Arsenic distribution modeling method based on multi-source data fusion

Abstract

The invention discloses an arsenic distribution modeling method based on multi-source data fusion, and the method comprises the steps: obtaining multi-source detection data, carrying out the time synchronization of the multi-source detection data, carrying out the fusion of the data after the time synchronization, obtaining a comprehensive data set, carrying out the feature extraction of the multi-source detection data through a principal component analysis method according to the comprehensive data set, and carrying out the feature extraction of the multi-source detection data. The method comprises the steps of obtaining a feature data set, constructing a three-dimensional space interpolation model according to the feature data set, calculating arsenic concentration in the three-dimensional space interpolation model for spatial analysis to obtain a three-dimensional concentration distribution model, and analyzing concentration gradient change data in the three-dimensional concentration distribution model to obtain concentration distribution rule data; performing spatial clustering according to the concentration distribution rule data to obtain an aggregation area distribution diagram, generating a diffusion path according to the aggregation area distribution diagram, performing clustering analysis according to the diffusion path to obtain pollution source distribution data, and obtaining a risk assessment result of arsenic pollution in food according to the pollution source distribution data and the three-dimensional concentration distribution model.

Description

Technical Field

[0001] This invention belongs to the field of rice detection technology, and in particular relates to a method for modeling arsenic distribution based on multi-source data fusion. Background Technology

[0002] Food testing is a crucial area for ensuring food safety and safeguarding public health. Arsenic, a highly toxic heavy metal, has its distribution and concentration in food directly impacting the assessment and control of human health risks. Accurately understanding the distribution of arsenic in rice can not only effectively identify contamination sources but also provide a scientific basis for formulating food safety standards. However, current methods for analyzing arsenic distribution in rice testing have significant limitations, primarily in the difficulty of integrating data generated by different testing devices, leading to insufficient completeness and reliability of analytical results. For example, analysis based on a single detection method often fails to fully reflect the complex distribution patterns of arsenic in rice cultivation, and the format and temporal differences between different data sources further exacerbate the difficulty of data fusion. Against this backdrop, the core challenge lies in how to effectively integrate multi-source detection data and accurately model its spatial distribution. When detecting arsenic content in rice products, spectral data may record concentration information on the sample surface, while mass spectrometry data provides deeper chemical speciation information. However, due to time asynchrony or inconsistent formats, the two methods are difficult to use together to analyze the distribution patterns of arsenic. Furthermore, this difficulty in data integration directly affects the modeling accuracy of the three-dimensional spatial distribution of arsenic concentration in food samples. Particularly when identifying arsenic contamination diffusion paths and accumulation areas, the lack of dynamic analysis of concentration gradient changes leads to a decrease in the accuracy of pollution source tracing and risk assessment. Therefore, how to construct a three-dimensional concentration distribution model based on gradient analysis to accurately identify arsenic contamination diffusion paths and accumulation areas has become a key issue in arsenic distribution modeling in food testing. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes an arsenic distribution modeling method based on multi-source data fusion, thereby resolving the issues present in the existing technologies.

[0004] To achieve the above objectives, this invention provides a method for arsenic distribution modeling based on multi-source data fusion, comprising:

[0005] Acquire multi-source detection data, including arsenic detection data in food from different devices. Synchronize the multi-source detection data over time and fuse the synchronized data to obtain a comprehensive dataset. Based on the comprehensive dataset, extract features from the multi-source detection data using principal component analysis to obtain a feature dataset. Based on the feature dataset, construct a three-dimensional spatial interpolation model and calculate the arsenic concentration in the three-dimensional spatial interpolation model for spatial analysis to obtain a three-dimensional concentration distribution model. Analyze the concentration gradient change data in the three-dimensional concentration distribution model to obtain concentration distribution pattern data. Perform spatial clustering based on the concentration distribution pattern data to obtain a clustering region distribution map. Based on the clustering region distribution map, generate diffusion paths. Perform cluster analysis based on the diffusion paths to obtain pollution source distribution data. Based on the pollution source distribution data and the three-dimensional concentration distribution model, obtain the risk assessment result of arsenic contamination in food.

[0006] Optionally, the process of time synchronization of the multi-source detection data includes:

[0007] Time series data and corresponding timestamps are obtained from the multi-source detection data. Based on the timestamps, the multi-source detection data is time-aligned using a linear interpolation method to obtain preliminary aligned data. The preliminary aligned data is then offset-corrected using a dynamic time correction algorithm to obtain time-synchronized data.

[0008] Optionally, the process of merging time-synchronized data includes:

[0009] The corresponding time-synchronized data from the spectrometer, chromatograph, and mass spectrometer are obtained to obtain the original dataset. The arsenic concentrations detected in the original dataset are then fused using a weighted average method to generate a comprehensive dataset.

[0010] Optionally, the process of feature extraction from multi-source detection data using principal component analysis includes:

[0011] The comprehensive dataset is cleaned and standardized to obtain a standardized dataset. Based on the standardized dataset, principal component analysis is applied to the time-synchronized data to calculate the correlation matrix between variables in the time-synchronized data, obtaining eigenvalues ​​and eigenvectors. The eigenvalues ​​and eigenvectors are then thresholded to obtain a dimensionality-reduced feature dataset. The feature dataset is then subjected to correlation analysis and sorting. Based on the sorting results, the eigenvalues ​​and eigenvectors are further filtered to obtain a simplified feature dataset. The simplified feature dataset is then validated to obtain the final feature dataset.

[0012] Optionally, the process of obtaining the three-dimensional concentration distribution model includes:

[0013] Based on the feature dataset, measurement data of arsenic concentration in food is obtained. Based on the measurement data, a three-dimensional spatial coordinate system is constructed to determine the position of the measurement data in the three-dimensional space, thus obtaining a spatial coordinate dataset. The three-dimensional spatial coordinate system is interpolated using the Kriging interpolation method, and the continuous distribution function of arsenic concentration is calculated to obtain an interpolation distribution model. The arsenic concentration is extracted from the interpolation distribution model, and the arsenic concentration value at each point in the three-dimensional space is calculated to obtain a three-dimensional concentration dataset. The three-dimensional concentration dataset is then visualized to generate a three-dimensional concentration distribution model.

[0014] Optionally, the process of obtaining concentration distribution data includes:

[0015] In the three-dimensional concentration distribution model, spatial coordinates and arsenic concentration time series data are obtained. The gradient value of arsenic concentration at each coordinate point is calculated based on the spatial coordinates and arsenic concentration time series data to obtain concentration gradient data. The concentration gradient data is optimized using the gradient descent method to obtain optimized distribution parameters. The changing trend of arsenic concentration in the arsenic concentration time series is analyzed based on the optimized distribution parameters to obtain the concentration change trend. It is determined whether the concentration change trend needs to be re-optimized using the gradient descent method. Based on the optimized distribution parameters and combined with the spatial coordinate parameters, the dynamic distribution law data of arsenic concentration is obtained.

[0016] Optionally, the process of obtaining the clustered area distribution map includes:

[0017] Arsenic concentration data and spatial coordinates are extracted from the concentration distribution data to obtain a spatial distribution set. The data points in the spatial distribution set are clustered using a clustering method to obtain preliminary regional divisions. The arsenic concentration in the preliminary regional divisions is judged and marked to obtain a set of high-value regions. Density analysis is performed on the high-value region set to obtain the boundary range of the clustered regions. Based on the boundary range, the network data is divided to obtain gridded concentration distribution data. The gridded concentration distribution data is rendered and visualized to obtain a distribution map of arsenic concentration clustered regions.

[0018] Optionally, the process of acquiring pollution source distribution data includes:

[0019] Environmental monitoring data is acquired, and pollution area distribution data is obtained based on the environmental monitoring data. The pollution area distribution data is preprocessed, and pollutant concentration values ​​are extracted from the preprocessed pollution area distribution data. The spatial distribution characteristics of pollutant concentration values ​​are analyzed using the kernel density estimation method to obtain the pollution area range. Cluster analysis is performed on the pollution area range to extract pollution diffusion path characteristics and obtain diffusion paths. The diffusion paths are judged, and the temporal changes of diffusion paths are analyzed using the reverse tracing method to obtain the location set of pollution sources. The pollution source distribution probability is calculated based on the location set of pollution sources using the kernel density estimation method. Based on the pollution source distribution probability, pollution source distribution data is obtained.

[0020] Optionally, the process for obtaining risk assessment results for arsenic contamination in food includes:

[0021] The pollution source distribution data is divided into grids to obtain the spatial distribution characteristics of the pollution sources. The concentration of arsenic is analyzed through a three-dimensional concentration distribution model, and the concentration distribution of arsenic is judged. Based on the judgment results, boundary decisions are made for the concentration distribution of arsenic to obtain high-risk areas. According to the spatial distribution characteristics of high-risk areas and pollution sources, the potential risk level of food contamination is judged. The Monte Carlo simulation method is used to calculate risk assessment parameters based on the potential risk level, and the quantitative risk value, i.e., the risk assessment result of arsenic contamination in food, is obtained.

[0022] On the other hand, the present invention provides an arsenic distribution modeling system based on multi-source data fusion for performing the above-described method.

[0023] Compared with the prior art, the present invention has the following advantages and technical effects:

[0024] This invention discloses a multi-source data analysis and risk assessment method for arsenic contamination in food. The method integrates detection data from spectrometers, chromatographs, and mass spectrometers to generate a comprehensive dataset. Principal component analysis is used to extract arsenic concentration characteristics, a three-dimensional spatial interpolation model is constructed to calculate the arsenic concentration distribution, and its dynamic distribution patterns are analyzed. High-value clustering areas are further identified to determine the location of contamination sources. Finally, risk assessment parameters are calculated using the three-dimensional concentration distribution model. This invention enables comprehensive analysis and assessment of arsenic contamination in rice, providing strong technical support for food safety supervision. Attached Figure Description

[0025] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0026] Figure 1 This invention provides an arsenic distribution modeling method based on multi-source data fusion.

[0027] Figure 2 This is an arsenic distribution modeling system based on multi-source data fusion, as described in this embodiment of the invention. Detailed Implementation

[0028] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0029] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0030] like Figure 1 As shown, this embodiment of an arsenic distribution modeling method based on multi-source data fusion may specifically include:

[0031] Step S101: Data is acquired from multi-source detection devices through a pre-established standardized access protocol, and the data format and timestamp are unified to generate a multi-source dataset with a consistent format.

[0032] Raw data is acquired from multiple sources using standardized protocols, and the data content is parsed using preset protocol configurations to obtain a structured raw dataset. If the formats of the raw datasets are inconsistent, a unified processing mechanism is used to convert the data into a standard format, resulting in a consistent multi-source dataset.

[0033] In one possible implementation, acquiring raw data from multiple sources is fundamental to achieving unified data processing. These multiple sources include spectrometers, chromatographs, and mass spectrometers. These devices provide data to relevant computing devices through corresponding standardized access protocols for data processing. Data of the same type is formatted using a unified data format, which generates a dataset containing device IDs, timestamps, and corresponding data formats.

[0034] Step S102: Based on the multi-source dataset, a time series alignment algorithm is used to correct the time synchronization problem and obtain the time-synchronized dataset.

[0035] Time series data and corresponding timestamps are obtained from multi-source datasets to determine the time range and sampling frequency of each dataset. If timestamp discrepancies exist, linear interpolation is used to align the time series, resulting in a preliminary aligned dataset. Based on the preliminary aligned dataset, the time offset between each time series is calculated, and it is determined whether the offset exceeds a preset threshold. If the offset exceeds the preset threshold, a dynamic time warping algorithm is used to adjust the time series, generating the final synchronized dataset.

[0036] For example, when acquiring time-series data and corresponding timestamps from a multi-source dataset, consider spectrometers, chromatographs, and mass spectrometers. Time-series data may be a record of measured data changing over time, with timestamps recorded in milliseconds. During data acquisition, data is extracted from each instrument through a predefined protocol interface, ensuring that timestamps and numerical values ​​appear in pairs.

[0037] For example, a spectrometer sensor records data once every 1 / 10 of a second, generating a time series such as 2025-07-23 12:00:00.000. A chromatograph records 100 times per second, generating a time series such as 2025-07-23 12:00:00.000.

[0038] In one possible implementation, the start and end times of each dataset are analyzed when determining the time range and sampling frequency.

[0039] For example, spectrometer data is sampled from 12:00:00.000 to 12:10:00.000 on July 23, 2025, at a sampling frequency of 10 times per second; chromatograph data is sampled from 12:00:01.000 to 12:10:01.000 on July 23, 2025, at a sampling frequency of 100 times per second. A comparison reveals an inconsistency in sampling frequencies. Linear interpolation methods are used to align time series, for example, interpolating the high-frequency data from the spectrometer to an interval of 100 times per second to generate time-point data aligned with the chromatograph data.

[0040] Specifically, when calculating the time offset, the starting timestamps of each time series are compared.

[0041] For example, if the timestamp discrepancy between the spectrometer and the chromatograph is 1 second, and the preset threshold is 0.5 seconds, further adjustments are needed if the discrepancy exceeds the threshold. The dynamic time warping algorithm adjusts the chromatograph's time series by matching similar data patterns, aligning it with the spectrometer's time series at key points.

[0042] For example, peak values ​​or corresponding abrupt change points are aligned to generate a corrected dataset.

[0043] Step S103: Obtain the detection data from the spectrometer, chromatograph, and mass spectrometer from the time-synchronized dataset, and fuse them to generate a comprehensive dataset.

[0044] A time-synchronized dataset is acquired by extracting detection data from spectrometers, chromatographs, and mass spectrometers to generate a raw dataset. Based on this raw dataset, a weighted average method is applied for data fusion to generate a comprehensive dataset. For this comprehensive dataset, an anomaly detection algorithm is used to determine data integrity, resulting in a validation comprehensive dataset.

[0045] For example, when extracting detection data from spectrometers, chromatographs, and mass spectrometers, specific data can be obtained from actual chemical analysis scenarios. Suppose in a laboratory, we need to analyze arsenic in a rice sample. The spectrometer provides UV-Vis spectral data, recording the absorbance at specific wavelengths; the ion chromatograph generates the corresponding content of substances related to arsenic ions; the mass spectrometer provides molecular weight and fragment ion information, such as the detection of arsenic. These data, each with a timestamp, constitute the original dataset. The data have different temporal resolutions. By integrating these data onto a unified timeline, for example, at 1-second intervals, the original dataset is generated, providing a foundation for subsequent fusion.

[0046] In one possible implementation, when applying a weighted average method for data fusion, weights can be assigned based on the detection accuracy of each instrument.

[0047] For example, spectrometer absorbance data has high precision and is assigned a weight of 0.5; chromatograph data is affected by baseline drift and has a weight of 0.3; mass spectrometer fragment ion data has a higher weight due to background noise and is assigned a weight of 0.2. For data at a specific time point, assuming the spectrometer records an absorbance of 0.8, the chromatograph detects an arsenic concentration of 2.0 ppm, and the mass spectrometer detects a signal intensity of 1000, the corresponding arsenic element values ​​are calculated using conventional analytical methods based on the above information. Then, a weighted average is taken to obtain the comprehensive concentration value. This method balances the reliability of different instruments through weighting, generating a comprehensive dataset that more fully reflects the chemical characteristics of the sample.

[0048] Specifically, when using anomaly detection algorithms to determine data integrity for comprehensive datasets, statistical anomaly detection methods can be used.

[0049] For example, by setting the mean and standard deviation of the data as benchmarks, data points deviating from three times the standard deviation are detected. Suppose the arsenic concentration sequence in the comprehensive dataset is [2.0, 2.1, 2.0, 5.0, 2.2] ppm, where 5.0 is significantly higher. The calculated mean is 2.66, and the standard deviation is 1.26. 5.0 exceeds three times the standard deviation and is marked as an outlier. By removing or correcting outliers, a validation comprehensive dataset is obtained. This method ensures the reliability and consistency of the data, providing a higher-quality data foundation for subsequent analysis.

[0050] For example, in practice, anomaly detection can also be combined with time-series trend analysis. Suppose that the concentration value in the monitoring data suddenly increases over a certain period, which may be due to instrument malfunction or sample contamination. By examining the trend before and after the initial time point, if the concentration value jumps from 2.0 ppm to 5.0 ppm in a short period and then quickly falls back, and the fragment ion data from the mass spectrometer does not show any anomalies, then it can be inferred that this is a temporary baseline drift of the chromatograph. Through this multi-faceted verification, the accuracy of anomaly detection is improved, and the generated comprehensive verification dataset can better support the conclusions of environmental analysis.

[0051] Step S104: For the comprehensive dataset, principal component analysis algorithm is used to extract the main features of arsenic concentration and generate a feature dataset.

[0052] Data preprocessing methods are employed to obtain raw data from a comprehensive dataset, perform data cleaning to remove missing and outlier values, resulting in a cleaned dataset. Based on the cleaned dataset, data standardization is performed to scale all variables to a uniform range, resulting in a standardized dataset. For the standardized dataset, principal component analysis (PCA) is used to calculate the correlation matrix between variables, obtaining eigenvalues ​​and eigenvectors. Based on the eigenvalues ​​and eigenvectors, principal components with a cumulative contribution rate reaching a preset threshold are selected to generate a dimensionality-reduced feature dataset. If the dimensionality of the feature dataset meets preset conditions, a feature selection method is used to retain features highly correlated with arsenic concentration, resulting in a simplified feature dataset. Based on the simplified feature dataset, data transformation is performed to map the features to a new space, generating the final feature dataset. For the final feature dataset, a result validation method is used to calculate the variance explained by the feature dataset, determining whether the extracted features meet the requirements for arsenic detection.

[0053] For example, in time-synchronized integrated datasets generated by spectrometers, chromatographs, and mass spectrometers, data preprocessing is a crucial step in ensuring data quality. Data cleaning aims to remove missing and outlier values. Suppose that in the light intensity data acquired from the spectrometer, at a certain time point, the light intensity value is abnormally high, reaching 10 times the normal value due to equipment vibration. Data cleaning would remove such outliers by setting a threshold, such as 3 times the standard deviation, and simultaneously impute missing values ​​using interpolation methods, such as filling in missing spectral intensities based on the average of preceding and following time points. This method ensures the continuity and reliability of the dataset, laying the foundation for subsequent analysis.

[0054] In one possible implementation, data standardization unifies variables with different dimensions to the same scale.

[0055] For example, the light intensity data range of a spectrometer might be 0 to 1000, the peak area range of a chromatograph might be 0 to 100, and the ion intensity range of a mass spectrometer might be 10^3 to 10^5. Min-max normalization scales all variables to the range of 0 to 1.

[0056] For example, a spectrometer light intensity value of 500 will be converted to 0.5. This standardized operation facilitates the comparison of characteristics of different instruments on the same scale, improving the accuracy of subsequent analyses.

[0057] Specifically, principal component analysis (PCA) is used for dimensionality reduction and key feature extraction. Assuming the comprehensive dataset contains 10 variables, including spectral intensity, chromatographic peak area, and mass spectrometry ion intensity, the correlation matrix reveals a high correlation between spectral intensity and chromatographic peak area, with a correlation coefficient of 0.9. PCA generates eigenvalues ​​and eigenvectors, selecting the top three principal components with a cumulative contribution rate of 85% to form the dimensionality-reduced feature dataset. This method effectively reduces data dimensionality while retaining key information related to arsenic concentration.

[0058] For example, feature selection focuses on features highly correlated with arsenic concentration. If analysis reveals that the absorption peak of the spectrometer at a specific wavelength has a correlation coefficient of 0.95 with arsenic concentration, while other features have correlation coefficients below 0.5, then this absorption peak feature is preferentially retained, and irrelevant variables are eliminated to form a streamlined feature dataset. This selection ensures that the features in the dataset are strongly correlated with the arsenic detection task.

[0059] In one possible implementation, data transformation maps features to a new space to optimize the analysis.

[0060] For example, kernel functions can be used to map a simplified feature dataset to a high-dimensional space, thereby enhancing the separability between features.

[0061] For example, the nonlinear relationship between the spectral absorption peak and arsenic concentration in the original features can be transformed into a linear relationship through kernel transformation, which facilitates subsequent modeling. This transformation enhances the expressiveness of the features.

[0062] Specifically, the results were validated by calculating the variance explained rate to assess the quality of the feature dataset. Assuming the final feature dataset has a variance explained rate of 90%, this indicates that the extracted features effectively captured the information related to arsenic concentration in the original data, meeting the detection requirements.

[0063] For example, in one experiment, the three retained principal components explained 95% of the variance, validating the effectiveness of the feature extraction. This validation ensures that the dataset is suitable for arsenic concentration analysis tasks.

[0064] In one possible implementation, the feature selection method can be extended if the feature dataset dimension is not as expected.

[0065] For example, by combining mutual information methods to further screen secondary features related to arsenic concentration, such as secondary ion intensity features from mass spectrometry (correlation coefficient 0.7), these features were included in the simplified feature dataset. This extended approach enriches the feature diversity based on the core approach, ensuring the comprehensiveness of the analysis.

[0066] Step S105: Based on the feature dataset, construct a three-dimensional spatial interpolation model to calculate the spatial distribution of arsenic concentration in the food sample, and obtain a three-dimensional concentration distribution model.

[0067] A feature dataset is obtained by extracting arsenic concentration measurement data from rice samples, resulting in the original measurement dataset. Based on the original measurement data, a three-dimensional spatial coordinate system is constructed to determine the position of each data point in the three-dimensional space of rice cultivation, resulting in a spatial coordinate dataset. The Kriging interpolation algorithm is used to interpolate the spatial coordinate dataset of rice cultivation, generating a continuous distribution function of arsenic concentration, resulting in an interpolated distribution model. Arsenic concentration values ​​are extracted from the interpolated distribution model, and the concentration values ​​at each point in the three-dimensional space of rice cultivation are calculated, resulting in a three-dimensional concentration dataset. Stereoscopic microscopy is used to visualize the three-dimensional concentration dataset, generating a three-dimensional concentration distribution model. If the resolution of the three-dimensional concentration distribution model is lower than a preset threshold, the model parameters are adjusted using a mesh optimization algorithm to obtain an optimized three-dimensional concentration distribution model.

[0068] For example, when extracting arsenic concentration measurement data from rice samples, the original measurement dataset can be obtained through laboratory analysis. Assuming 100 rice samples are collected in a rice-growing area, the arsenic concentration of each sample is determined by atomic absorption spectrometry, resulting in a dataset with concentration values ​​ranging from 0.01 to 0.5 mg / kg. This data includes sample numbers, sampling point latitude and longitude, and arsenic concentration values, constituting the original measurement dataset. Data collection must ensure sample representativeness, for example, covering planting areas with different soil types and irrigation conditions to reflect spatial differences in arsenic concentration. In this invention, using the data in the aforementioned feature dataset as a standard, data corresponding to the feature datasets in different rice samples are collected. A deep learning model is used to establish the relationship between the feature datasets and arsenic concentrations, and trained using historical sample data to directly obtain the data corresponding to the feature datasets. The trained deep learning model then obtains the arsenic concentration measurement data.

[0069] In one possible implementation, when constructing a three-dimensional spatial coordinate system, the rice planting area can be divided into a three-dimensional grid. Assuming the planting area is a field 100 meters long, 100 meters wide, and 1 meter high, the location of the sampling points is determined by latitude, longitude, and soil depth.

[0070] For example, a sample point with coordinates (50, 30, 0.2) represents a location at a soil depth of 0.2 meters, with horizontal coordinates of (50 meters, 30 meters). Using Geographic Information System (GIS) software, the coordinates of all sampling points are mapped to three-dimensional space, generating a spatial coordinate dataset. This dataset provides the spatial location basis for subsequent interpolation calculations.

[0071] Specifically, when using the Kriging interpolation algorithm to generate a continuous distribution function for arsenic concentration, the interpolation model can be calculated based on a spatial coordinate dataset. The Kriging algorithm estimates the concentration values ​​of unsampled points by analyzing the spatial correlation between sample points.

[0072] For example, assuming the arsenic concentrations at sample points A(10,10,0.1) and B(20,10,0.1) are known to be 0.2 and 0.3 mg / kg, respectively, an algorithm calculates the spatial variability function between the two points, generating an interpolation distribution model covering the entire region. This model can predict the concentration value at any point, such as (15,10,0.1), generating a continuous concentration distribution.

[0073] For example, when extracting a three-dimensional concentration dataset, the arsenic concentration value for each grid point can be calculated from an interpolation distribution model. Assuming the grid is divided into 1-meter × 1-meter × 0.1-meter cells, the model estimates the concentration value for each grid point, forming a three-dimensional concentration dataset.

[0074] For example, the concentration value at a grid point (50, 50, 0.3) is 0.25 mg / kg. This dataset reflects the spatial distribution characteristics of arsenic within the planting area.

[0075] In one possible implementation, when using stereomicroscopy to visualize the three-dimensional concentration distribution, a concentration distribution model can be generated using 3D rendering software. Assuming a color gradient is used to represent high and low concentrations, with red representing a high concentration of 0.5 mg / kg and blue representing a low concentration of 0.01 mg / kg, an intuitive 3D distribution map is generated. This model is convenient for analyzing high-risk areas of arsenic concentration, such as areas near irrigation water sources where concentrations are higher.

[0076] Specifically, if the resolution of the three-dimensional concentration distribution model is lower than a preset threshold, such as a grid resolution of less than 0.5 meters, the model parameters can be adjusted through a grid optimization algorithm.

[0077] For example, by employing adaptive mesh refinement technology, the mesh density in high-concentration areas is increased to 0.2 meters, enhancing the model's detail representation. The optimized model can more accurately depict spatial variations in arsenic concentration, such as concentration gradient changes in the vicinity of a pollution source.

[0078] It should be noted that the above method ensures the comprehensive extraction and intuitive presentation of arsenic concentration characteristics through a complete process from data acquisition to visualization.

[0079] For example, the optimized three-dimensional concentration distribution model can be used for agricultural management of rice to detect arsenic, identify high-risk areas, and take targeted control measures.

[0080] Step S106: Obtain concentration gradient change data from the three-dimensional concentration distribution model, use the gradient descent algorithm to analyze the dynamic distribution law of arsenic concentration, and generate a distribution law dataset.

[0081] The spatial coordinate parameters and time series data of rice cultivation are obtained from a three-dimensional concentration distribution model. The gradient value of arsenic concentration at each coordinate point is calculated to obtain concentration gradient data. Based on the concentration gradient data, the gradient descent algorithm is used to optimize the parameters of arsenic concentration distribution by setting the learning rate parameter and the number of model iterations, resulting in optimized distribution parameters. The trend of arsenic concentration change in the time series data is analyzed using the optimized distribution parameters to obtain the concentration change trend. If the fluctuation range of the concentration change trend is greater than a preset threshold, the learning rate parameter is adjusted, and the gradient descent algorithm is re-executed to obtain new optimized distribution parameters. Based on the new optimized distribution parameters and the spatial coordinate parameters of rice cultivation, the dynamic distribution law of arsenic concentration is generated, resulting in the dynamic distribution law. Using the dynamic distribution law, the concentration distribution characteristics of each time point in the time series data are extracted to generate a distribution law dataset. The convergence condition of the optimization is verified using the distribution law dataset. If it is not met, the learning rate parameter is adjusted, and the distribution law dataset is regenerated.

[0082] For example, when obtaining spatial coordinate parameters and time series data of rice cultivation from a three-dimensional concentration distribution model, the spatial coordinate points and corresponding time point arsenic concentration values ​​can be extracted from the model through a data interface. Assuming the rice cultivation area is a cube with dimensions of 100 meters on each side, the coordinate points are recorded in meters, such as (10, 20, 30) representing the location of a point in space. Time series data can be collected daily, such as arsenic concentration values ​​for 30 consecutive days, in milligrams per kilogram (mg / kg). After extraction, the data format might be (10, 20, 30, t1, 0.05), indicating that the concentration at a point on day 1 is 0.05 mg / kg. This method ensures the integrity of both spatial and temporal data, facilitating subsequent analysis.

[0083] In one possible implementation, the arsenic concentration gradient value can be estimated by the concentration difference between adjacent coordinate points when calculating the arsenic concentration gradient value.

[0084] For example, the concentration at coordinates (10, 20, 30) is 0.05 mg / kg, and the concentration at the adjacent point (11, 20, 30) is 0.06 mg / kg. The gradient along the x-axis is approximately (0.06 - 0.05) / 1 = 0.01 mg / kg / m. Similarly, the gradients along the y-axis and z-axis are calculated to obtain a three-dimensional gradient vector. This method intuitively reflects the direction and magnitude of concentration variation in space, providing a foundation for optimizing distribution.

[0085] Specifically, when using the gradient descent algorithm to optimize the distribution parameters, a learning rate of 0.01 and 1000 iterations can be set. Gradient descent iteratively adjusts the parameters to minimize the error between the concentration values ​​predicted by the model and the actual measured values.

[0086] For example, the initial distribution parameters might assume that the concentration varies linearly with altitude, but through gradient descent, they can be optimized to a more complex nonlinear distribution pattern. This optimization can more accurately describe the spatial distribution of arsenic concentration.

[0087] For example, when analyzing concentration trends in time-series data, the mean and standard deviation of concentration at each coordinate point over 30 days can be calculated. Suppose the concentration at a certain point fluctuates from 0.05 mg / kg to 0.07 mg / kg, with a standard deviation of 0.015, exceeding a preset threshold of 0.01, indicating significant fluctuation. After adjusting the learning rate to 0.005, gradient descent is run again to obtain more stable distribution parameters. This method effectively captures the patterns of concentration changes over time.

[0088] In one possible implementation, when generating dynamic distribution patterns, the optimized parameters and spatial coordinates can be combined to construct a four-dimensional dataset of time-space-concentration.

[0089] For example, if the concentration at a certain point is 0.05 mg / kg on day 1 and 0.06 mg / kg on day 30, the dynamic pattern shows that the concentration increases slowly over time. This pattern can be used to predict future concentration trends.

[0090] Specifically, when extracting distribution features from time series data, the mean concentration, variance, and spatial correlation at each time point can be calculated.

[0091] For example, the mean concentration on day 10 was 0.055 mg / kg with a variance of 0.002, and spatial correlation showed that the concentration was higher at lower altitudes. This feature extraction helps to understand the spatiotemporal characteristics of arsenic distribution.

[0092] For example, when verifying the convergence condition of the optimization, one can check whether the value of the gradient descent loss function is lower than a preset threshold, such as 0.001. If convergence is not achieved, the learning rate is adjusted to 0.002, and a new dataset with a regular distribution is generated. This verification ensures the reliability and stability of the model.

[0093] Step S107: Based on the distribution pattern dataset, identify high-value clusters of arsenic concentration and generate a distribution map of the clusters.

[0094] Arsenic concentration data and location coordinates within the rice planting area were obtained from the dataset to generate a spatial distribution set containing the data points. K-means clustering was used to cluster the data points in the spatial distribution set, resulting in a preliminary division of high-value regions. If the arsenic concentration value of a data point was greater than a preset concentration threshold, it was marked as a high-value region, resulting in a high-value region set. Density analysis was performed on the high-value region set, and kernel density estimation was used to determine the boundary range of the clustered regions. Based on the boundary range of the clustered regions, a spatially distributed grid data was generated, resulting in gridded concentration distribution data. A rendering tool was used to render the gridded concentration distribution data, generating a distribution map of the high-arsenic concentration clustered regions. Finally, a graphical visualization technique was used to output the distribution map as a visual image, resulting in the final clustered region distribution map.

[0095] For example, when acquiring arsenic concentration data and location coordinates within a rice paddy's spatial distribution, data can be extracted from the paddy field's soil monitoring system. Suppose a paddy field is divided into 100 grid points, each recording its latitude and longitude coordinates and the corresponding arsenic concentration value, such as an arsenic concentration of 0.5 mg / kg at coordinates (10, 20). A spatial distribution dataset can be constructed from these data points, containing the coordinates and concentration value of each point, forming a three-dimensional dataset.

[0096] It should be noted that the accuracy of the data depends on the sampling density and the precision of the monitoring equipment. Therefore, evenly distributed sampling points are usually selected to ensure coverage of the entire planting area.

[0097] In one possible implementation, when using the K-means clustering algorithm to process the spatial distribution set, K can be set to 3, representing low, medium, and high concentration regions respectively. The algorithm will automatically cluster adjacent points with similar concentrations based on the concentration values ​​and spatial distances of the data points.

[0098] For example, several points located in the southeast corner of the paddy field were classified as high-concentration areas due to their high concentration values, such as above 0.6 mg / kg. This method can quickly identify potentially high-risk areas, facilitating subsequent analysis.

[0099] Specifically, when marking high-value areas, assuming a preset concentration threshold of 0.4 mg / kg, points with concentrations higher than this value are marked as high-value area data points.

[0100] For example, points at coordinates (15,25) and (20,30) were marked with concentrations of 0.55 and 0.65 mg / kg, respectively. These points form a set of high-value areas, reflecting the distribution of regions with more severe arsenic pollution.

[0101] In one embodiment, the kernel density estimation method is used to analyze the density distribution of a set of high-value regions. The density value at each point can be calculated by setting an appropriate bandwidth parameter, such as 0.5 meters, thereby determining the boundaries of the high-concentration regions.

[0102] For example, multiple high-value points in the southeast corner may form an elliptical cluster area, with a boundary covering approximately 50 square meters. This method helps to accurately delineate contaminated areas, facilitating targeted remediation.

[0103] For example, when generating gridded concentration distribution data, the paddy field can be divided into a 1-meter × 1-meter grid. Each grid is interpolated based on the concentration values ​​of the surrounding data points to obtain a smooth concentration distribution.

[0104] For example, the concentration value at the center of a grid might be estimated as 0.45 mg / kg using a weighted average of neighboring points. This gridded data provides the basis for subsequent visualization.

[0105] In one possible implementation, rendering tools such as Geographic Information System (GIS) software are used to process the gridded data to generate a distribution map of areas with high arsenic concentrations. Concentration differences can be represented by color gradients, such as red indicating concentrations above 0.5 mg / kg and yellow indicating concentrations between 0.3 and 0.5 mg / kg. The final output visualization clearly shows pollution hotspots; for example, a red area in the southeast corner indicates a high risk of pollution. This visualization method is intuitive and helps decision-makers quickly identify problem areas and take appropriate action.

[0106] Step S108: Obtain pollution diffusion path characteristics from the cluster area distribution map, use cluster analysis algorithm to determine the location of pollution sources, and generate a pollution source distribution dataset.

[0107] Pollution area distribution data is obtained from environmental monitoring data. Noise is removed using preprocessing methods to obtain a clean dataset. Pollutant concentration values ​​are extracted from the clean dataset, and spatial distribution characteristics are analyzed using kernel density estimation to determine the extent of the polluted area. Based on the extent of the polluted area, K-means clustering is used to analyze geographical location information and extract pollution diffusion path characteristics, resulting in a diffusion path set. If the pollutant concentration value of at least one path in the diffusion path set is higher than a preset threshold, a reverse tracing algorithm is used to analyze time series changes to determine the pollution source location set. Based on the pollution source location set, kernel density estimation is used to calculate the pollution source distribution probability, generating a pollution source distribution dataset. The correlation between spatial distribution characteristics and time series changes is analyzed using the pollution source distribution dataset to obtain the pollution diffusion trend. Based on the pollution diffusion trend, interpolation methods are used to predict the future distribution of pollutant concentration values, generating a prediction dataset.

[0108] For example, in the field of environmental monitoring, the process of obtaining data on the distribution of polluted areas can begin with raw data from monitoring stations. Raw data may include the concentration of pollutants in air, soil, or water, such as PM2.5 or heavy metal concentrations, along with their corresponding geographic coordinates and timestamps. Assuming monitoring stations cover a certain urban area, the data might contain PM2.5 concentration values ​​from 1000 stations, ranging from 10 to 200 micrograms per cubic meter. When preprocessing to remove noise, median filtering techniques can be used to eliminate outliers.

[0109] For example, if a site records a sudden spike in data to 500 micrograms per cubic meter, which is significantly outside the normal range, it can be identified as noise by comparing it with data from neighboring sites and removed to obtain a clean dataset.

[0110] In one possible implementation, after extracting pollutant concentration values, a kernel density estimation method is used to analyze spatial distribution characteristics. Kernel density estimation generates a smooth density distribution map by weighting the concentration values ​​around each data point.

[0111] For example, by taking a concentration value within a 5-kilometer radius of a specific site as the center, a density heat map is generated, showing that the polluted area is concentrated in the urban industrial zone, with a boundary covering approximately 20 square kilometers. This method can clearly reflect the spatial aggregation characteristics of pollution.

[0112] Specifically, when using the K-means clustering algorithm to analyze geographical location information, the sites can be divided into three clusters based on their coordinates and concentration values ​​to extract pollution diffusion path features.

[0113] For example, clusters near industrial areas show concentrations exceeding 100 micrograms per cubic meter, and path characteristics indicate that the pollutants are spreading northeastward along the wind direction, with the path set covering a range of 10 kilometers. If the concentration value along a certain path exceeds the threshold of 80 micrograms per cubic meter, a reverse tracing algorithm is triggered. By analyzing time-series data, such as concentration changes over the past 7 days, it is inferred that the pollution source may be located at a chemical plant, generating a set of pollution source locations.

[0114] For example, based on a set of pollution source locations, kernel density estimation methods can calculate the probability of pollution source distribution. Suppose three potential pollution sources are identified, located at a chemical plant, a power plant, and a transportation hub, respectively. The probability distribution shows that the chemical plant has the highest pollution contribution rate, reaching 60%. The pollution source distribution dataset is further combined with time series data to analyze the correlation between concentration values ​​and changes over time.

[0115] For example, peak emissions from chemical plants are positively correlated with increased concentrations in the northeastern part of the city, indicating that the trend of pollution spread is affected by wind direction and emission time.

[0116] In one possible implementation, interpolation methods can predict the future distribution of pollutant concentrations based on diffusion trends.

[0117] For example, using Kriging interpolation, concentration values ​​in unmonitored areas can be predicted based on existing site data, generating a predictive dataset. The prediction shows that concentrations around industrial areas may rise to 120 micrograms per cubic meter within the next three days. This method helps to provide early warning of highly polluted areas and offers data support for remediation efforts.

[0118] It should be noted that the implementation of the reverse tracing algorithm requires high-precision time series data.

[0119] For example, by combining hourly concentration changes with meteorological data such as wind speed and direction, the accuracy of inferring the location of pollution sources can be improved to over 90%. This method is particularly effective for areas with complex terrain. The generation of the prediction dataset depends on the appropriate selection of interpolation methods.

[0120] For example, Kriging interpolation is better able to adapt to spatial heterogeneity, generating more accurate predictions. The combination of these techniques forms a chain of pollution distribution analysis, providing comprehensive data support for environmental management.

[0121] Step S109: Based on the pollution source distribution dataset and combined with the three-dimensional concentration distribution model, calculate the risk assessment parameters to obtain the risk assessment results of arsenic pollution in food.

[0122] A dataset of pollution source distribution was acquired, and a pre-defined sampling point gridding method was used to determine the spatial distribution characteristics of the pollution sources. Using a three-dimensional concentration distribution model, the concentration distribution of arsenic in food was calculated based on these spatial characteristics. If the concentration distribution value exceeded a pre-defined threshold, a boundary decision algorithm was used to classify high-risk areas, resulting in high-risk area delineation results. Based on the high-risk area delineation results and the spatial distribution of food production areas, the potential risk level of arsenic contamination in food was determined. Monte Carlo simulation was used to calculate risk assessment parameters for the potential risk level, obtaining a quantified risk value. Based on the quantified risk value and combined with food type and consumption data, the final risk assessment result of arsenic contamination in food was determined. Based on the final risk assessment result, a spatial distribution map of arsenic contamination risk was generated using Geographic Information System (GIS) technology, resulting in a visualized risk distribution.

[0123] For example, in the field of environmental monitoring, when acquiring a dataset of pollution source distribution, a pre-defined sampling point gridding method can be used.

[0124] Specifically, the target area is divided into a uniform grid, for example, an agricultural area divided into a 1 km × 1 km grid. A sampling point is set at the center of each grid to collect arsenic concentration data in soil samples. Assuming 1000 sampling points are collected, the arsenic concentration ranges from 0.01 to 5 mg / kg. The gridding method, through systematic sampling, ensures coverage of the entire area and reflects the spatial distribution characteristics of arsenic. This method is convenient for identifying concentrated areas of pollution sources; for example, a grid near a mining area showing arsenic concentrations higher than 2 mg / kg indicates the potential presence of a pollution source.

[0125] In one possible implementation, a three-dimensional concentration distribution model is constructed based on sampling point data to analyze the spatial distribution characteristics of arsenic.

[0126] For example, by combining geographic coordinates and concentration values, a three-dimensional model can be generated to show the variation of arsenic concentration with spatial altitude and horizontal distribution. The model can reveal that the concentration is high in the surface soil near the mining area, while the concentration decreases in deeper layers, indicating that the pollution source mainly affects the surface soil. If the threshold is set at 1.5 mg / kg, the model can identify areas with excessive concentrations, such as within a 5-kilometer radius of the mining area.

[0127] Specifically, for areas exceeding the standard, a boundary decision algorithm is used to delineate high-risk areas.

[0128] For example, by analyzing concentration gradients using algorithms, areas with concentrations higher than 1.5 mg / kg were designated as high-risk zones, covering approximately 10 square kilometers. This classification method, through spatial clustering, accurately locates risk boundaries, which is helpful for subsequent risk management.

[0129] For example, by combining the spatial distribution of high-risk areas and food production areas, the potential risk level of arsenic contamination in food can be determined. Assuming that the agricultural area contains rice and wheat growing regions, after overlaying the high-risk areas, it is found that the rice growing areas overlap significantly with the high-risk areas, indicating a high potential risk level. Analysis shows that soil arsenic can enter rice plants through irrigation water, increasing the risk of contamination.

[0130] In one possible implementation, the Monte Carlo simulation method is used to quantify the risk. Assuming 10,000 simulations are run, input variables include soil arsenic concentration, irrigation water uptake rate, and crop enrichment coefficient, risk parameters are calculated.

[0131] For example, simulation results show that the probability of arsenic concentration exceeding the standard in rice is 70%, with a quantified risk value of high. This method improves the reliability of the assessment by comprehensively considering uncertainties through stochastic simulation.

[0132] Specifically, the final risk assessment result is determined based on quantitative risk values ​​and food consumption data.

[0133] For example, with an annual rice consumption of 100 kg per person, combined with the probability of arsenic concentration exceeding the standard, the assessment results indicate that long-term consumption may pose health risks. The risks are more pronounced in regions with high consumption levels.

[0134] For example, geographic information system (GIS) technology can be used to generate a spatial distribution map of arsenic pollution risk. Based on quantified risk values, a visual map is created, showing that rice-growing areas surrounding the mining area are designated as high-risk (red) zones, while wheat-growing areas on the outskirts are designated as medium-risk (yellow) zones. The map visually presents the risk distribution, facilitating the development of targeted mitigation measures. This visualization method helps decision-makers quickly identify key areas for prevention and control.

[0135] like Figure 2 As shown, this invention provides an arsenic distribution modeling system based on multi-source data fusion, which mainly includes:

[0136] The data acquisition module is used to acquire data from multi-source detection devices through a pre-established standardized access protocol, unify the data format and timestamp, and generate a multi-source dataset with a consistent format.

[0137] The time synchronization module is used to correct time synchronization problems based on multi-source datasets using a time series alignment algorithm, and obtain a time-synchronized dataset.

[0138] The data fusion module is used to acquire detection data from spectrometers, chromatographs, and mass spectrometers from the time-synchronized dataset and fuse them to generate a comprehensive dataset.

[0139] The feature extraction module is used to extract the main features of arsenic concentration from the comprehensive dataset using principal component analysis algorithm, and generate a feature dataset.

[0140] The spatial modeling module is used to construct a three-dimensional spatial interpolation model based on the feature dataset, calculate the spatial distribution of arsenic concentration in food samples, and obtain a three-dimensional concentration distribution model.

[0141] The dynamic analysis module is used to obtain concentration gradient change data from the three-dimensional concentration distribution model, and to analyze the dynamic distribution law of arsenic concentration using the gradient descent algorithm to generate a distribution law dataset.

[0142] The region identification module is used to identify high-value clusters of arsenic concentration based on the distribution pattern dataset and generate a distribution map of the clustered regions.

[0143] The pollution source tracing module is used to obtain pollution diffusion path characteristics from the distribution map of clustered areas, use cluster analysis algorithms to determine the location of pollution sources, and generate a pollution source distribution dataset.

[0144] The risk assessment module is used to calculate risk assessment parameters based on the pollution source distribution dataset and a three-dimensional concentration distribution model, thereby obtaining the risk assessment results of arsenic contamination in food.

[0145] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for modeling arsenic distribution based on multi-source data fusion, characterized in that, include: Acquire multi-source detection data, including arsenic detection data in food from different devices. Synchronize the multi-source detection data over time and fuse the synchronized data to obtain a comprehensive dataset. Based on the comprehensive dataset, extract features from the multi-source detection data using principal component analysis to obtain a feature dataset. Based on the feature dataset, construct a three-dimensional spatial interpolation model and calculate the arsenic concentration in the three-dimensional spatial interpolation model for spatial analysis to obtain a three-dimensional concentration distribution model. Analyze the concentration gradient change data in the three-dimensional concentration distribution model to obtain concentration distribution pattern data. Perform spatial clustering based on the concentration distribution pattern data to obtain a clustering region distribution map. Based on the clustering region distribution map, generate diffusion paths. Perform cluster analysis based on the diffusion paths to obtain pollution source distribution data. Based on the pollution source distribution data and the three-dimensional concentration distribution model, obtain the risk assessment result of arsenic contamination in food.

2. The method according to claim 1, characterized in that, The process of time synchronization of the multi-source detection data includes: Time series data and corresponding timestamps are obtained from the multi-source detection data. Based on the timestamps, the multi-source detection data is time-aligned using a linear interpolation method to obtain preliminary aligned data. The preliminary aligned data is then offset-corrected using a dynamic time correction algorithm to obtain time-synchronized data.

3. The method according to claim 1, characterized in that, The process of merging time-synchronized data includes: The corresponding time-synchronized data from the spectrometer, chromatograph, and mass spectrometer are obtained to obtain the original dataset. The arsenic concentrations detected in the original dataset are then fused using a weighted average method to generate a comprehensive dataset.

4. The method according to claim 1, characterized in that, The process of feature extraction from multi-source detection data using principal component analysis includes: The comprehensive dataset is cleaned and standardized to obtain a standardized dataset. Based on the standardized dataset, principal component analysis is applied to the time-synchronized data to calculate the correlation matrix between variables in the time-synchronized data, obtaining eigenvalues ​​and eigenvectors. The eigenvalues ​​and eigenvectors are then thresholded to obtain a dimensionality-reduced feature dataset. The feature dataset is then subjected to correlation analysis and sorting. Based on the sorting results, the eigenvalues ​​and eigenvectors are further filtered to obtain a simplified feature dataset. The simplified feature dataset is then validated to obtain the final feature dataset.

5. The method according to claim 1, characterized in that, The process of obtaining the three-dimensional concentration distribution model includes: Based on the feature dataset, measurement data of arsenic concentration in food is obtained. Based on the measurement data, a three-dimensional spatial coordinate system is constructed to determine the position of the measurement data in the three-dimensional space, thus obtaining a spatial coordinate dataset. The three-dimensional spatial coordinate system is interpolated using the Kriging interpolation method, and the continuous distribution function of arsenic concentration is calculated to obtain an interpolation distribution model. The arsenic concentration is extracted from the interpolation distribution model, and the arsenic concentration value at each point in the three-dimensional space is calculated to obtain a three-dimensional concentration dataset. The three-dimensional concentration dataset is then visualized to generate a three-dimensional concentration distribution model.

6. The method according to claim 1, characterized in that, The process of obtaining concentration distribution data includes: In the three-dimensional concentration distribution model, spatial coordinates and arsenic concentration time series data are obtained. The gradient value of arsenic concentration at each coordinate point is calculated based on the spatial coordinates and arsenic concentration time series data to obtain concentration gradient data. The concentration gradient data is optimized using the gradient descent method to obtain optimized distribution parameters. The changing trend of arsenic concentration in the arsenic concentration time series is analyzed based on the optimized distribution parameters to obtain the concentration change trend. It is determined whether the concentration change trend needs to be re-optimized using the gradient descent method. Based on the optimized distribution parameters and combined with the spatial coordinate parameters, the dynamic distribution law data of arsenic concentration is obtained.

7. The method according to claim 1, characterized in that, The process of obtaining the cluster distribution map includes: Arsenic concentration data and spatial coordinates are extracted from the concentration distribution data to obtain a spatial distribution set. The data points in the spatial distribution set are clustered using a clustering method to obtain preliminary regional divisions. The arsenic concentration in the preliminary regional divisions is judged and marked to obtain a set of high-value regions. Density analysis is performed on the high-value region set to obtain the boundary range of the clustered regions. Based on the boundary range, the network data is divided to obtain gridded concentration distribution data. The gridded concentration distribution data is rendered and visualized to obtain a distribution map of arsenic concentration clustered regions.

8. The method according to claim 1, characterized in that, The process of obtaining pollution source distribution data includes: Environmental monitoring data is acquired, and pollution area distribution data is obtained based on the environmental monitoring data. The pollution area distribution data is preprocessed, and pollutant concentration values ​​are extracted from the preprocessed pollution area distribution data. The spatial distribution characteristics of pollutant concentration values ​​are analyzed using the kernel density estimation method to obtain the pollution area range. Cluster analysis is performed on the pollution area range to extract pollution diffusion path characteristics and obtain diffusion paths. The diffusion paths are judged, and the temporal changes of diffusion paths are analyzed using the reverse tracing method to obtain the location set of pollution sources. The pollution source distribution probability is calculated based on the location set of pollution sources using the kernel density estimation method. Based on the pollution source distribution probability, pollution source distribution data is obtained.

9. The method according to claim 1, characterized in that, The process of obtaining risk assessment results for arsenic contamination in food includes: The pollution source distribution data is divided into grids to obtain the spatial distribution characteristics of the pollution sources. The concentration of arsenic is analyzed through a three-dimensional concentration distribution model, and the concentration distribution of arsenic is judged. Based on the judgment results, boundary decisions are made for the concentration distribution of arsenic to obtain high-risk areas. According to the spatial distribution characteristics of high-risk areas and pollution sources, the potential risk level of food contamination is judged. The Monte Carlo simulation method is used to calculate risk assessment parameters based on the potential risk level, and the quantitative risk value, i.e., the risk assessment result of arsenic contamination in food, is obtained.

10. An arsenic distribution modeling system based on multi-source data fusion, characterized in that, Used to perform the method described in any one of claims 1-9.

Images