjusted according to the error value. During verification, the correlation coefficient between the measured salt density value and the predicted salt density value of the data samples in the verification set is calculated. When the correlation coefficient is greater than the preset coefficient threshold, it is determined that the verification passes the accuracy requirement.

[0075] The above model training method uses the aerosol optical thickness data, near-infrared index data, reanalysis meteorological data and daily average resolution surface data obtained in the target training study area as a data sample, inputs a data sample into the regression tree model, and obtains the output predicted salt density value; calculates the error value based on the measured salt density value and the predicted salt density value of the target training study area, and adjusts the parameters of the regression tree model based on the error value, and repeats this process until the verification passes the preset accuracy requirements, and obtains the trained regression tree model. The regression tree model can predict the salt density value of the study area.

[0076] like Figure 2 As shown, Figure 2 FIG. 1 is a flow chart of a method for monitoring pollution distribution in an embodiment. The method for monitoring pollution distribution in this embodiment includes the following steps:

[0077] Step 202, obtain aerosol optical thickness data, near infrared index data, reanalysis meteorological data and daily average resolution surface data of the target study area for a preset year, and input them into the regression tree model to obtain the output of the equivalent salt density distribution of the target study area.

[0078] The Cubist model is trained by the above-mentioned Cubist model training method.

[0079] First, determine the preset year for which the pollution spatial distribution map needs to be drawn, input the year, and retrieve the corresponding aerosol optical thickness data, near-infrared index data, reanalysis meteorological data, and daily average resolution surface data, and input them into the trained Cubist model to obtain the output of the equivalent salt density distribution of the target study area.

[0080] Step 204, draw a pollution spatial distribution map based on the equivalent salt density distribution of the fundamental target study area.

[0081] Among them, the pollution spatial distribution map is used to indicate the pollution distribution in the target study area.

[0082] Then use Matlab's function package m_map to draw a pollution spatial distribution map based on the equivalent salt density distribution of the fundamental target study area.

[0083] It can be seen that the present invention can quickly and efficiently obtain high-resolution pollution conditions in the study area, reduce the consumption of manpower, material resources and funds, and provide reference and indicative significance for the pollution cleaning of relevant parts.

[0084] Figure 3 FIG. 2 shows an internal structure diagram of a terminal device in an embodiment. Figure 3 As shown, the terminal device includes a processor, a memory and a network interface connected via a system bus. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the terminal device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor may implement a method for monitoring the distribution of contamination and model training. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor may implement a method for monitoring the distribution of contamination and model training. Those skilled in the art will appreciate that Figure 3 The structure shown in the figure is only a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the terminal device to which the scheme of the present application is applied. The specific terminal device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0085] A terminal device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program: in a plurality of preset training study areas, obtaining aerosol optical thickness data and infrared data in a target training study area, and calculating near infrared index data based on the infrared data; obtaining reanalysis meteorological data in the target training study area; obtaining daily average resolution ground station data in the target training study area, and upscaling each daily average resolution ground station data to daily average resolution surface data covering the target training study area through a transfer learning algorithm; taking the aerosol optical thickness data, near infrared index data, reanalysis meteorological data and daily average resolution surface data in the target training study area as a data sample, inputting a data sample into a regression tree model, and obtaining an output predicted salt density value; calculating an error value based on the measured salt density value and the predicted salt density value of the target training study area, adjusting the parameters of the regression tree model based on the error value, returning to execute the step of inputting a data sample into the regression tree model, and obtaining the output predicted salt density value and subsequent steps, until the preset accuracy requirement is verified and the trained regression tree model is obtained.

[0086] And execute: obtain aerosol optical thickness data, near-infrared index data, reanalysis meteorological data and daily average resolution surface data of the target study area in the preset year, and input them into the regression tree model to obtain the output of the equivalent salt density distribution of the target study area; draw the pollution spatial distribution map based on the equivalent salt density distribution of the target study area.

[0087] A computer-readable storage medium stores a computer program, which implements the following steps when executed by a processor: in a plurality of preset training study areas, obtain aerosol optical thickness data and infrared data in a target training study area, and calculate near infrared index data based on the infrared data; obtain reanalysis meteorological data in the target training study area; obtain daily average resolution ground station data in the target training study area, and upscale each daily average resolution ground station data to daily average resolution surface data covering the target training study area through a transfer learning algorithm; take the aerosol optical thickness data, near infrared index data, reanalysis meteorological data and daily average resolution surface data in the target training study area as a data sample, input a data sample into a regression tree model, and obtain an output predicted salt density value; calculate an error value based on the measured salt density value and the predicted salt density value of the target training study area, adjust the parameters of the regression tree model based on the error value, return to execute the step of inputting a data sample into the regression tree model, and obtain the output predicted salt density value and subsequent steps, until the preset accuracy requirement is verified and the trained regression tree model is obtained.

[0088] And execute: obtain aerosol optical thickness data, near-infrared index data, reanalysis meteorological data and daily average resolution surface data of the target study area in the preset year, and input them into the regression tree model to obtain the output of the equivalent salt density distribution of the target study area; draw the pollution spatial distribution map based on the equivalent salt density distribution of the target study area.

[0089] It should be noted that the above-mentioned pollution distribution monitoring method, model training method, equipment and medium belong to a general inventive concept, and the contents of the pollution distribution monitoring method, model training method, equipment and medium embodiments are applicable to each other.

[0090] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0091] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0092] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A regression tree model training method, characterized in that: The method comprises: In a plurality of preset training study areas, obtaining aerosol optical thickness data and infrared data in a target training study area, and calculating near infrared index data according to the infrared data; wherein the target training study area is any one of the plurality of training study areas; Obtain reanalysis meteorological data within the target training study area; wherein the reanalysis meteorological data includes monthly average temperature, monthly maximum temperature, precipitation, and solar downward radiation; Obtain daily average resolution ground station data within the target training study area, and upscale each daily average resolution ground station data to daily average resolution surface data covering the target training study area through a transfer learning algorithm; wherein the daily average resolution ground station data includes daily average PM2.5 data, daily average PM10 data, daily average SO2 data, daily average NO2 data, and daily average NO3 data covering the ground station; The aerosol optical thickness data, near infrared index data, reanalysis meteorological data and daily average resolution surface data in the target training study area are taken as a data sample, and a data sample is input into the regression tree model to obtain the output predicted salt density value; wherein, the data sample input each time is different; An error value is calculated based on the measured salt density value and the predicted salt density value of the target training study area, and the parameters of the regression tree model are adjusted according to the error value. The step of inputting a data sample into the regression tree model to obtain the output predicted salt density value and subsequent steps are returned to execute until the preset accuracy requirements are verified and the trained regression tree model is obtained.

2. The method according to claim 1, characterized in that: The step of obtaining aerosol optical thickness data and infrared data in the target training study area, and calculating near infrared index data based on the infrared data, includes: Acquire the daily average resolution aerosol optical thickness data of multiple days in the target training study area, and calculate the monthly maximum value of each pixel value of the daily average resolution aerosol optical thickness data of the preset days by the maximum value synthesis method to obtain the monthly average resolution aerosol optical thickness data, and use the daily average resolution aerosol optical thickness data and the monthly average resolution aerosol optical thickness data as the aerosol optical thickness data; The reflectance value of the near-infrared band and the reflectance value of the red light band in the target training study area are obtained, the normalized vegetation index is calculated according to the reflectance value of the near-infrared band and the reflectance value of the red light band, and the product of the normalized vegetation index and the near-infrared band reflectance is calculated, and the obtained product value is used as the near-infrared index data.

3. The method according to claim 1, characterized in that After obtaining the aerosol optical thickness data and infrared data in the target training study area and calculating the near infrared index data according to the infrared data, the method further includes: The near-infrared index data are resampled by the nearest neighbor sampling method so that the spatial resolution of the resampled aerosol optical thickness data and the near-infrared index data remain consistent.

4. The method according to claim 3, characterized in that After obtaining the reanalysis meteorological data in the target training study area, the method further includes: The reanalysis meteorological data are cropped, and the cropped reanalysis meteorological data are resampled by Kriging interpolation method, so that the spatial resolution of the resampled reanalysis meteorological data is consistent with the resampled aerosol optical thickness data and near infrared index data.

5. The method according to claim 1, characterized in that The method further comprises: The daily average resolution surface data is synthesized into monthly average resolution surface data by a maximum value synthesis method, and the monthly average resolution surface data is added to the data sample.

6. The method according to claim 1, characterized in that The aerosol optical thickness data, near infrared index data, reanalysis meteorological data and daily average resolution surface data in the target training study area are taken as a data sample, and a data sample is input into the regression tree model to obtain the output predicted salt density value, including: Creating multiple rules based on data samples of the target training study area; wherein each rule corresponds to a type of input data and is associated with a related multiple linear regression model and judgment condition, and all multiple linear regression models and judgment conditions constitute the regression tree model; At the terminal node of the regression tree model, the regression values ​​output by all the multivariate linear regression models are summed and averaged, and the obtained average value is used as the predicted salt density value.

7. The method according to claim 1, characterized in that The method further comprises: The data samples are divided into a training set and a validation set by the leave-one-out method, and the correlation coefficient between the measured salt density value and the predicted salt density value of the data samples in the validation set is calculated. When the correlation coefficient is greater than the preset coefficient threshold, it is determined that the validation passes the accuracy requirement.

8. A method for monitoring pollution distribution, characterized in that: The method comprises: Obtain aerosol optical thickness data, near infrared index data, reanalysis meteorological data and daily average resolution surface data of the target study area in a preset year, and input them into a regression tree model to obtain the output equivalent salt density distribution of the target study area; wherein the regression tree model is trained by the regression tree model training method according to claim 1; A pollution space distribution map is drawn based on the equal salt density distribution of the target study area; wherein the pollution space distribution map is used to indicate the pollution distribution situation of the target study area.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 8.

10. A device for monitoring pollution distribution, comprising a memory and a processor, characterized in that: The memory stores a computer program, and when the computer program is executed by the processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 8.

Images