Processing method, device and processing equipment for PM2.5 (Particulate Matter 2.5) estimation model
By using atmospheric apparent reflectivity data and meteorological geological environment data for spatiotemporal matching, the PM2.5 estimation model was trained, and data loss and insufficient accuracy in aerosol optical thickness observation were solved, and PM2 with higher accuracy and coverage was achieved .5 Observation, supporting air pollution prevention and control and health assessment.
Patent Information
- Application Number
- CN202411180003.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-12-06
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
When using aerosol optical thickness (AOD) to observe PM2.5, there are problems of data loss and insufficient processing accuracy, resulting in a lack of coverage and accuracy, especially in the case of cloud cover, surface brightness and snow cover. The correlation between non-uniform aerosol composition and local climatic conditions has poor influence.
Using the apparent reflectivity of the atmosphere as the basis, a novel PM2.5 estimation model is configured, and space-time matching is performed by obtaining sample atmospheric apparent reflectivity data, meteorological observation data and geological environment observation data, training the PM2.5 estimation model to bypass the air Error transfer of sol optical thickness for higher spatial resolution and coverage.
It effectively avoids the error transmission caused by AOD inversion, improves the accuracy and coverage of PM2.5 observations, meets the needs of high-precision PM2.5 observations, and provides strong data support for air pollution prevention and control and health effect assessment.
Smart Images

Figure CN119089200A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of air quality detection, and specifically to a processing method, device and processing equipment for a PM2.5 estimation model. Background Art
[0002] PM2.5 (fine particulate matter) refers to particles with an aerodynamic diameter of less than 2.5 μm. The higher the concentration in the air, the more serious the pollution. Severe PM2.5 pollution is associated with many health problems, such as cardiovascular disease and respiratory diseases. Obviously, high spatial resolution ground PM2.5 estimation is an important support for the refined prevention and control of air pollution and the assessment of health effects.
[0003] Traditional PM2.5 monitoring technology relies on PM2.5 observation stations (ground monitoring stations) to provide accurate spatiotemporal data. Without PM2.5 observation stations, there is no support for PM2.5 observation data. However, in reality, PM2.5 observation stations are often unevenly distributed, which makes it difficult to continuously monitor ground PM2.5 pollution.
[0004] Therefore, it is crucial to use satellite remote sensing methods for high-resolution, high-precision, and full-coverage PM2.5 inversion. Based on satellite remote sensing, the most commonly used method is to estimate the high-resolution distribution of ground PM2.5 concentration based on aerosol optical depth (AOD).
[0005] However, the inventors of the present application found that for aerosol optical depth observations, on the one hand, about 70% of the data was lost due to cloud cover, surface brightness and snow cover, and on the other hand, the non-uniform aerosol composition and local geographic climatic conditions would further lead to a poor correlation between aerosol optical depth and PM2.5.
[0006] In other words, the existing technology for observing PM2.5 based on aerosol optical thickness has problems with lack of coverage and processing accuracy in practical applications. Summary of the invention
[0007] The present application provides a processing method, device and processing equipment for a PM2.5 estimation model, which is used to bypass aerosol optical thickness and is based on atmospheric apparent reflectivity. A PM2.5 estimation model with more novel processing logic is configured, which can effectively avoid the error transmission caused by AOD inversion, and can also obtain higher spatial resolution and larger coverage. In practical applications, it can meet the needs of high-precision PM2.5 observations and provide strong data support for the refined prevention and control of air pollution and the assessment of health effects.
[0008] In a first aspect, the present application provides a method for processing a PM2.5 estimation model, characterized in that the method comprises:
[0009] Obtain sample atmospheric apparent reflectance data and sample meteorological observation data within the observation period;
[0010] Obtaining sample geological environment observation data related to sample atmospheric apparent reflectance data;
[0011] Obtain sample PM2.5 observation data obtained from PM2.5 observation sites during the observation period, wherein the sample PM2.5 observation data is used as the true value in the model training process;
[0012] Performing temporal and spatial matching on the sample atmospheric apparent reflectance data, the sample meteorological observation data, the sample geological environment observation data and the sample PM2.5 observation data to obtain an atmospheric apparent reflectance data set as a training sample;
[0013] Based on the atmospheric apparent reflectance dataset, a PM2.5 estimation model is trained. The PM2.5 estimation model is used to predict PM2.5 data at different locations based on the input atmospheric apparent reflectance data, meteorological observation data, and geological environment observation data.
[0014] In a second aspect, the present application provides a processing device for a PM2.5 estimation model, the device comprising:
[0015] A first acquisition unit is used to acquire sample atmospheric apparent reflectance data and sample meteorological observation data within an observation period;
[0016] A second acquisition unit is used to acquire sample geological environment observation data related to the sample atmospheric apparent reflectance data;
[0017] A third acquisition unit is used to acquire sample PM2.5 observation data obtained by observing PM2.5 observation sites during an observation period, wherein the sample PM2.5 observation data is used as a true value in a model training process;
[0018] A time-space matching unit is used to perform time-space matching on the sample atmospheric apparent reflectance data, the sample meteorological observation data, the sample geological environment observation data and the sample PM2.5 observation data to obtain an atmospheric apparent reflectance data set as a training sample;
[0019] The model training unit is used to train a PM2.5 estimation model based on an atmospheric apparent reflectance data set, wherein the PM2.5 estimation model is used to predict PM2.5 data at different locations based on input atmospheric apparent reflectance data, meteorological observation data, and geological environment observation data.
[0020] In a third aspect, the present application provides a processing device, including a processor and a memory, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the method provided in the first aspect of the present application or any possible implementation method of the first aspect of the present application is executed.
[0021] In a fourth aspect, the present application provides a computer-readable storage medium, which stores multiple instructions, and the instructions are suitable for a processor to load to execute the method provided by the first aspect of the present application or any possible implementation of the first aspect of the present application.
[0022] From the above content, it can be concluded that the present application has the following beneficial effects:
[0023] For the PM2.5 observation target, the present application obtains sample atmospheric apparent reflectance data and sample meteorological observation data within the observation period on the one hand, obtains sample geological environment observation data related to the sample atmospheric apparent reflectance data on the other hand, and obtains sample PM2.5 observation data observed by PM2.5 observation stations within the observation period on the other hand. The sample PM2.5 observation data is used as the true value in the model training process, and then the sample atmospheric apparent reflectance data, sample meteorological observation data, sample geological environment observation data and sample PM2.5 observation data are matched in time and space to obtain an atmospheric apparent reflectance data set as a training sample. At this time, based on the atmospheric apparent reflectance Data set, training PM2.5 estimation model, wherein the PM2.5 estimation model is used to predict PM2.5 data at different locations based on the input atmospheric apparent reflectance data, meteorological observation data and geological environment observation data. Under this model processing mechanism, this application bypasses aerosol optical depth and configures a PM2.5 estimation model with more novel processing logic based on atmospheric apparent reflectance. It can effectively avoid the error transmission caused by AOD inversion, and can also obtain higher spatial resolution and larger coverage. In practical applications, it can meet the needs of high-precision PM2.5 observations and provide strong data support for the refined prevention and control of air pollution and the assessment of health effects. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.
[0025] Figure 1 A flow chart of a processing method for the PM2.5 estimation model of this application;
[0026] Figure 2 A schematic diagram of the structure of a processing device for the PM2.5 estimation model of this application;
[0027] Figure 3 A schematic diagram of the structure of the processing equipment of this application. DETAILED DESCRIPTION
[0028] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.
[0029] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. The naming or numbering of steps in this application does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The process steps that have been named or numbered can change the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.
[0030] The division of modules in this application is a logical division. There may be other division methods when it is implemented in actual applications. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection between modules can be electrical or other similar forms, which are not limited in this application. In addition, the modules or submodules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed in multiple circuit modules, and some or all of the modules may be selected according to actual needs to achieve the purpose of the present application.
[0031] Before introducing the processing method of the PM2.5 estimation model provided in this application, the background content involved in this application is first introduced.
[0032] The processing method, device and computer-readable storage medium of the PM2.5 estimation model provided in this application can be applied to processing equipment to bypass aerosol optical thickness and, based on the atmospheric apparent reflectivity, configure a PM2.5 estimation model with more novel processing logic. It can effectively avoid the error transmission caused by AOD inversion, and can also obtain higher spatial resolution and larger coverage. In practical applications, it can meet the needs of high-precision PM2.5 observations and provide strong data support for the refined prevention and control of air pollution and the assessment of health effects.
[0033] The processing method of the PM2.5 estimation model mentioned in this application can be executed by a processing device of the PM2.5 estimation model, or a server, a physical host, or a user equipment (UE) or other different types of processing equipment that integrates the processing device of the PM2.5 estimation model. Among them, the processing device of the PM2.5 estimation model can be implemented in hardware or software, and the UE can be a terminal device such as a smart phone, a tablet computer, a laptop computer, a desktop computer, or a personal digital assistant (PDA), and the processing device can be set in a device cluster.
[0034] Among them, considering that the processing method of the PM2.5 estimation model provided in this application is mainly data processing based on existing data, therefore, the processing method of the PM2.5 estimation model provided in this application, or the processing device equipped with the application service corresponding to the processing method of the PM2.5 estimation model provided in this application, only needs to meet the requirements of data processing capabilities. The specific equipment type and equipment deployment form can be flexibly adjusted according to actual conditions, and this application does not make specific limitations.
[0035] Next, we will begin to introduce the processing method of the PM2.5 estimation model provided by this application.
[0036] First, see Figure 1 , Figure 1 A flow chart of a method for processing a PM2.5 estimation model of the present application is shown. The method for processing a PM2.5 estimation model provided by the present application may specifically include the following steps S101 to S105:
[0037] Step S101, obtaining sample atmospheric apparent reflectance data and sample meteorological observation data within an observation period;
[0038] It can be understood that, corresponding to the training process of the PM2.5 estimation model, the present application scheme needs to configure corresponding training samples, and the training samples take into account the different types of data involved, so they can be constrained to an observation time period, and the model training can be carried out with training samples involving different types of data within the observation time period.
[0039] It is easy to see that the samples obtained here are two types of satellite observation data products, namely, sample atmospheric apparent reflectance data and sample meteorological observation data. From this, it can be seen that the PM2.5 data estimation logic to be implemented in this application, in addition to atmospheric apparent reflectance data, may also involve other aspects of meteorological observation data. In other words, meteorological influencing factors are introduced as auxiliary data for estimating PM2.5 data.
[0040] As an example here, the observation time period that can be adjusted according to actual conditions may be specifically 2018-2020.
[0041] As another example here, the sample atmospheric apparent reflectance data may specifically be an atmospheric apparent reflectance observation product output by a multispectral camera (WFV) carried on the Gaofen-1 (GF-1) satellite, with a spatial resolution of 16m×16m.
[0042] As another example here, the sample meteorological observation data may specifically be ERA5 meteorological data output by the European Centre for Medium-Range Weather Forecasts (ECMWF).
[0043] As an exemplary embodiment of the present application, the sample atmospheric apparent reflectance data includes atmospheric apparent reflectance data of the blue light band (which can be recorded as BLUE_TOAR), atmospheric apparent reflectance data of the green light band (which can be recorded as GREEN_TOAR), atmospheric apparent reflectance data of the red light band (which can be recorded as RED_TOAR), and atmospheric apparent reflectance data of the near infrared band (which can be recorded as NIR-_RED_TOAR). These four types of atmospheric apparent reflectance data can be directly referred to as TOAR.
[0044] In addition, the sample atmospheric apparent reflectance data also includes longitude (lon), latitude (lat), Normalized Difference Vegetation Index (NDVI), satellite altitude angle (Satellite Azimuth), satellite zenith angle (Satellite Zenith), solar altitude angle (Solar Azimuth) and solar zenith angle (Solar Zenith). The latter four angles can be referred to as angles.
[0045] As an exemplary embodiment of the present application, the sample meteorological observation data may specifically include 10m u-component of wind (which can be recorded as u10), 10m v-component of wind (which can be recorded as v10), 2m temperature (which can be recorded as 2tm), boundary layer height (which can be recorded as b1h), surface pressure (which can be recorded as sp), total columnar ozone (which can be recorded as tco3), total columnar water vapor (which can be recorded as tcwv) and relative humidity (which can be recorded as rh).
[0046] It should be understood that the specific data parameters involved do not need to be obtained from specific satellite observation products together, and can actually be obtained from different satellite observation products.
[0047] Step S102, obtaining sample geological environment observation data related to sample atmospheric apparent reflectance data;
[0048] It can be understood that the present application believes that the geological environment or ground topography factors can affect the distribution of PM2.5. Therefore, the training samples of the present application can also involve sample geological environment observation data to provide another data support for high-precision PM2.5 estimation.
[0049] Among them, the sample geological environment observation data and the sample atmospheric apparent reflectance data are correlated. Theoretically, the two should have the same time range and regional range. However, considering that the frequency of geological environment changes is low in actual situations, or even there may be no change, it is feasible to obtain sample geological environment observation data with the same / similar range as the sample atmospheric apparent reflectance data.
[0050] As an exemplary embodiment of the present application, the sample geological environment observation data may specifically include a digital elevation model (DEM) and a land cover dataset.
[0051] As an example, the digital elevation model may adopt the Endeavour Space Shuttle DEM product, and the land cover dataset may adopt the Annual China Land Cover Dataset (CLCD) product.
[0052] Step S103, obtaining sample PM2.5 observation data observed by PM2.5 observation sites during the observation period, wherein the sample PM2.5 observation data is used as the true value in the model training process;
[0053] It can be understood that corresponding to the training of the model, the training samples need to be configured with annotations so that they can be used as true values in the model training process. In layman's terms, this is the PM2.5 estimation result that the model theoretically needs to output.
[0054] In this regard, the present application can use sample PM2.5 observation data obtained from a small number of unevenly distributed PM2.5 observation stations in actual situations, which has the effect of conveniently implementing labeling processing.
[0055] It should be understood that, for the above three steps, there is no specific limitation on the timing of data acquisition among the three. They can be performed simultaneously, in the order of the steps given, in other orders or even in random time sequences. These are all possible in actual situations as long as the data acquisition goal is achieved.
[0056] In addition, for the data acquisition processing in the above three steps, it can be understood that in actual situations, it is usually direct acquisition of ready-made data, such as manual input or direct extraction (including local or remote extraction methods).
[0057] Step S104, performing spatiotemporal matching on the sample atmospheric apparent reflectance data, the sample meteorological observation data, the sample geological environment observation data and the sample PM2.5 observation data to obtain an atmospheric apparent reflectance data set as a training sample;
[0058] It can be understood that the sample atmospheric apparent reflectance data, sample meteorological observation data, sample geological environment observation data and sample PM2.5 observation data obtained earlier, these different types / aspects of data, have scattered characteristics / situations. In order to form data that is convenient for model use, time-space matching can be carried out to form a highly integrated atmospheric apparent reflectance data set.
[0059] As an exemplary embodiment of the present application, the sample atmospheric apparent reflectance data, the sample meteorological observation data, the sample geological environment observation data and the sample PM2.5 observation data are temporally and spatially matched to obtain an atmospheric apparent reflectance data set as a training sample, which may specifically include:
[0060] Orthorectify the sample atmospheric apparent reflectance data to form atmospheric apparent reflectance data with a resolution of 30m×30m;
[0061] Resample the sample geological environment observation data to geological environment observation data with a resolution of 30m×30m;
[0062] The atmospheric apparent reflectance data with a resolution of 30m×30m and the geological environment observation data with a resolution of 30m×30m are matched in time and space;
[0063] The sample meteorological observation data were calculated by bilinear interpolation and resampled to meteorological observation data with a resolution of 30m×30m;
[0064] The atmospheric apparent reflectance data with a resolution of 30m×30m and the meteorological observation data with a resolution of 30m×30m are matched in time and space. During the matching process, the grid where the PM2.5 observation station is located is taken as the center, and the average value of the atmospheric apparent reflectance data, meteorological data and geological environment observation data of 5×5 pixels is taken.
[0065] The spatiotemporal matching processing of the average value involved here can be understood as converting the longitude and latitude of the PM2.5 observation site to the corresponding pixel position on the three raster data (atmospheric surface reflectance data, meteorological data, and geological environment observation data), and then taking the 5*5 pixel values around this position as the value matching the PM2.5 data at the corresponding position.
[0066] It can be understood that time-space matching is mainly about matching in time and space, so as to complete the data integration under the same dimension. The subsequent model can obtain the corresponding various data based on the same time and space, and carry out PM2.5 data estimation.
[0067] Among them, it can be seen that the embodiment here involves data alignment with a resolution of 30m×30m, which means that the PM2.5 data estimation processing that can be achieved by the subsequent PM2.5 estimation model can achieve a spatial resolution of 30m×30m resolution, and has the characteristics of PM2.5 inversion with high spatial resolution.
[0068] It is worth adding that the PM2.5 data estimation and processing that can be achieved by the PM2.5 model, due to the adoption of an inversion mechanism, also has the characteristic of obvious high temporal resolution compared to the PM2.5 observations carried out by PM2.5 observation points. That is, the PM2.5 data estimation results given by the PM2.5 model have a smaller time granularity. In layman's terms, the PM2.5 data can be estimated more densely in time, with the characteristics of continuity and smoothness, thereby better capturing short-term / instantaneous PM2.5 fluctuations.
[0069] At the same time, it can also be seen that the spatiotemporal matching of various types of data is based on the atmospheric apparent reflectance data, and this application also provides a more specific implementation plan on how to better perform spatiotemporal matching between atmospheric apparent reflectance data and meteorological observation data.
[0070] Furthermore, the present application also takes into account that there is a long time interval for the radiation calibration of satellite sensors on-orbit satellites, and the number of calibrations in a year is limited. For example, the GF-1 satellite radiation calibration adopts an annual calibration coefficient (that is, the calibration interval is one year). Therefore, in actual situations, it is difficult to monitor changes in the radiation performance of satellite sensors, and therefore there may be certain calibration errors.
[0071] Therefore, in order to ensure the high-precision PM2.5 inversion effect based on the atmospheric apparent reflectivity, the present application can also calibrate and optimize the atmospheric apparent reflectivity data in the atmospheric apparent reflectivity dataset.
[0072] In this regard, as an exemplary embodiment of the present application, before training the PM2.5 estimation model based on the atmospheric apparent reflectance data set, the processing method of the PM2.5 estimation model of the present application may also include:
[0073] The atmospheric apparent reflectance data in the atmospheric apparent reflectance dataset is optimized by the following formula:
[0074] L=G×D+O,
[0075] P=(π×L×d 2 ) / (E×cosθ),
[0076] Among them, L represents the radiation brightness value, G and O represent different radiation calibration coefficients, D represents the initial digital signal value (that is, the input atmospheric apparent reflectance data), P represents the optimization result of the atmospheric apparent reflectance data (that is, the output atmospheric apparent reflectance data), d is the distance between the sun and the earth, E represents the average solar irradiance, and θ represents the solar zenith angle.
[0077] It can be understood that the embodiment herein provides a specific implementation scheme for how to better correct and optimize the atmospheric apparent reflectance data in the atmospheric apparent reflectance data set, starting from a specific quantitative formula.
[0078] Among them, the radiation calibration coefficients G and O in the formula have a more continuous and stable correction optimization effect from the actual performance, which can promote more accurate and refined PM2.5 data estimation results.
[0079] As an example, for the radiation calibration coefficient G, the specific coefficient value adopted may be as follows:
[0080] Table 1 - Examples of radiation calibration coefficients G
[0081]
[0082] Among them, WFVi represents different multispectral cameras.
[0083] In addition, invalid data in the atmospheric apparent reflectance dataset can also be processed through corresponding removal operations, where it should be understood that the invalid data referred to here refers to data that cannot or is difficult to contribute to the subsequent training model, rather than data that has problems itself.
[0084] As an exemplary embodiment of the present application, before training the PM2.5 estimation model based on the atmospheric apparent reflectance data set, the processing method of the PM2.5 estimation model of the present application may further include:
[0085] Based on the atmospheric apparent reflectance data in the atmospheric apparent reflectance dataset, the existing cloud pixels are identified through the cloud pixel recognition strategy, the existing water pixels are identified through the water pixel recognition strategy, and the corresponding data of the cloud pixels and water pixels are removed from the atmospheric apparent reflectance dataset.
[0086] On the one hand, the cloud pixel recognition strategy can have the following contents:
[0087] H=B1-0.5B3,
[0088] V=(min(B1,B2,B3)) / (max(B1,Bb2,B3)),
[0089] Among them, H and V are two discrimination indicators (indexes), B1, B2 and B3 are the atmospheric apparent reflectance values of the red, green and blue bands respectively. If the current pixel meets the conditions of H>0.13, V>0.7 and B3>0.07, it is judged as a cloud pixel.
[0090] It can be understood that for the H value, considering that the spectral response of most surfaces to clouds is different in the blue and red bands, and the values of cloud pixels are usually greater than those of clear sky pixels, the H value is used to separate cloud pixels from clear sky pixels. V represents the ratio of the minimum reflectance to the maximum reflectance of visible light, which can also be used to separate cloud pixels from clear sky pixels.
[0091] On the other hand, the water pixel identification strategy can be as follows:
[0092] N=(B2-B4) / (B2+B4),
[0093] Among them, N is another discrimination index, B4 is the value of the atmospheric apparent reflectance in the near-infrared band, and if the current pixel satisfies the condition of N>0, it is judged as a water pixel.
[0094] It can be understood that N is another indicator designed to effectively separate water pixels from non-water pixels.
[0095] It is easy to see that in the embodiment here, the present application believes that cloud and water pixels will affect the estimation of PM2.5 data during the model training process. Therefore, the corresponding data can be deleted and only the other data can be retained, thereby achieving content optimization of the atmospheric apparent reflectance data set at the overall level, thereby further improving the training efficiency and prediction accuracy of the model.
[0096] In addition, as an exemplary embodiment, the data involved in the atmospheric apparent reflectance dataset may also be annotated with corresponding annual accumulation days.
[0097] Day of Year (DOY) refers to the number of days calculated continuously from January 1 of the current year. For example, the DOY corresponding to February 15 is 46 days.
[0098] That is to say, the processing logic of the PM2.5 estimation model of this application, in addition to the basic time of data annotation, can also use the annual accumulated day as a new auxiliary parameter in terms of the time dimension. It can be understood that this application believes that at the detail / micro level, PM2.5 will also show corresponding correlation between time points of the same annual accumulated day in different years. Therefore, the labeled annual accumulated days can be used to guide the model to pay attention to this situation and achieve a more accurate prediction effect.
[0099] The annual accumulated day may be carried by one or more of the previous data in the atmospheric apparent reflectance data set, or may be obtained by further marking on the basis of one or more of the previous data in the atmospheric apparent reflectance data set.
[0100] Step S105, training a PM2.5 estimation model based on the atmospheric apparent reflectance data set, wherein the PM2.5 estimation model is used to predict PM2.5 data at different locations based on the input atmospheric apparent reflectance data, meteorological observation data and geological environment observation data.
[0101] After obtaining a highly integrated training sample, namely the atmospheric apparent reflectance dataset, it is obvious that specific model training can be carried out on the initialized PM2.5 estimation model.
[0102] In each round of training, each group of atmospheric apparent reflectance data sets (atmospheric apparent reflectance data, meteorological observation data and geological environment observation data) is input into the model, so that the model can carry out PM2.5 estimation processing and complete forward propagation. Then, according to the PM2.5 estimation results, the loss function is calculated in combination with the corresponding sample PM2.5 observation data, and the model parameters are optimized according to the loss function calculation results to realize reverse propagation. In this way, when the preset model training requirements such as the number of training times, training time, and estimation accuracy are met, the model training can be completed. At this time, the model is a PM2.5 estimation model that can be put into practical use.
[0103] The processing logic of the PM2.5 estimation model, taking the specific parameters involved in the above exemplary embodiment as an example, can be expressed as the following mapping relationship:
[0104] PM 2.5 =f(DOY,TOAR,lon,lat,NDVI,angles,u10,v10,t2m,blh,sp,tco3,tcwv,rh,dem,lucc).
[0105] Among them, it is necessary to understand that the input parameters of the PM2.5 estimation model in actual application correspond to the parameters involved in the training samples during the model training process. For example, the mapping relationship expressed by the formula above involves DOY, TOAR (BLUE_TOAR, GREEN_TOAR, RED_TOAR and NIR-_RED_TOAR)), lon, lat, NDVI, angles (SatelliteAzimuth, Satellite Zenith, Solar Azimuth and Solar Zenith)), u10, v10, t2m, blh, sp, tco3, tcwv, rh and lucc.
[0106] In addition, the focus of this application is on how to configure training samples and the estimation logic of PM2.5 based on training samples. Therefore, how to carry out model training in the model training phase can adopt any type of existing model training scheme according to actual conditions. Of course, it is not ruled out that improved or even completely self-developed model training schemes can be introduced in specific operations. This is also possible under actual circumstances. This is also the case for the specific neural network model architecture used by the PM2.5 estimation model.
[0107] As an example, the PM2.5 estimation model can specifically adopt an existing deep neural network (DNN) model architecture, or adopt a model architecture improved on the basis of the existing deep neural network model architecture.
[0108] As an example, the PM2.5 estimation model can be divided into three parts in terms of model structure: input layer, hidden layer, and output layer. The input layer has a total of 22 input nodes, and the hidden layer is designed to be 4. The number of nodes in each layer is 512, 256, 128, and 64 respectively. The activation function of each hidden layer is a nonlinear rectification function.
[0109] In addition, the present application can use the random dropout method to prevent model overfitting. The basic idea of the Dropout method is to randomly set the output values of some neurons to 0 during the model training process, thereby making the structure of the neural network unstable and forcing the model to learn a more robust feature representation.
[0110] Specifically, in the Dropout method, during the model training process, each neuron is retained or discarded with a certain probability. Usually, the retention probability is P and the discard probability is 1-P, which can be achieved through some random sampling methods. During the forward propagation process, the output value of the discarded neuron will be set to 0. During the back propagation process, the discarded neurons will not participate in the error back propagation, thereby reducing the number of parameters and model complexity of the neural network, and thus avoiding overfitting.
[0111] The benefit of the Dropout method is that it allows the model to learn more robust feature representations, thereby improving the generalization ability of the model. At the same time, the Dropout method can also play a certain regularization role, making the weight parameters in the neural network smoother and further reducing the risk of overfitting.
[0112] It should be noted that the Dropout method is only used in the model training process, and the application of the Dropout method should be turned off during the testing process in order to obtain more accurate output results.
[0113] In addition, for the adopted model training scheme, as an example, the present application can specifically use ten-fold cross-validation to evaluate the generalization ability of the model. In the specific operation, the PM2.5 estimation model is cross-validated based on samples and sites respectively. In brief, the logic of the ten-fold cross-validation is to randomly divide the data set into ten subsets, and select nine subsets for model training each time, and one subset for model verification.
[0114] Specifically, R 2 , RMSE and MAE are used to evaluate the model performance (the indicators used for model evaluation and loss function are not necessarily the same, and the specific indicators used in the two links can be adjusted according to the actual situation). It can be understood that considering R 2 , RMSE and MAE are the existing indicators, so they will not be described in detail here. Of course, it should also be understood that in specific operations, the evaluation indicators may also be improved on the basis of existing indicators or even completely self-developed indicators.
[0115] After the model training is completed, it can obviously be put into practical application. In this regard, as an exemplary embodiment of the present application, after step S105 trains the PM2.5 estimation model based on the atmospheric apparent reflectance data set, the processing method of the PM2.5 estimation model of the present application may also include:
[0116] Obtain target atmospheric apparent reflectivity data, target meteorological observation data and target geological environment observation data at the prediction time;
[0117] The target atmospheric apparent reflectance data, target meteorological observation data and target geological environment observation data are input into the PM2.5 estimation model, so that the PM2.5 estimation model estimates the PM2.5 data corresponding to different locations.
[0118] It can be understood that the time range of the prediction time can be configured according to the PM2.5 data estimation needs under actual conditions. Correspondingly, it is necessary to configure the corresponding target atmospheric apparent reflectance data, target meteorological observation data and target geological environment observation data as model inputs, so that the PM2.5 estimation model can predict the PM2.5 data for the prediction time or a period of time later.
[0119] Among them, it should be understood that the PM2.5 estimation made by the PM2.5 estimation model of the present application is not limited to the location of the PM2.5 observation sites involved in the training samples, but covers the range involved in the training samples. Therefore, in actual situations, PM2.5 estimation can be performed at different locations to achieve PM2.5 estimation in mountainous areas, industrial areas, rural areas and urban areas, and due to the significant improvement in spatial resolution, the estimation effect in a local range can be further achieved. For example, in urban areas, it is also possible to achieve precise positioning of PM2.5 pollution at the regional scale or even the building scale within the city.
[0120] After obtaining the PM2.5 data estimation results output by the PM2.5 estimation model, further data processing can be performed according to the corresponding data processing strategies, such as local storage, remote storage, result push, output of completed estimation prompts, result display and secondary processing (for example, based on the daily average PM2.5 processing to obtain the monthly average PM2.5, seasonal average PM2.5, annual average PM2.5), etc.
[0121] Among them, in order to achieve a more vivid result presentation effect (not limited to the result display link, other data processing links may also indirectly involve the configuration of the result presentation effect), the PM2.5 data estimation results may also be converted into an image format, that is, mapping / drawing is performed based on the PM2.5 data estimation results, and the severity of PM2.5 may also be suddenly shown through different colors and other presentation effects, so that the estimation of the PM2.5 estimation model can be intuitively presented later.
[0122] In general, for the above scheme content (including various exemplary embodiments), for the PM2.5 observation target, the present application obtains sample atmospheric apparent reflectance data and sample meteorological observation data within the observation time period on the one hand, obtains sample geological environment observation data related to the sample atmospheric apparent reflectance data on the other hand, and obtains sample PM2.5 observation data observed by PM2.5 observation stations within the observation time period on the other hand. The sample PM2.5 observation data is used as the true value in the model training process, and then the sample atmospheric apparent reflectance data, sample meteorological observation data, sample geological environment observation data and sample PM2.5 observation data are matched in time and space to obtain the atmospheric apparent reflectance data as the training sample. Data set, at this time, based on the atmospheric apparent reflectance dataset, the PM2.5 estimation model is trained, wherein the PM2.5 estimation model is used to predict PM2.5 data at different locations based on the input atmospheric apparent reflectance data, meteorological observation data and geological environment observation data. Under this model processing mechanism, the present application bypasses the aerosol optical depth and is based on the atmospheric apparent reflectance. A PM2.5 estimation model with more novel processing logic is configured, which can effectively avoid the error transmission caused by AOD inversion, and can also obtain higher spatial resolution and larger coverage. In practical applications, it can meet the needs of high-precision PM2.5 observations, and provide strong data support for the refined prevention and control of air pollution and the assessment of health effects.
[0123] The above is an introduction to the processing method of the PM2.5 estimation model provided in this application. In order to facilitate better implementation of the processing method of the PM2.5 estimation model provided in this application, this application also provides a processing device for the PM2.5 estimation model from the perspective of functional modules.
[0124] See also Figure 2 , Figure 2 Schematic diagram of a structure of a processing device for a PM2.5 estimation model of the present application. In the present application, the processing device 200 for a PM2.5 estimation model may specifically include the following structure:
[0125] The first acquisition unit 201 is used to acquire sample atmospheric apparent reflectance data and sample meteorological observation data within an observation period;
[0126] The second acquisition unit 202 is used to acquire sample geological environment observation data related to the sample atmospheric apparent reflectance data;
[0127] The third acquisition unit 2038 is used to acquire sample PM2.5 observation data observed by the PM2.5 observation site during the observation period, wherein the sample PM2.5 observation data is used as the true value in the model training process;
[0128] A time-space matching unit 204 is used to perform time-space matching on the sample atmospheric apparent reflectance data, the sample meteorological observation data, the sample geological environment observation data and the sample PM2.5 observation data to obtain an atmospheric apparent reflectance data set as a training sample;
[0129] The model training unit 205 is used to train a PM2.5 estimation model based on an atmospheric apparent reflectance data set, wherein the PM2.5 estimation model is used to predict PM2.5 data at different locations based on input atmospheric apparent reflectance data, meteorological observation data, and geological environment observation data.
[0130] In another exemplary embodiment, the sample atmospheric apparent reflectance data includes atmospheric apparent reflectance data of a blue light band, atmospheric apparent reflectance data of a green light band, atmospheric apparent reflectance data of a red light band, and atmospheric apparent reflectance data of a near infrared band, and the sample atmospheric apparent reflectance data also includes longitude, latitude, normalized difference vegetation index, satellite altitude angle, satellite zenith angle, solar altitude angle, and solar zenith angle;
[0131] Sample meteorological observation data include 10-meter u-direction wind speed, 10-meter v-direction wind speed, 2-m temperature, boundary layer height, surface pressure, total columnar ozone, total columnar water vapor, and relative humidity;
[0132] The sample geological environment observation data include digital elevation models and land cover datasets.
[0133] In yet another exemplary embodiment, the data involved in the atmospheric apparent reflectance dataset are also annotated with corresponding annual accumulation days.
[0134] In another exemplary embodiment, the spatiotemporal matching unit 204 is specifically configured to:
[0135] Orthorectify the sample atmospheric apparent reflectance data to form atmospheric apparent reflectance data with a resolution of 30m×30m;
[0136] Resample the sample geological environment observation data to geological environment observation data with a resolution of 30m×30m;
[0137] The atmospheric apparent reflectance data with a resolution of 30m×30m and the geological environment observation data with a resolution of 30m×30m are matched in time and space;
[0138] The sample meteorological observation data were calculated by bilinear interpolation and resampled to meteorological observation data with a resolution of 30m×30m;
[0139] The atmospheric apparent reflectance data with a resolution of 30m×30m and the meteorological observation data with a resolution of 30m×30m are matched in time and space. During the matching process, the grid where the PM2.5 observation station is located is taken as the center, and the average value of the atmospheric apparent reflectance data, meteorological data and geological environment observation data of 5×5 pixels is taken.
[0140] In yet another exemplary embodiment, the apparatus further includes a data optimization unit 206, configured to:
[0141] The atmospheric apparent reflectance data in the atmospheric apparent reflectance dataset is optimized by the following formula:
[0142] L=G×D+O,
[0143] P=(π×L×d 2 ) / (E×cosθ),
[0144] Among them, L represents the radiation brightness value, G and O represent different radiation calibration coefficients, D represents the initial digital signal value, P represents the optimization result of atmospheric apparent reflectance data, d is the distance between the sun and the earth, E represents the average solar irradiance, and θ represents the solar zenith angle.
[0145] In yet another exemplary embodiment, the apparatus further includes a data optimization unit 206, configured to:
[0146] Based on the atmospheric apparent reflectance data in the atmospheric apparent reflectance dataset, the cloud pixels are identified by the cloud pixel identification strategy, the water pixels are identified by the water pixel identification strategy, and the corresponding data of the cloud pixels and the water pixels are removed from the atmospheric apparent reflectance dataset;
[0147] The cloud pixel identification strategy includes the following:
[0148] H=B1-0.5B3,
[0149] V=(min(B1,B2,B3)) / (max(B1,Bb2,B3)),
[0150] Among them, H and V are two discrimination indicators, B1, B2 and B3 are the atmospheric apparent reflectance values of the red, green and blue bands respectively. If the current pixel meets the conditions of H>0.13, V>0.7 and B3>0.07, it is judged as a cloud pixel;
[0151] The water pixel identification strategy includes the following:
[0152] N=(B2-B4) / (B2+B4),
[0153] Among them, N is another discrimination index, B4 is the value of the atmospheric apparent reflectance in the near-infrared band, and if the current pixel satisfies the condition of N>0, it is judged as a water pixel.
[0154] In yet another exemplary embodiment, the apparatus further includes a model application unit 207, configured to:
[0155] Obtain target atmospheric apparent reflectivity data, target meteorological observation data and target geological environment observation data at the prediction time;
[0156] The target atmospheric apparent reflectance data, target meteorological observation data and target geological environment observation data are input into the PM2.5 estimation model, so that the PM2.5 estimation model estimates the PM2.5 data corresponding to different locations.
[0157] This application also provides a processing device from the perspective of hardware structure, see Figure 3 , Figure 3 301, a memory 302, and an input / output device 303. The processor 301 is used to execute the computer program stored in the memory 302 to implement the following Figure 1 The steps of the processing method of the PM2.5 estimation model in the corresponding embodiment; or, the processor 301 is used to execute the computer program stored in the memory 302 to implement the following Figure 2 Corresponding to the functions of each unit in the embodiment, the memory 302 is used to store the processor 301 executing the above Figure 1 The computer program required for the processing method of the PM2.5 estimation model in the corresponding embodiment.
[0158] Exemplarily, the computer program may be divided into one or more modules / units, one or more modules / units are stored in the memory 302, and executed by the processor 301 to complete the present application. One or more modules / units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the computer device.
[0159] The processing device may include, but is not limited to, a processor 301, a memory 302, and an input / output device 303. Those skilled in the art will appreciate that the illustration is merely an example of a processing device and does not constitute a limitation on the processing device, and may include more or fewer components than shown in the illustration, or a combination of certain components, or different components. For example, the processing device may also include a network access device, a bus, etc., and the processor 301, the memory 302, the input / output device 303, etc. are connected via a bus.
[0160] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the processing device, and uses various interfaces and lines to connect various parts of the entire device.
[0161] The memory 302 can be used to store computer programs and / or modules. The processor 301 implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory 302 and calling the data stored in the memory 302. The memory 302 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the processing device, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.
[0162] When the processor 301 is used to execute the computer program stored in the memory 302, the following functions can be implemented:
[0163] Obtain sample atmospheric apparent reflectance data and sample meteorological observation data within the observation period;
[0164] Obtaining sample geological environment observation data related to sample atmospheric apparent reflectance data;
[0165] Obtain sample PM2.5 observation data obtained from PM2.5 observation sites during the observation period, wherein the sample PM2.5 observation data is used as the true value in the model training process;
[0166] Performing temporal and spatial matching on the sample atmospheric apparent reflectance data, the sample meteorological observation data, the sample geological environment observation data and the sample PM2.5 observation data to obtain an atmospheric apparent reflectance data set as a training sample;
[0167] Based on the atmospheric apparent reflectance dataset, a PM2.5 estimation model is trained. The PM2.5 estimation model is used to predict PM2.5 data at different locations based on the input atmospheric apparent reflectance data, meteorological observation data, and geological environment observation data.
[0168] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the processing device, processing equipment and corresponding units of the PM2.5 estimation model described above can refer to Figure 1 The description of the processing method of the PM2.5 estimation model in the corresponding embodiment will not be repeated here.
[0169] A person of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.
[0170] To this end, the present application provides a computer-readable storage medium, in which a plurality of instructions are stored, and the instructions can be loaded by a processor to execute the present application as follows: Figure 1 The steps of the processing method of the PM2.5 estimation model in the corresponding embodiment, the specific operation can refer to the following Figure 1 The description of the processing method of the PM2.5 estimation model in the corresponding embodiment will not be repeated here.
[0171] The computer-readable storage medium may include: a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.
[0172] Due to the instructions stored in the computer-readable storage medium, the present application can be executed. Figure 1 The steps of the processing method of the PM2.5 estimation model in the corresponding embodiment, therefore, the present application can be implemented as follows Figure 1 The beneficial effects that can be achieved by the processing method of the PM2.5 estimation model in the corresponding embodiment are detailed in the previous description and will not be repeated here.
[0173] The above is a detailed introduction to the processing method, device, processing equipment and computer-readable storage medium of the PM2.5 estimation model provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for technical personnel in this field, according to the idea of the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.
Claims
1. A processing method for PM2.5 estimation model, characterized in that: The method comprises: Obtain sample atmospheric apparent reflectance data and sample meteorological observation data within the observation period; Acquiring sample geological environment observation data related to the sample atmospheric apparent reflectance data; Obtaining sample PM2.5 observation data obtained from observations at PM2.5 observation sites during the observation period, wherein the sample PM2.5 observation data is used as the true value in the model training process; Performing temporal and spatial matching on the sample atmospheric apparent reflectance data, the sample meteorological observation data, the sample geological environment observation data and the sample PM2.5 observation data to obtain an atmospheric apparent reflectance data set as a training sample; Based on the atmospheric apparent reflectance data set, a PM2.5 estimation model is trained, wherein the PM2.5 estimation model is used to predict PM2.5 data at different locations based on input atmospheric apparent reflectance data, meteorological observation data and geological environment observation data.
2. The method according to claim 1, characterized in that: The sample atmospheric apparent reflectance data includes atmospheric apparent reflectance data of a blue light band, atmospheric apparent reflectance data of a green light band, atmospheric apparent reflectance data of a red light band, and atmospheric apparent reflectance data of a near infrared band. The sample atmospheric apparent reflectance data also includes longitude, latitude, normalized difference vegetation index, satellite altitude angle, satellite zenith angle, solar altitude angle, and solar zenith angle. The sample meteorological observation data include 10m u-direction wind speed, 10m v-direction wind speed, 2m temperature, boundary layer height, surface pressure, total columnar ozone, total columnar water vapor and relative humidity; The sample geological environment observation data includes a digital elevation model and a land cover dataset.
3. The method according to claim 1, characterized in that: The data involved in the atmospheric apparent reflectance dataset are also marked with corresponding annual accumulation days.
4. The method according to claim 1, characterized in that: The step of performing spatiotemporal matching on the sample atmospheric apparent reflectance data, the sample meteorological observation data, the sample geological environment observation data and the sample PM2.5 observation data to obtain an atmospheric apparent reflectance data set as a training sample includes: Ortho-correcting the sample atmospheric apparent reflectance data to form atmospheric apparent reflectance data with a resolution of 30m×30m; Resampling the sample geological environment observation data to geological environment observation data with a resolution of 30m×30m; Performing temporal and spatial matching on the atmospheric apparent reflectance data with a resolution of 30m×30m and the geological environment observation data with a resolution of 30m×30m; The sample meteorological observation data is calculated by bilinear interpolation and resampled to meteorological observation data with a resolution of 30m×30m; The atmospheric apparent reflectance data with a resolution of 30m×30m and the meteorological observation data with a resolution of 30m×30m are matched in time and space, and in the matching process, the grid where the PM2.5 observation site is located is used as the center to take the average value of the atmospheric apparent reflectance data, meteorological data and geological environment observation data of 5×5 pixels.
5. The method according to claim 1, characterized in that Before training the PM2.5 estimation model based on the atmospheric apparent reflectance dataset, the method further includes: The atmospheric apparent reflectance data in the atmospheric apparent reflectance dataset is optimized by the following formula: L=G×D+O, P=(π×L×d 2 ) / (E×cosθ), Among them, L represents the radiation brightness value, G and O represent different radiation calibration coefficients, D represents the initial digital signal value, P represents the optimization result of atmospheric apparent reflectance data, d is the distance between the sun and the earth, E represents the average solar irradiance, and θ represents the solar zenith angle.
6. The method according to claim 5, characterized in that Before training the PM2.5 estimation model based on the atmospheric apparent reflectance dataset, the method further includes: Based on the atmospheric apparent reflectance data in the atmospheric apparent reflectance data set, existing cloud pixels are identified by a cloud pixel identification strategy, existing water pixels are identified by a water pixel identification strategy, and corresponding data of the cloud pixels and the water pixels are removed from the atmospheric apparent reflectance data set; The cloud pixel identification strategy has the following contents: H=B1-0.5B3, V=(min(B1,B2,B3)) / (max(B1,Bb2,B3)), Among them, H and V are two discrimination indicators, B1, B2 and B3 are the atmospheric apparent reflectance values of the three bands of red, green and blue respectively. If the current pixel meets the conditions of H>0.13, V>0.7 and B3>0.07, it is determined to be the cloud pixel; The water pixel identification strategy has the following contents: N=(B2-B4) / (B2+B4), Wherein, N is another discrimination index, B4 is the value of the atmospheric apparent reflectance in the near-infrared band, and if the current pixel satisfies the condition of N>0, it is determined to be the water pixel.
7. The method according to claim 1, characterized in that After training the PM2.5 estimation model based on the atmospheric apparent reflectance data set, the method further includes: Obtain target atmospheric apparent reflectivity data, target meteorological observation data and target geological environment observation data at the prediction time; The target atmospheric apparent reflectance data, the target meteorological observation data and the target geological environment observation data are input into the PM2.5 estimation model, so that the PM2.5 estimation model estimates the PM2.5 data corresponding to different locations.
8. A processing device for a PM2.5 estimation model, characterized in that: The device comprises: A first acquisition unit is used to acquire sample atmospheric apparent reflectance data and sample meteorological observation data within an observation period; A second acquisition unit is used to acquire sample geological environment observation data related to the sample atmospheric apparent reflectance data; A third acquisition unit is used to acquire sample PM2.5 observation data observed by the PM2.5 observation site during the observation period, wherein the sample PM2.5 observation data is used as the true value in the model training process; A time-space matching unit, used for performing time-space matching on the sample atmospheric apparent reflectance data, the sample meteorological observation data, the sample geological environment observation data and the sample PM2.5 observation data to obtain an atmospheric apparent reflectance data set as a training sample; A model training unit is used to train a PM2.5 estimation model based on the atmospheric apparent reflectance data set, wherein the PM2.5 estimation model is used to predict PM2.5 data at different locations based on input atmospheric apparent reflectance data, meteorological observation data and geological environment observation data.
9. A processing device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a computer program, and the processor executes the method according to any one of claims 1 to 7 when calling the computer program in the memory.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to execute the method according to any one of claims 1 to 7.
Citation Information
Patent Citations
Method for estimating PM2.5 and PM10 by use of satellite multispectral image data
CN106979911A
Remote sensing estimation method for high-resolution near-real-time PM2.5 concentration
CN113159141A
Method and system for acquiring near-ground PM2.5 concentration in stationary meteorological satellite observation
CN113468799A
Satellite polarization PM2.5 estimation method and system based on machine learning fusion model
CN115356249A
PM2.5 concentration estimation method based on atmospheric apparent reflectance
CN117932255A
