Satellite observation sulfur dioxide vertical column total amount stripe correction method, system and equipment

Abstract

The invention discloses a satellite observation sulfur dioxide vertical column total amount stripe correction method, system and device, and relates to the technical field of satellite remote sensing data processing, and the method comprises the steps: obtaining a correction result based on any solar spectrum data in a solar spectrum data set, collected by a satellite, of a target area within a preset time length; determining SO2 column total amount original data-stripe correction data pairing samples, and further forming a training sample set; constructing a neural network model fusing multi-scale feature extraction, a space attention mechanism and an adaptive high-value protection weight; training the neural network model by adopting the training sample set to obtain a sulfur dioxide vertical column total amount stripe correction model; and inputting SO2 column total amount original data corresponding to the new solar spectrum data into the sulfur dioxide vertical column total amount stripe correction model to obtain corresponding SO2 column total amount stripe correction data. According to the method and the device, SO2 stripe correction considering noise removal precision and signal fidelity can be realized.

Description

Technical Field

[0001] This application relates to the field of satellite remote sensing data processing technology, and in particular to a method, system and equipment for correcting the vertical column total amount of sulfur dioxide observed by satellite. Background Technology

[0002] Sulfur dioxide, as one of the major air pollutants, has a significant impact on atmospheric environmental quality, climate systems, and the ecological environment due to its global distribution and concentration variations. Satellite remote sensing technology, with its advantages of wide coverage and short observation cycles, has become the core means of global SO2 concentration monitoring. Two-dimensional satellite spectrometers (such as the EMI instrument on the Gaofen-5 satellite and the Tropomi instrument on the Sentinel-5 satellite) can acquire high spatial resolution observation data on the vertical distribution of SO2. However, due to differences in the pixel performance of two-dimensional CCD detectors and interference from the complex space environment (such as cosmic rays and temperature fluctuations) on the instrument's optical system, significant strip noise is commonly found in the observed total SO2 vertical column data. This strip noise manifests as a systematic deviation along the detector column direction, severely interfering with the accurate inversion of SO2 concentration, leading to a decrease in the accuracy of distinguishing high-concentration areas from background areas, and even affecting the reliability of global SO2 source emission estimates. Existing stripe removal methods mainly include traditional algorithms based on statistical filtering (such as median filtering and wavelet denoising) and correction methods based on physical models. Traditional statistical filtering methods tend to cause smoothing distortion of the true SO2 concentration signal, especially in high-concentration pollution areas, resulting in peak attenuation. Physical model correction methods rely on accurate instrument response functions and environmental parameter modeling, but the performance drift of instruments in the complex space environment is difficult to model accurately, leading to limited stripe removal effectiveness. Therefore, existing methods cannot simultaneously achieve both noise removal accuracy and signal fidelity. Summary of the Invention

[0003] The purpose of this application is to provide a method, system, and device for correcting the total amount of sulfur dioxide vertical column in satellite observations, which can achieve SO2 strip correction that balances noise removal accuracy and signal fidelity.

[0004] To achieve the above objectives, this application provides the following solution.

[0005] Firstly, this application provides a satellite-based method for correcting the vertical column total sulfur dioxide stripe, including: Acquire a set of solar spectral data of the target area collected by satellite within a preset time period; Based on any solar spectral data in the solar spectral data set, a paired sample of SO2 total column raw data and strip correction data is determined; multiple paired samples of SO2 total column raw data and strip correction data constitute a training sample set. A neural network model integrating multi-scale feature extraction, spatial attention mechanism and adaptive high-value protection weights was constructed. The neural network model is trained using the training sample set to obtain a vertical column total sulfur dioxide strip correction model. The original data of total SO2 column corresponding to the new solar spectral data is input into the sulfur dioxide vertical column total strip correction model to obtain the corresponding SO2 column total strip correction data.

[0006] Secondly, this application provides a satellite-based vertical column total sulfur dioxide strip correction system, comprising: The data acquisition module is used to acquire a set of solar spectral data of the target area collected by the satellite within a preset time period; The sample generation module is used to determine paired samples of SO2 total column raw data and strip correction data based on any solar spectral data in the solar spectral data set; multiple paired samples of SO2 total column raw data and strip correction data constitute a training sample set. The model building module is used to build a neural network model that integrates multi-scale feature extraction, spatial attention mechanism and adaptive high-value protection weights; The model training module is used to train the neural network model using the training sample set to obtain a vertical column total sulfur dioxide strip correction model. The strip correction module is used to input the original data of SO2 column total amount corresponding to the new solar spectral data into the sulfur dioxide vertical column total amount strip correction model to obtain the corresponding SO2 column total amount strip correction data.

[0007] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a satellite observation method for vertical column total sulfur dioxide strip correction.

[0008] According to the specific embodiments provided in this application, the following technical effects are disclosed: Based on any solar spectral data from a set of solar spectral data collected by satellite within a preset time period for the target area, this application determines paired samples of raw SO2 column total data and strip correction data, thus achieving data pairing. Then, a neural network model integrating multi-scale feature extraction, spatial attention mechanism, and adaptive high-value protection weights is constructed and trained. Through this "data pairing training + neural network learning" technical solution, compared with traditional statistical filtering and physical model methods, strip removal accuracy is higher and the true signal of high-concentration areas is effectively preserved, avoiding peak attenuation. It can be directly used for research such as global SO2 source emission inventory compilation and cross-border atmospheric pollution transport simulation. Furthermore, the setting of integrating multi-scale feature extraction in the neural network is specifically designed for the strip noise characteristics of SO2 observation data, and the processing integrating multi-scale feature extraction, spatial attention mechanism, and adaptive high-value protection weights can achieve a balance between accurate noise removal and signal fidelity.

[0009] This application effectively solves the strip noise problem caused by differences in instrument performance and interference from the space environment, significantly improving the accuracy and data availability of SO2 vertical column total inversion, and providing reliable data support for atmospheric pollution monitoring, climate change research, and other fields. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating a satellite-based method for correcting the vertical column total amount of sulfur dioxide in an embodiment of this application.

[0012] Figure 2 This is a schematic diagram of an example training dataset.

[0013] Figure 3 This diagram illustrates a comparison between traditional strip correction methods and deep learning methods for strip correction.

[0014] Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] To address the problem that the vertical distribution of sulfur dioxide observed by existing two-dimensional satellite spectrometers exhibits obvious bands that are difficult to remove, this application provides a satellite observation method for SO2 vertical column total concentration band correction using neural networks. This method achieves accurate removal of band noise and efficient correction of SO2 vertical column concentration, while ensuring signal fidelity in high-concentration regions, ultimately outputting a high-quality SO2 column total concentration product with regular grids.

[0017] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] In one exemplary embodiment, such as Figure 1 As shown, a method for correcting the total vertical column of sulfur dioxide by satellite observation is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, it includes the following steps 101 to 105.

[0019] Step 101: Acquire a set of solar spectral data for the target area collected by satellite within a preset time period. Since there are areas of high SO2 concentration in the global SO2 concentration distribution, the data acquisition process before constructing the training set needs to cover both clean SO2 areas and areas with SO2 pollution to cover different SO2 concentration scenarios. This ensures the universality of the model trained in subsequent steps for strip correction of clean and polluted areas. Therefore, the target area includes at least clean SO2 areas and areas with SO2 pollution.

[0020] In a specific application example, step 101 includes the following steps.

[0021] (1) Download the set of global solar spectral data observed by satellite within the preset time period.

[0022] (2) Determine the collection frequency of the training sample set; for example, select satellite observation data for 12 months of the year, 3 days per month (36 days in total) to ensure coverage of SO2 concentration distribution characteristics in different seasons.

[0023] (3) Based on the acquisition frequency, extract multiple solar spectral data from the global solar spectral dataset; the solar spectral data includes two-dimensional spectral data corresponding to clean SO2 areas on the same day and two-dimensional spectral data corresponding to areas with SO2 pollution.

[0024] The clean SO2 area can be an open sea area without volcanic activity, where there is no human-caused SO2 emission and the SO2 concentration is close to zero during non-volcanic periods. The polluted SO2 area can be a region with high SO2 concentration. Correspondingly, the solar spectral data can include one sample from a clean area and two samples from polluted areas (volcanic activity areas / industrially dense areas) on the same day to ensure the balance of samples between the two scenarios.

[0025] (4) Multiple solar spectral data constitute a solar spectral data set.

[0026] Step 102: Based on any solar spectral data in the solar spectral data set, determine the SO2 column total raw data-strip correction data paired sample; multiple SO2 column total raw data-strip correction data paired samples constitute a training sample set.

[0027] In a specific application example, step 102 includes: performing the following processing on the two-dimensional spectral data corresponding to both clean SO2 areas and areas with contaminated SO2 on the same day: (21) Inversion is performed to determine the original matrix of total SO2 column; specifically, based on the two-dimensional spectral data observed by satellite, an adaptive iterative technique is adopted to simultaneously iterate and fit multiple parameters including total SO2 column, and combined with the atmospheric radiative transfer model and atmospheric extinction law, so that the observed and simulated atmospheric albedo data reach the optimal fit, and the original matrix of total SO2 column is obtained.

[0028] In practical applications, this process can also be completed after step (1) in step 101. That is, after determining the original matrix of total SO2 column, data collection and sample data construction are carried out.

[0029] (22) Based on preset conditions, the original matrix of the total SO2 column is divided into valid pixels and invalid pixels; wherein, the preset conditions are that the cloud cover is less than 30% and the total SO2 column is greater than -0.5 DU (1 DU = 2.69e16 mole / cm). 2 Pixels with a DU value less than 0.5 are considered valid pixels, while those with a DU value greater than 0.5 are considered invalid pixels.

[0030] (23) Based on all the effective pixels, the formula is used. Calculate the SO2 stripe value; in short, calculate the median as the SO2 stripe value for the corresponding column.

[0031] (24) Using the formula SO2 striping correction was performed to obtain the corrected SO2 values; for invalid pixels, the corrected SO2 value was set to 0; where, For the first SO2 stripe values ​​of the column, For the first A list of all rows of cells in a column, where MEDIAN is the mean function; This represents the corrected SO2 value in column i.

[0032] (25) Extract multiple two-dimensional matrices from the two-dimensional spectral data by random cropping.

[0033] (26) The original SO2 column total amount matrix and the corrected SO2 value corresponding to any of the two-dimensional matrices constitute a paired sample of original SO2 column total amount data and strip correction data.

[0034] After the above correction process is performed on both clean SO2 areas and areas with SO2 pollution, 200 and 100 two-dimensional matrices of 450×450 pixels are randomly selected respectively (corresponding to the single observation field of the TROPOMI instrument) to form 10,800 pairs of "raw data-strip correction data" samples, achieving full coverage in both time and space dimensions.

[0035] To address the striping phenomenon caused by differences in 2D CCD performance and the space environment, the original distribution matrix of SO2 column totals was manually destriped to construct a paired training set of "original data - striped correction data". The matrix of the original SO2 column totals served as the input features for the neural network model, while the destriped data matrix after strip correction served as the model training labels.

[0036] Step 103: Construct a neural network model that integrates multi-scale feature extraction, spatial attention mechanism, and adaptive high-value protection weights. The neural network model includes a multi-scale feature extraction module, a feature fusion and scale recovery module, an attention and adaptive weights module, and a noise prediction and signal recovery module.

[0037] (i) The multi-scale feature extraction module is used to capture the strip noise and SO2 concentration signal features at different scales in the original data of total SO2 column.

[0038] Before inputting the raw SO2 column total data into the multi-scale feature extraction module, data preprocessing is generally required. Specifically, the input features and labels are subjected to min-max normalization, with a normalization range of [0,1], and the formula is as follows: . These are the original input features. The value after normalization. , These are the minimum and maximum values ​​in the sample set, respectively, to ensure the consistency of the input data distribution and improve the model convergence speed.

[0039] The multi-scale feature extraction module includes a convolutional enhancement unit, an original scale branch, a 2x downsampling branch, and a 4x downsampling branch.

[0040] The convolutional enhancement unit is used to perform initial feature extraction and original scale feature enhancement on the original data of SO2 column total amount to obtain a convolutionally enhanced feature map. Specifically, the input data (input features after normalization) is initially extracted through a 7×7 convolutional layer, and then four 7×7 convolutional blocks are concatenated to enhance the spatial correlation capture of strip noise at the original scale, resulting in a convolutionally enhanced feature map.

[0041] The original scale branch is used to perform convolution processing on the convolutionally enhanced feature map to obtain the original scale feature map; specifically, the convolutionally enhanced feature map is output as the original scale feature map through a 3×3 convolutional layer.

[0042] The 2x downsampling branch is used to perform convolution and downsampling on the enhanced convolution feature map to obtain a 2x downsampling feature map. Specifically, the enhanced convolution feature map is downsampled using a 3×3 stride 2 convolution to focus on the distribution pattern of mesoscale strip noise and obtain a 2x downsampling feature map.

[0043] The 4x downsampling branch is used to perform two convolution and downsampling processes on the convolution-enhanced feature map to obtain a 4x downsampled feature map. Specifically, after downsampling the convolution-enhanced feature map using a 3×3 stride 2 convolution, it is downsampled again using a 3×3 stride 2 convolution to capture the trend of strip noise at the global scale, thus obtaining a 4x downsampled feature map.

[0044] (ii) The feature fusion and scale recovery module is used to align and fuse strip noise and SO2 concentration signal features at different scales using a combination of deconvolution and interpolation to obtain a comprehensive feature map. Specifically, the feature fusion and scale recovery module includes a 4x upsampling branch, a 2x upsampling branch, and a fusion unit.

[0045] The 4x upsampling branch is used to perform deconvolution upsampling on the 4x downsampling feature map to obtain a 4x restored original scale feature map; specifically, it is upsampled by 4×4 stride 2 deconvolution, and then restored to the original scale by 4×4 stride 2 deconvolution.

[0046] The 2x upsampling branch is used to perform deconvolution upsampling on the 2x downsampling feature map to obtain a 2x restored original scale feature map; specifically, it is directly restored to the original scale through a 4×4 stride 2 deconvolution.

[0047] The fusion unit is used to fuse the original scale feature map, the 2x restored original scale feature map, and the 4x restored original scale feature map by adding elements together, integrating multi-dimensional noise and signal features to obtain a comprehensive feature map.

[0048] (iii) The attention and adaptive weight module is used to dynamically learn the spatial attention weight distribution of the comprehensive feature map and assign coefficients to each pixel position, thereby adjusting the noise removal intensity in different SO2 concentration regions to obtain the attention-enhanced feature map. The attention and adaptive weight module includes a spatial attention mechanism unit and an adaptive weight prediction unit.

[0049] To address the specificity of SO2 strip noise distribution along the column direction, a dedicated spatial attention mechanism unit is designed to enhance the model's ability to focus on noisy regions. The spatial attention mechanism unit comprises multiple convolutional layers to compress the channel features of the synthesized feature map to obtain a single-channel attention map, which is then normalized to the [0,1] interval using an activation function to obtain the spatial attention weight matrix. Specifically, the spatial attention mechanism unit consists of three convolutional layers, sequentially compressing 64-channel features to 32 channels and then 16 channels, ultimately outputting a 1-channel attention map, which is then normalized to the [0,1] interval using a Sigmoid activation function.

[0050] To address the signal fidelity requirements in high SO2 concentration areas (such as volcanoes and industrial zones), an adaptive weighted prediction unit is designed. Its core function is to dynamically adjust the noise removal intensity for different concentration areas. The adaptive weighted prediction unit is used for: Based on the statistical characteristics (such as standard deviation and mean) of the total SO2 column data observed by satellite in the training sample set, the total SO2 column matrix is ​​divided into multiple regions, and a preset adaptive strategy is used to set corresponding differentiated noise removal intensity coefficients for different regions. The adjustment coefficient matrix is ​​obtained by multiplying the spatial attention weight matrix with the noise removal intensity coefficients to achieve an adaptive strategy of "weak denoising" in high-value regions and "strong denoising" in basic regions, ensuring that high-concentration signals are not attenuated. The adjustment coefficient matrix is ​​then weighted and fused with the comprehensive feature map to obtain the attention-enhanced feature map.

[0051] Based on different SO2 concentrations (X), the region is divided into an extreme high value region, a high value region, a basic region, a low value region, and a zero value region.

[0052] Extremely high value area: X ≥ mean + 8 times standard deviation (corresponding to strong emission scenarios such as volcanic eruption).

[0053] High value area: mean + 3 standard deviations ≤ X < mean + 8 standard deviations (corresponding to industrially dense areas).

[0054] Base area: mean - 3 standard deviations ≤ X < mean + 3 standard deviations (corresponding to background area).

[0055] Low value region: X < mean - 3 times standard deviation (corresponding to abnormally low value noise).

[0056] Zero value region: X=0 (corresponding to the ideal clean region).

[0057] (iv) The noise prediction and signal recovery module is used to predict the strip noise component by convolution (e.g., by predicting the strip noise component by 2 layers of 3×3 convolution) based on the attention-enhanced feature map, and to recover the clean signal by using a residual learning mechanism to obtain SO2 column total strip correction data.

[0058] The noise prediction and signal recovery module uses the following function formula: ;in, For SO2 column total amount strip correction data, This is the raw data for the total SO2 column volume. For the strip noise components predicted by convolution, This is for adjusting the coefficient.

[0059] Compared to directly predicting clean images, residual learning focuses on the accurate modeling of noise components, reducing the difficulty of model training, while dynamically adjusting the degree of noise removal, balancing strip removal effect and signal fidelity.

[0060] In this application, a strip noise correction network based on multi-scale feature fusion is adopted, which is designed specifically for the strip noise characteristics of SO2 observation data. The core point is to integrate multi-scale feature extraction, spatial attention mechanism and adaptive high value protection weight to achieve a balance between accurate noise removal and signal fidelity.

[0061] Step 104: Train the neural network model using the training sample set to obtain a vertical column total sulfur dioxide strip correction model.

[0062] Based on the model constructed in step 103 above, the model is trained. If the validation set loss does not decrease for 15 consecutive rounds (the decrease may be adjusted as needed), the early stopping mechanism is triggered, and the current optimal model is saved.

[0063] Step 105: Input the original data of SO2 total column amount corresponding to the new solar spectral data into the sulfur dioxide vertical column total amount strip correction model to obtain the corresponding SO2 column total amount strip correction data.

[0064] Based on the optimal model after training, strip correction is performed on the newly acquired raw data of SO2 column totals observed by satellite. The specific steps are as follows: Ensure that the input data is a two-dimensional matrix, and call up the data saved during the training phase. , The parameters are used to perform min-max normalization on the original input data, consistent with the training set. After normalization, the data range is forcibly constrained to [0,1]. Outliers outside this range are truncated at boundary values, and values ​​less than [0,1] are removed. The value is 0, greater than The value is set to 1. Then, the preprocessed data is input into the sulfur dioxide vertical column total amount striping correction model, outputting the clean image after striping and the predicted noise component. Finally, the normalized result of the clean image is restored to the actual SO2 column total value, using the following formula: .in, This is the total amount of secondary SO2 vertical column data after strip correction. This is the normalized total amount of secondary SO2 vertical columns after strip correction.

[0065] In a specific application example, after step 105, the method further includes: performing gridding processing on the SO2 column total amount strip correction data to obtain a regularly gridded three-level SO2 column total amount product. Specifically, the area to be gridded is divided into a 0.25°×0.25° (approximately 28 km×28 km) grid. For each target grid point, all secondary product pixels (i.e., SO2 column total amount strip correction data) within a radius of 18 km centered on the grid point are selected, and their weighted average value is calculated as the SO2 column total amount value after strip correction for that grid point. The weight is the reciprocal of the distance from the pixel to the grid point, and finally, a regularly gridded three-level SO2 column total amount product is obtained.

[0066] In summary, this application addresses the stripe noise problem in SO2 observation data from two-dimensional satellite spectrometers by proposing a "data pairing training + neural network learning" technical solution. This solution sequentially performs three levels of processing: SO2 column totality inversion based on the two-dimensional satellite spectrometer, acquisition of a training sample set for the SO2 column totality distribution, neural network model construction, and model training followed by SO2 prediction and grid point acquisition. Compared to traditional statistical filtering and physical model methods, this approach achieves higher stripe removal accuracy, reducing the inversion accuracy to below 0.05 DU, while effectively preserving the true signal in high-concentration areas and avoiding peak attenuation. It can be directly applied to research such as global SO2 source emission inventory compilation and cross-border atmospheric pollution transport simulation.

[0067] In a specific practical application, this application also provides a strip correction for the total SO2 column concentration in a densely industrial area in May of the target year, using TROPOMI observation Band 3 two-dimensional spectral data. The specific process of this application example is as follows: (1) Download the TROPOMI Band 3 two-dimensional spectral data from December of the historical year to November of the target year, invert the clean area and the industrial dense area, and use the adaptive iterative technique to invert the original matrix of SO2 total column, so as to obtain the monthly average SO2 distribution in May of the target year.

[0068] (2) Select clean areas from the 1st to the 3rd of each month, and filter out pixels with cloud cover less than 30% and total SO2 column greater than -0.5 DU but less than 0.5 DU as valid pixels, otherwise invalid pixels. Select the spectrum with the most valid SO2 pixels in the area for SO2 strip correction. After strip correction is completed, randomly crop 100 two-dimensional matrices of 450×450 pixels.

[0069] Similarly, corresponding to the selected 1-3 days mentioned above, three distinct high SO2 concentration areas were selected each month, and the band values ​​were calculated and manually corrected. After the band correction was completed, 200 two-dimensional matrices of 450×450 pixels were also randomly cropped.

[0070] (3) The original SO2 column total matrix is ​​used as the input feature of the neural network, and the striped data matrix after manual correction is used as the model training label. The resulting training dataset is as follows: Figure 2 As shown, this diagram illustrates the comparison between clean areas, contaminated areas, and the corresponding statistically corrected data.

[0071] First, the data is preprocessed, then a neural network is constructed. Based on this model architecture, the model is trained. The current optimal model is identified when the loss on the validation set does not decrease for 15 consecutive rounds. Based on the trained optimal model, strip correction is performed on the raw data of total SO2 column from satellite observations in May of the target year.

[0072] If a trained model already exists, it can be directly called. The preprocessed data is input into the current best model, which outputs a clean image after striping and the predicted noise components. Finally, the normalized result of the clean image is restored to the actual total SO2 column value.

[0073] (4) Select a portion of the area and divide it into a 0.25°×0.25° grid. For each target grid point, select all secondary product pixels within a radius of 18km centered on the grid point, and calculate their weighted average as the total SO2 column value after strip correction for that grid point. The weight is the reciprocal of the distance from the pixel to the grid point. Finally, the total SO2 column product of the regular gridded tertiary grid is obtained. Figure 3 As shown, compared with the original results and traditional statistical filtering methods, the strip correction and destripping effect provided in this application based on neural networks is more obvious, and can effectively preserve the true signal in high-concentration areas and avoid peak attenuation.

[0074] Based on the same inventive concept, this application also provides a system. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more system embodiments provided below can be found in the limitations of the method above, and will not be repeated here.

[0075] In one exemplary embodiment, a satellite-based vertical column total sulfur dioxide striping correction system is provided, comprising: The data acquisition module is used to acquire a set of solar spectral data of the target area collected by satellite within a preset time period.

[0076] The sample generation module is used to determine paired samples of SO2 column total raw data and strip correction data based on any solar spectral data in the solar spectral data set; multiple paired samples of SO2 column total raw data and strip correction data constitute a training sample set.

[0077] The model building module is used to build neural network models that integrate multi-scale feature extraction, spatial attention mechanisms, and adaptive high-value protection weights.

[0078] The model training module is used to train the neural network model using the training sample set to obtain a vertical column total sulfur dioxide strip correction model.

[0079] The strip correction module is used to input the original data of SO2 column total amount corresponding to the new solar spectral data into the sulfur dioxide vertical column total amount strip correction model to obtain the corresponding SO2 column total amount strip correction data.

[0080] In summary, addressing the technical challenge of removing noticeable bands in the vertical column total distribution of sulfur dioxide (SO2) observed by two-dimensional satellite spectrometers, this application constructs a paired training set of raw data and de-banded data, designs a dedicated neural network model, and achieves band removal and accurate correction of SO2 total column totals. Finally, gridding processing yields a regularly gridded three-level SO2 total column product. Compared with traditional mathematical methods, this application effectively solves the band noise problem caused by instrument performance differences and space environment interference, significantly improving the accuracy and data usability of SO2 vertical column total inversion, and providing reliable data support for atmospheric pollution monitoring, climate change research, and other fields.

[0081] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 4As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and databases. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media to run. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a satellite-based method for correcting the vertical column total sulfur dioxide concentration.

[0082] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0083] In one exemplary embodiment, a computer device is also provided, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in the above-described method embodiments.

[0084] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0085] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0086] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0087] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0088] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0089] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0090] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method of swath correction for satellite observations of total vertical column amounts of sulfur dioxide, characterized by, The method includes: Acquire a set of solar spectral data of the target area collected by satellite within a preset time period; Based on any solar spectral data in the solar spectral data set, a paired sample of SO2 total column raw data and strip correction data is determined; multiple paired samples of SO2 total column raw data and strip correction data constitute a training sample set. A neural network model integrating multi-scale feature extraction, spatial attention mechanism and adaptive high-value protection weights was constructed. The neural network model is trained using the training sample set to obtain a vertical column total sulfur dioxide strip correction model. The original data of total SO2 column corresponding to the new solar spectral data is input into the sulfur dioxide vertical column total strip correction model to obtain the corresponding SO2 column total strip correction data.

2. The method of claim 1, wherein the satellite observed sulfur dioxide vertical column total swath correction is based on a ratio of a first satellite observed sulfur dioxide vertical column total swath correction to a second satellite observed sulfur dioxide vertical column total swath correction. The target area includes clean SO2 areas and areas containing SO2 pollution. Acquire a set of solar spectral data of the target area collected by satellite within a preset time period, including: Download the set of global solar spectral data observed by satellite within a preset time period; Determine the collection frequency of the training sample set; Based on the acquisition frequency, multiple solar spectral data are extracted from the global solar spectral dataset; the solar spectral data includes two-dimensional spectral data corresponding to clean SO2 areas and two-dimensional spectral data corresponding to areas with SO2 pollution on the same day; The solar spectral data from multiple sources constitute a solar spectral data set.

3. The method of claim 2, wherein the satellite observed sulfur dioxide vertical column total swath correction is determined by: Based on any solar spectral data from the aforementioned solar spectral data set, determine a paired sample of raw SO2 column total data and strip correction data, including: For the two-dimensional spectral data corresponding to both clean SO2 areas and areas with SO2 contamination on the same day, the following processing was performed: Inversion is performed to determine the original matrix of total SO2 column volume; Based on preset conditions, the original matrix of the total amount of SO2 columns is divided into effective pixels and invalid pixels; Based on all the valid pixels, the SO2 strip value is calculated using the formula SO2 strip value = 0.5 * (SO2 max - SO2 min) SO2 band correction was performed using the formula to obtain the corrected SO2 value. Multiple two-dimensional matrices are extracted from the two-dimensional spectral data using a random pruning method; The original SO2 column total amount matrix and the corrected SO2 value corresponding to any of the two-dimensional matrices constitute a paired sample of original SO2 column total amount data and strip correction data. wherein, is the column of SO2 band values, is the column of all row elements, MEDIAN is the mean function; is the i-th column of corrected SO2 values.

4. The method of claim 1, wherein the satellite observed SO2 vertical column total swath correction is based on a ratio of a first SO2 vertical column total swath correction to a second SO2 vertical column total swath correction. The neural network model includes a multi-scale feature extraction module, a feature fusion and scale recovery module, an attention and adaptive weight module, and a noise prediction and signal recovery module. The multi-scale feature extraction module is used to capture the strip noise and SO2 concentration signal features at different scales in the original data of total SO2 column volume; The feature fusion and scale restoration module is used to align and fuse strip noise and SO2 concentration signal features at different scales using a combination of deconvolution and interpolation to obtain a comprehensive feature map. The attention and adaptive weight module is used to dynamically learn the spatial attention weight distribution of the comprehensive feature map and assign coefficients to each pixel position, thereby adjusting the noise removal intensity in different SO2 concentration regions to obtain the attention-enhanced feature map. The noise prediction and signal recovery module is configured to predict a strip noise component by convolution based on the attention-enhanced feature map, recover a clean signal by using a residual learning mechanism, and obtain SO2 column total amount strip correction data.

5. The satellite observation vertical column total sulfur dioxide strip correction method according to claim 4, characterized in that, The multi-scale feature extraction module comprises a convolution reinforcement unit, an original scale branch, a 2-fold downsampling branch, and a 4-fold downsampling branch. The convolution reinforcement unit is configured to perform initial feature extraction and original scale feature reinforcement on the SO2 column total amount original data to obtain a convolution-reinforced feature map; the original scale branch is configured to perform convolution processing on the convolution-reinforced feature map to obtain an original scale feature map; the 2-fold downsampling branch is configured to perform convolution and downsampling processing on the convolution-reinforced feature map to obtain a 2-fold downsampling feature map; and the 4-fold downsampling branch is configured to perform twice convolution and downsampling processing on the convolution-reinforced feature map to obtain a 4-fold downsampling feature map. The feature fusion and scale recovery module comprises a 4-fold upsampling branch, a 2-fold upsampling branch, and a fusion unit. The 4-fold upsampling branch is configured to perform inverse convolution and upsampling processing on the 4-fold downsampling feature map to obtain a 4-fold recovered original scale feature map; the 2-fold upsampling branch is configured to perform inverse convolution and upsampling processing on the 2-fold downsampling feature map to obtain a 2-fold recovered original scale feature map; and the fusion unit is configured to realize fusion of the original scale feature map, the 2-fold recovered original scale feature map, and the 4-fold recovered original scale feature map by element addition to obtain a comprehensive feature map.

6. The satellite observation vertical column total sulfur dioxide strip correction method according to claim 4, characterized in that, The attention and adaptive weight module comprises a spatial attention mechanism unit and an adaptive weight prediction unit. The spatial attention mechanism unit comprises multiple convolution layers to compress channel features of the comprehensive feature map to obtain a single-channel attention map, and normalize the single-channel attention map to the interval [0, 1] by using an activation function to obtain a spatial attention weight matrix. The adaptive weight prediction unit is configured to divide the SO2 column total amount matrix into multiple regions based on data statistical characteristics of the SO2 column total amount observed by a satellite, set corresponding noise removal intensity coefficients for different regions by using a preset adaptive strategy, multiply the spatial attention weight matrix and the noise removal intensity coefficients to obtain an adjustment coefficient matrix, and perform weighted fusion of the adjustment coefficient matrix and the comprehensive feature map to obtain an attention-enhanced feature map. Different regions include an extreme high-value region, a high-value region, a basic region, a low-value region, and a zero-value region.

7. The satellite observation vertical column total sulfur dioxide strip correction method according to claim 4, characterized in that, The noise prediction and signal recovery module adopts the following function formula: ; wherein, is the SO2 column total strip correction data, is the SO2 column total original data; is the strip noise component predicted by convolution, is the adjustment coefficient.

8. The method of claim 1, wherein the satellite observed SO2 vertical column total swath correction is based on a look-up table of SO2 vertical column total swath correction values. The method further comprises: performing grid processing on the SO2 column total amount strip correction data to obtain regular gridded three-level SO2 column total amount products.

9. A system for strip correction of satellite observed total vertical column amounts of sulfur dioxide, characterized by, The system comprises: a data acquisition module configured to acquire a set of solar spectrum data of a target region within a preset time length collected by a satellite; a sample generation module configured to determine SO2 column total amount original data-strip correction data paired samples based on any solar spectrum data in the set of solar spectrum data; and multiple SO2 column total amount original data-strip correction data paired samples constitute a training sample set. The model construction module is configured to construct a neural network model that integrates multi-scale feature extraction, a spatial attention mechanism, and adaptive high-value protection weights. The model training module is configured to train the neural network model using the training sample set to obtain a sulfur dioxide vertical column total amount strip correction model. The strip correction module is configured to input original SO2 column total amount data corresponding to new solar spectrum data into the sulfur dioxide vertical column total amount strip correction model to obtain corresponding SO2 column total amount strip correction data.

10. A computer device comprising: The memory, the processor, and the computer program stored in the memory and executable on the processor are characterized in that the processor executes the computer program to implement the satellite observation sulfur dioxide vertical column total amount strip correction method in any one of claims 1-8.

Images