Particulate matter concentration inversion method based on satellite data and meteorological data

Abstract

The invention relates to the technical field of deep learning and the technical field of concentration inversion, and discloses a particulate matter concentration inversion method based on satellite data and meteorological data. The method comprises the following steps: acquiring meteorological data of a station area based on a numerical meteorological model; acquiring grid ground features, time-space features and aerosol optical thickness of the site area; the particulate matter concentration of the site area is obtained based on the ground site observation data; the meteorological data, the grid ground features, the time-space features, the aerosol optical thickness and the particulate matter concentration of the site area serve as a training set, a deep learning model is trained, and a particulate matter concentration inversion model is generated; the particulate matter concentration inversion model is used for inverting the particulate matter concentration of the non-site area based on the real-time meteorological data of the non-site area and the virtual constellation observation satellite data, and the real-time meteorological data of the non-site area is obtained based on the numerical meteorological model. According to the method, the numerical meteorological model and the deep learning model are combined, so that the inversion precision is improved.

Description

Technical Field

[0001] The present disclosure relates to the fields of deep learning technology and concentration inversion technology, and in particular to a particle concentration inversion method based on satellite data and meteorological data. Background Art

[0002] With the frequent occurrence of haze weather, people pay more and more attention to air quality, and the impact of particulate matter on the global atmospheric environment is becoming more and more serious. Although ground environmental monitoring stations have been established to detect particulate matter concentrations, the established stations are relatively sparse and cannot reflect the particulate matter concentrations in all regions. Therefore, the inversion of particulate matter concentrations is crucial for environmental monitoring. Summary of the invention

[0003] In view of the above situation, an embodiment of the present application provides a method for inverting particulate matter concentration based on satellite data and meteorological data, aiming to solve the above problem or at least partially solve the above problem.

[0004] In a first aspect, an embodiment of the present application provides a method for inverting particulate matter concentration based on satellite data and meteorological data, the method comprising: acquiring meteorological data of a site area based on a numerical meteorological model; acquiring grid ground characteristics, temporal and spatial characteristics, and aerosol optical thickness of the site area based on virtual constellation observation satellite data; acquiring the particulate matter concentration of the site area based on ground site observation data; using the meteorological data, grid ground characteristics, temporal and spatial characteristics, aerosol optical thickness, and particulate matter concentration of the site area as a training set, training a deep learning model to generate a particulate matter concentration inversion model, the particulate matter concentration inversion model being used to invert the particulate matter concentration of a non-site area based on real-time meteorological data of the non-site area and virtual constellation observation satellite data, the real-time meteorological data of the non-site area being acquired based on the numerical meteorological model.

[0005] In the second aspect, an embodiment of the present application also provides a particle concentration inversion device based on satellite data and meteorological data, including: an acquisition module, used to acquire meteorological data of a site area based on a numerical meteorological model; acquire grid ground characteristics, time-space characteristics and aerosol optical thickness of the site area based on virtual constellation observation satellite data; acquire the particle concentration of the site area based on ground site observation data; a training module, used to train a deep learning model using the meteorological data, grid ground characteristics, time-space characteristics, aerosol optical thickness and particle concentration of the site area as a training set to generate a particle concentration inversion model, wherein the particle concentration inversion model is used to invert the particle concentration of a non-site area based on real-time meteorological data of the non-site area and virtual constellation observation satellite data, and the real-time meteorological data of the non-site area is acquired based on the numerical meteorological model.

[0006] In a third aspect, an embodiment of the present application further provides an electronic device, comprising: a processor; and a memory arranged to store computer executable instructions, wherein the executable instructions, when executed, cause the processor to perform the steps of the first aspect described above.

[0007] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, which stores one or more programs. When the one or more programs are executed by an electronic device including multiple application programs, the electronic device executes the steps of the first aspect above.

[0008] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: real-time meteorological data can be obtained through the numerical meteorological model, and then the deep learning model is trained in combination with the virtual constellation observation data and the ground station observation data to obtain a particle concentration inversion model, so as to obtain real-time meteorological data of non-station areas through the numerical meteorological model and combine the virtual constellation observation data, and use the particle concentration inversion model to predict the particle concentration in the non-station area. The present application combines the numerical meteorological model with the deep learning model. The numerical meteorological model can be used to obtain real-time dynamic and high-resolution meteorological data, providing important background information for the inversion of particle concentration. The nonlinear fitting ability and feature extraction ability of the deep learning model can be used to learn the complex relationship between particle concentration and other related parameters from a large amount of data. The combination of the two can invert more accurate particle concentrations and improve inversion accuracy and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0010] Figure 1 A schematic diagram of a process for inverting particulate matter concentration based on satellite data and meteorological data provided in an embodiment of the present application is shown;

[0011] Figure 2 A flow chart of a method for inverting particulate matter concentration based on satellite data and meteorological data provided by another embodiment of the present application is shown;

[0012] Figure 3 The structure diagram of the particle concentration inversion device based on satellite data and meteorological data provided in an embodiment of the present application is shown;

[0013] Figure 4 A schematic structural diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0014] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. 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 ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0015] It should be noted that the terms "first", "second", etc. in the specification and claims of the present 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 such use is interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the term "including" and its variants are to be interpreted as open-ended terms meaning "including but not limited to".

[0016] Before introducing the embodiments of the present application in detail, the following technical terms are introduced first.

[0017] 1. Numerical meteorological model: Numerical meteorological model plays a vital role in the inversion of particulate matter concentration. Taking the inversion of FY-4A satellite data as an example, the numerical meteorological model can provide real-time atmospheric temperature, humidity, air pressure and other parameters, which are crucial for the inversion of particulate matter concentration.

[0018] 2. Deep learning model: Deep learning has powerful nonlinear inversion capabilities. Through the combination of multiple linear layers and activation functions, it can learn complex feature representations, thereby better fitting the nonlinear relationship between input features and target results. For example, assuming there are input features and target results, the input data is 100 groups of observation data, each with 10 features, and the target data is 100 groups of corresponding target results. Using the mean square error as the loss function and the optimizer (Adam) to train the model, the relationship between input features and target results is learned in 1000 epochs, which can effectively improve the inversion accuracy.

[0019] Deep learning has great advantages in prediction. Deep learning models can effectively process large amounts of satellite data, fully exploit the spatiotemporal features in the data, and provide more accurate spatiotemporal information for particle concentration inversion. At the same time, deep learning can also be combined with other data sources, such as ground observation data and numerical meteorological model outputs, to further improve the accuracy and reliability of inversion.

[0020] 3. Feasibility of combining numerical meteorological models with deep learning: Numerical meteorological models can usually provide macro data such as physical parameters of the atmosphere and meteorological conditions, which have high physical accuracy and spatiotemporal continuity. The data required for deep learning is more diverse and can include satellite remote sensing images, ground observation data, etc. There are significant advantages in fusing numerical model data with the data required for deep learning. For example, temperature, humidity, air pressure and other data provided by numerical meteorological models can be used as one of the input features of deep learning models to help the model better understand the formation and transmission process of particulate matter. At the same time, satellite remote sensing image data can provide rich spatial information for deep learning models, enabling the model to better capture the spatial distribution characteristics of particulate matter concentrations. According to relevant studies, in meteorological forecasts based on deep learning, the fusion of numerical meteorological model data and satellite remote sensing image data can improve the prediction accuracy by about 10%.

[0021] Numerical meteorological models are built based on physical principles and can describe the physical processes of the atmosphere more accurately, but they may have limitations when dealing with complex nonlinear relationships and large amounts of observational data. Deep learning models, on the other hand, have powerful nonlinear fitting and feature extraction capabilities and can learn complex patterns and relationships from large amounts of data. The combination of the two can improve inversion accuracy. For example, in the fault classification model based on the sparrow algorithm-optimized probabilistic neural network (SSA-PPN), the sparrow algorithm is used to optimize the parameters of the neural network so that it can better adapt to different types of data. Similarly, in the inversion of particulate matter concentration, the atmospheric physical parameters provided by the numerical meteorological model can be used as constraints to optimize the parameters of the deep learning model and improve the inversion accuracy. At the same time, the deep learning model can further optimize and adjust the output of the numerical meteorological model to make up for the shortcomings of the numerical model in dealing with complex situations.

[0022] The present application is described in detail below through specific embodiments.

[0023] Figure 1 The schematic diagram of the process of the particle concentration inversion method based on satellite data and meteorological data provided in the embodiment of the present application is shown. Figure 1 It can be seen that the present application at least includes steps S101 to S104:

[0024] Step S101: Acquire meteorological data of the site area based on a numerical meteorological model.

[0025] Meteorological data includes temperature, humidity, air pressure, wind speed, precipitation, and stability information of each layer of the atmosphere. The stability information is used to indicate whether each layer of the atmosphere is stable. For example, the greater the stability index, the more stable the atmosphere is. For example, the stability information can be indicators such as the K index and the lift index.

[0026] In some embodiments, the numerical meteorological model is used to predict real-time dynamic meteorological data based on weather station observation data, sounding data, satellite data, radar data, etc.

[0027] Step S102: Obtaining the grid ground features, temporal and spatial features, and aerosol optical depth of the site area based on the virtual constellation observation satellite data.

[0028] Among them, the virtual constellation is usually composed of satellites from multiple countries, which may include satellites with different resolutions and different sensors. For example, Landsat 8 / 9 and Sentinel 2A / B satellites can be observed together to form a virtual constellation. Specifically, the satellite data observed by the virtual constellation may include remote sensing images, and grid ground features and aerosol optical depth (AOD) can be obtained by preprocessing the remote sensing images.

[0029] Grid ground features refer to the division of the earth's surface into a series of regular or irregular grids, each with unique codes and attributes. Specifically, grid ground features include surface thermal anomalies, industrial enterprise density, historical mean particle concentration, terrain elevation, built-up area ratio, industrial park distribution, and potential industrial plant area. Temporal and spatial features refer to parameters such as month, season, week, hour, longitude and latitude used to represent temporal features.

[0030] Step S103: Obtaining the particle concentration in the site area based on ground site observation data.

[0031] Among them, the ground stations include national control stations and provincial control stations. The first particle matter concentration of the station area is obtained through the national control station, and the second particle matter concentration of the station area is obtained through the provincial control station. The final particle matter concentration of the station area is determined based on the first particle matter concentration and the second particle matter concentration, for example, the average of the first particle matter and the second particle matter concentration is taken, or the maximum value is taken, or a certain calculation method is used to determine the final particle matter concentration.

[0032] The particulate matter in this application may be PM2.5, PM10, etc.

[0033] Step S104: The meteorological data, grid ground features, temporal and spatial features, aerosol optical depth and particle concentration of the site area are used as a training set to train a deep learning model to generate a particle concentration inversion model.

[0034] The particle concentration inversion model is used to invert the particle concentration in the non-station area based on the real-time meteorological data of the non-station area and the virtual constellation observation satellite data, and the real-time meteorological data of the non-station area is obtained based on the numerical meteorological model. In some embodiments, the particle concentration inversion model outputs the longitude and latitude of the non-station area and its corresponding particle concentration.

[0035] Specifically, there are many ways to optimize the strategy when training deep learning models. The first is data enhancement, which increases the diversity of data and improves the generalization ability of deep learning models by rotating, flipping, scaling and other operations on satellite remote sensing image data. The second is to use optimization algorithms, such as the Adam optimizer, to adjust the parameters of the deep learning model to minimize the loss function. The third is to perform model fusion, which combines the outputs of multiple different deep learning models to improve the inversion accuracy. For example, the outputs of the fully connected neural network and the LSTM model can be fused to give full play to the advantages of both. In addition, methods such as cross-validation can be used to evaluate the effects of different parameter combinations and optimization strategies and select the optimal parameters and strategies. By continuously adjusting and optimizing parameters, the accuracy and reliability of the inversion algorithm can be improved.

[0036] from Figure 1 It can be seen from the method shown that the present application can obtain real-time meteorological data through the numerical meteorological model, and then train the deep learning model in combination with the virtual constellation observation data and the ground station observation data to obtain the particle concentration inversion model, so as to obtain the real-time meteorological data of the non-station area through the numerical meteorological model and combine the virtual constellation observation data, and use the particle concentration inversion model to predict the particle concentration in the non-station area. The present application combines the numerical meteorological model with the deep learning model. The numerical meteorological model can be used to obtain real-time dynamic and high-resolution meteorological data, providing important background information for the inversion of particle concentration. The nonlinear fitting ability and feature extraction ability of the deep learning model can be used to learn the complex relationship between particle concentration and other related parameters from a large amount of data. The combination of the two can invert more accurate particle concentrations and improve inversion accuracy and reliability.

[0037] In some embodiments of the present application, satellite remote sensing images are acquired based on virtual constellation observation satellite data, the satellite remote sensing images are preprocessed, and aerosol optical depth is acquired based on the preprocessed satellite remote sensing images.

[0038] Specifically, the satellite remote sensing image is converted from RGB space to HSI space to generate a saliency image of the remote sensing image; pixels in the saliency image are identified based on the pixel recognition method to obtain different types of pixels; different aerosol optical depth processing methods are used for different types of pixels to obtain the aerosol optical depth of different regional types.

[0039] In some embodiments, different types of pixels in the saliency image are identified based on the maximum inter-class variance method. Specifically, different pixel types in the image are identified by maximizing the inter-class variance between the image foreground and background.

[0040] In other embodiments, different types of pixels can be identified with the help of a medium-resolution imaging spectrometer. Specifically, cloud pixels are identified based on the highest confidence cloud of the medium-resolution imaging spectrometer; dense vegetation pixels are identified based on the apparent reflectance of the 2.1 μm channel and the vegetation index; and pixels with a 2.1 μm channel reflectance greater than 0.15 are regarded as bright target pixels.

[0041] Furthermore, different processing methods are used for different types of pixels. For dense vegetation pixels, the aerosol optical depth is determined based on the MODIS dark pixel algorithm. The dark pixel algorithm is suitable for areas with low surface reflectivity, such as dense vegetation coverage areas. There is a good linear relationship between the reflectivity of these areas in the short-wave infrared band and the reflectivity of the red and blue light bands. Because the reduction effect of atmospheric aerosols on the short-wave infrared band is relatively small and can be ignored, the surface reflectivity of the red and blue light bands can be indirectly obtained through the known apparent reflectivity of the short-wave infrared band. The part of energy loss due to the presence of atmospheric aerosols is obtained by subtracting the surface reflectivity indirectly obtained through the short-wave infrared band from the apparent reflectivity of the red and blue light bands. The aerosol optical depth is finally obtained through the surface reflectivity, aerosol single scattering albedo, and transmittance.

[0042] For bright target pixels, the aerosol optical depth is determined based on the visible red light band lookup table and the blue light band lookup table. Specifically, the geometric observation information of the bright target pixels is read; the atmospheric parameters are determined from the visible red light band lookup table or the blue light band lookup table based on the geometric observation information; the surface reflectance value is determined; and the aerosol optical depth is determined based on the atmospheric parameters and the surface reflectance value.

[0043] In some embodiments of the present application, a visible red light band lookup table and a blue light band lookup table are constructed based on a 6s radiation transfer model; the visible red light band lookup table includes a first correspondence between geometric observation information, aerosol optical thickness, and atmospheric parameters, the geometric observation information includes the solar zenith angle, the satellite observation angle, and the relative azimuth angle, and the atmospheric parameters include the atmospheric hemispheric reflectivity, the atmospheric path transmittance of the sun-ground, the atmospheric path transmittance of the ground-sensor, and the atmospheric path radiation reflectivity; the blue light band lookup table includes a second correspondence between geometric observation information, aerosol optical thickness, and atmospheric parameters.

[0044] In some embodiments of the present application, real-time meteorological data of non-site areas are obtained based on numerical meteorological models; the real-time meteorological data of non-site areas, grid ground characteristics, time and space characteristics, and aerosol optical thickness of non-site areas are input into a particle concentration inversion model to obtain the particle concentration of the non-site areas.

[0045] In order to fully explain the particle concentration inversion method based on satellite data and meteorological data provided by this application, the following is combined with Figure 2 Explanation: Through aerosol optical thickness, meteorological data, grid ground features, time and space features, and particulate matter concentration data, various types of data are normalized and preprocessed to construct a feature set. 70% is used as a training set and 30% is used as a validation set. The deep learning model is trained. The trained model can invert the particulate matter concentration in non-site areas based on satellite observation data in non-site areas. Figure 2 The data formats are nc, csv and txt, and the formats of various data need to be unified into csv format.

[0046] In order to further verify the reliability of the inversion results, specific regional cases are used for verification:

[0047] Taking the Anhui, Jiangsu and Shanghai regions as an example, the MODIS L1B data on May 16 and May 25, 2016 were used to invert the particle concentration using a method that combines numerical meteorological models and deep learning. First, the numerical meteorological model was used to obtain the atmospheric parameters of the region, including temperature, humidity, air pressure, etc. Then, the satellite remote sensing images were preprocessed and pixel recognized to distinguish different targets. For densely vegetated areas, the MODIS dark pixel algorithm was used for processing, and for bright targets, the aerosol optical thickness was obtained by linear interpolation based on the surface reflectance values ​​of the blue light band obtained from the surface reflectance library. Then, atmospheric parameters, aerosol optical thickness, ground features, etc. were input into the particle concentration inversion model to invert the particle concentration.

[0048] The inversion results were compared and verified with the sun photometer observation data of the Taihu, Xuzhou, Hefei, Nanjing and Shanghai monitoring stations in the AERONET monitoring network. The results show that the absolute error does not exceed 0.08, the absolute value of the relative error is within 18%, the inversion results meet the research accuracy requirements, and the deviation between the inversion results and the actual observation data is small, which can roughly reflect the distribution of aerosol optical thickness in Anhui, Jiangsu and Shanghai during the study period.

[0049] In addition, actual case verification can also be carried out in other regions. For example, in the Beijing-Tianjin-Hebei region and the Pearl River Delta region, the reliability of the combined method can be further verified by comparing it with ground observation data and other satellite data inversion results. At the same time, the experience of actual case verification is continuously accumulated to further optimize the inversion algorithm and improve the inversion accuracy and reliability.

[0050] In the embodiments of the present application, the combination of numerical meteorological model and deep learning has many advantages. Numerical meteorological model can provide physical parameters and meteorological conditions of the atmosphere, providing important background information for the inversion of particle concentration. Deep learning has powerful nonlinear fitting and feature extraction capabilities, and can learn the complex relationship between particle concentration and other related parameters from a large amount of satellite remote sensing data. The combination of the two gives full play to their respective advantages and improves the accuracy and reliability of the inversion.

[0051] In the specific inversion process, by building a reasonable algorithm flow, determining key parameters and optimizing them, efficient particle concentration inversion is achieved. Compared with traditional methods, the method combining numerical meteorological models and deep learning has significant advantages in accuracy, stability and robustness. Actual case verification also shows that this method can roughly reflect the distribution of particle concentration in the study area, and the inversion results meet the research accuracy requirements.

[0052] For example, in actual cases in Anhui, Jiangsu and Shanghai, the absolute error does not exceed 0.08, and the absolute value of the relative error is within 18%. In comparison with traditional methods, the correlation coefficient can be increased to above 0.9, and the root mean square error is reduced to below 0.15. At the same time, the improved method shows better stability and robustness when dealing with complex terrain and meteorological conditions.

[0053] In some embodiments of the present application, a particle concentration inversion device based on satellite data and meteorological data is provided, and the particle concentration inversion device based on satellite data and meteorological data corresponds one-to-one to the particle concentration inversion method based on satellite data and meteorological data in the above embodiment. Figure 3 As shown, the device for inverting the particle concentration based on satellite data and meteorological data includes an acquisition module 101 , a training module 102 and an inversion module 103 .

[0054] The acquisition module 101 is used to acquire meteorological data of the station area based on the numerical meteorological model; acquire the grid ground characteristics, time and space characteristics and aerosol optical thickness of the station area based on the virtual constellation observation satellite data; and acquire the particle concentration of the station area based on the ground station observation data;

[0055] The training module 102 is used to train a deep learning model using the meteorological data, grid ground features, time and space features, aerosol optical thickness and particle concentration of the site area as a training set to generate a particle concentration inversion model. The particle concentration inversion model is used to invert the particle concentration of the non-site area based on the real-time meteorological data of the non-site area and the virtual constellation observation satellite data. The real-time meteorological data of the non-site area is obtained based on the numerical meteorological model.

[0056] In some embodiments of the present application, in the above-mentioned device, the acquisition module 101 is specifically used to acquire satellite remote sensing images based on virtual constellation observation satellite data; convert the satellite remote sensing images from RGB space to HSI space to generate a saliency image of the remote sensing image; identify pixels in the saliency image based on a pixel recognition method to obtain different types of pixels; use different aerosol optical thickness processing methods for different types of pixels to obtain aerosol optical thickness of different area types.

[0057] In some embodiments of the present application, the acquisition module 101 is specifically used to identify different types of pixels in a saliency image based on the maximum inter-class variance method; or to identify cloud pixels based on the highest confidence cloud of a medium-resolution imaging spectrometer; to identify dense vegetation pixels based on the 2.1 μm channel apparent reflectance and vegetation index; and to use pixels with a 2.1 μm channel reflectance greater than 0.15 as bright target pixels.

[0058] In some embodiments of the present application, the acquisition module 101 is specifically used to determine the aerosol optical depth for dense vegetation pixels based on the moderate resolution imaging spectroradiometer MODIS dark pixel algorithm; for bright target pixels, the aerosol optical depth is determined based on the visible red light band lookup table and the blue light band lookup table.

[0059] In some embodiments of the present application, the visible red light band lookup table and the blue light band lookup table are constructed based on the 6s radiation transfer model; the visible red light band lookup table includes a first correspondence between geometric observation information, aerosol optical thickness, and atmospheric parameters, the geometric observation information includes the solar zenith angle, the satellite observation angle, and the relative azimuth angle, and the atmospheric parameters include the atmospheric hemispheric reflectivity, the atmospheric path transmittance of the sun-ground, the atmospheric path transmittance of the ground-sensor, and the atmospheric path radiation reflectivity; the blue light band lookup table includes a second correspondence between geometric observation information, aerosol optical thickness, and atmospheric parameters.

[0060] In some embodiments of the present application, the acquisition module 101 is specifically used to read the geometric observation information of the bright target pixel; determine the atmospheric parameters from the visible red light band lookup table or the blue light band lookup table based on the geometric observation information; determine the surface reflectance value; and determine the aerosol optical thickness based on the atmospheric parameters and the surface reflectance value.

[0061] In some embodiments of the present application, the inversion module 103 is used to obtain real-time meteorological data of non-site areas based on a numerical meteorological model; the real-time meteorological data of the non-site areas, grid ground characteristics, time and space characteristics, and aerosol optical thickness of the non-site areas are input into the particle concentration inversion model to obtain the particle concentration of the non-site areas.

[0062] It should be noted that any of the above-mentioned particle concentration inversion devices based on satellite data and meteorological data can implement the above-mentioned particle concentration inversion method based on satellite data and meteorological data in a one-to-one correspondence, which will not be repeated here.

[0063] Figure 4 FIG. 1 is a schematic diagram showing the structure of an electronic device provided in an embodiment of the present application. Figure 4 As shown, at the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. Among them, the memory may include a memory, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage, etc. Of course, the electronic device may also include hardware required for other services.

[0064] The processor, network interface and memory can be interconnected through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0065] The memory is used to store programs. Specifically, the program may include program code, and the program code includes computer operation instructions. The memory may include internal memory and non-volatile memory, and provides instructions and data to the processor.

[0066] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a particle concentration inversion device based on satellite data and meteorological data at the logical level. The processor executes the program stored in the memory and is specifically used to execute the above method.

[0067] The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware decoding processor to be executed, or a combination of hardware and software modules in the decoding processor to be executed. The software module may be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0068] The electronic device can execute the particle concentration inversion method based on satellite data and meteorological data provided in multiple embodiments of the present application, and realize a particle concentration inversion device based on satellite data and meteorological data. Figure 3 The functions of the illustrated embodiment will not be described in detail in the embodiments of the present application.

[0069] An embodiment of the present application also proposes a computer-readable storage medium, which stores one or more programs, and the one or more programs include instructions. When the instructions are executed by an electronic device including multiple application programs, the electronic device can execute the particulate matter concentration inversion method based on satellite data and meteorological data provided by multiple embodiments of the present application.

[0070] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0071] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0072] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0073] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0074] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0075] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0076] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0077] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity or device including the element.

[0078] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0079] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for inverting particulate matter concentration based on satellite data and meteorological data, characterized in that: The method comprises: Obtain meteorological data for the site area based on numerical meteorological models; Based on the virtual constellation observation satellite data, the grid ground characteristics, temporal and spatial characteristics and aerosol optical depth of the site area are obtained; Obtain the particle concentration in the station area based on ground station observation data; The meteorological data, grid ground features, temporal and spatial features, aerosol optical depth and particulate matter concentration of the site area are used as training sets to train a deep learning model to generate a particulate matter concentration inversion model. The particulate matter concentration inversion model is used to invert the particulate matter concentration in the non-site area based on the real-time meteorological data of the non-site area and the virtual constellation observation satellite data. The real-time meteorological data of the non-site area is obtained based on the numerical meteorological model.

2. The method according to claim 1, characterized in that: The method of obtaining the aerosol optical thickness of the station area based on the virtual constellation observation satellite data includes: Obtain satellite remote sensing images based on virtual constellation observation satellite data; Convert satellite remote sensing images from RGB space to HSI space to generate saliency images of remote sensing images; Identify pixels in salient images based on pixel recognition methods and obtain different types of pixels; Different aerosol optical depth processing methods are used for different types of pixels to obtain the aerosol optical depth of different regional types.

3. The method according to claim 2, characterized in that The pixel recognition method is based on identifying pixels in the saliency image to obtain different types of pixels, including: Identify different types of pixels in a saliency image based on the maximum inter-class variance method; or Highest confidence cloud identification cloud pixel based on moderate resolution imaging spectrometer; Dense vegetation pixels are identified based on the apparent reflectance and vegetation index of the 2.1 μm channel; The pixels with a reflectivity greater than 0.15 in the 2.1 μm channel are regarded as bright target pixels.

4. The method according to claim 1, characterized in that: The aerosol optical thickness of different types of pixels is obtained by using different aerosol optical thickness processing methods, including: For dense vegetation pixels, the aerosol optical depth is determined based on the MODIS dark pixel algorithm; For bright target pixels, the aerosol optical depth is determined based on the visible red light band lookup table and the blue light band lookup table.

5. The method according to claim 4, characterized in that The visible red light band lookup table and the blue light band lookup table are constructed based on the 6s radiation transmission model; The visible red light band lookup table includes a first correspondence between geometric observation information, aerosol optical thickness, and atmospheric parameters, wherein the geometric observation information includes the solar zenith angle, the satellite observation angle, and the relative azimuth angle, and the atmospheric parameters include the atmospheric hemispheric reflectivity, the atmospheric path transmittance of the sun-ground, the atmospheric path transmittance of the ground-sensor, and the atmospheric path radiation reflectivity; The blue light band lookup table includes the second correspondence between geometric observation information, aerosol optical thickness and atmospheric parameters.

6. The method according to claim 4 or 5, characterized in that: For bright target pixels, determining the aerosol optical depth based on a visible red light band lookup table and a blue light band lookup table includes: Read the geometric observation information of bright target pixels; Determine the atmospheric parameters from a visible red light band lookup table or a blue light band lookup table based on the geometric observation information; Determine the surface reflectance value; The aerosol optical depth is determined based on the atmospheric parameters and the surface reflectivity value.

7. The method according to any one of claims 1 to 6, characterized in that: The method further comprises: Acquire real-time meteorological data in non-station areas based on numerical meteorological models; The real-time meteorological data, grid ground characteristics, temporal and spatial characteristics and aerosol optical thickness of the non-station area are input into the particle concentration inversion model to obtain the particle concentration of the non-station area.

8. A particle concentration inversion device based on satellite data and meteorological data, characterized in that: The device comprises: The acquisition module is used to obtain meteorological data of the station area based on the numerical meteorological model; obtain the grid ground characteristics, time and space characteristics and aerosol optical thickness of the station area based on the virtual constellation observation satellite data; and obtain the particle concentration of the station area based on the ground station observation data; The training module is used to train a deep learning model using the meteorological data, grid ground features, time and space features, aerosol optical depth and particle concentration of the site area as a training set to generate a particle concentration inversion model. The particle concentration inversion model is used to invert the particle concentration of the non-site area based on the real-time meteorological data of the non-site area and the virtual constellation observation satellite data. The real-time meteorological data of the non-site area is obtained based on the numerical meteorological model.

9. An electronic device, comprising: processor; as well as A memory arranged to store computer executable instructions, wherein when the executable instructions are executed, the processor performs the steps of the particle concentration inversion method based on satellite data and meteorological data as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including multiple application programs, enables the electronic device to perform the steps of the particulate matter concentration inversion method based on satellite data and meteorological data as described in any one of claims 1 to 7.

Images