Three-dimensional visualization method and system for groundwater pollution based on artificial intelligence

Abstract

The present invention provides an artificial intelligence-based three-dimensional visualization method and system for groundwater pollution, which relates to the field of artificial intelligence technology. By obtaining a groundwater basic data set containing pollution monitoring information, a spatiotemporal feature is extracted using a spatiotemporal association learning model to generate a composite feature set reflecting the spatiotemporal evolution law of pollution. Furthermore, with the help of a three-dimensional visualization generation model, the composite feature set is mapped into a three-dimensional feature field of pollution distribution, and dynamic trend enhancement processing is performed to generate a dynamic three-dimensional visualization result containing pollution diffusion path information. Finally, the dynamic three-dimensional visualization result is output to a display device to present the spatial distribution and evolution process of groundwater pollution, thereby being able to accurately and intuitively display the three-dimensional spatial distribution and dynamic changes of groundwater pollution.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based three-dimensional visualization method and system for groundwater pollution. Background Art

[0002] As a vital water resource, groundwater pollution is a growing concern. Accurately understanding the spatial distribution and evolution of groundwater pollution is crucial for developing effective pollution prevention, control, and treatment measures. However, existing groundwater pollution monitoring and analysis methods have many limitations. Traditional methods typically rely on manual sampling and laboratory analysis, which is not only time-consuming and labor-intensive, but also makes it difficult to obtain continuous and comprehensive pollution data. Furthermore, while some technologies have begun to utilize tools such as geographic information systems (GIS) for two-dimensional visualization of groundwater pollution, these two-dimensional visualizations struggle to intuitively reflect the distribution and dynamic changes of pollution in three-dimensional space, and therefore cannot meet the needs for a deeper understanding of the complex evolution of groundwater pollution. Summary of the Invention

[0003] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a three-dimensional visualization method for groundwater pollution based on artificial intelligence, the method comprising:

[0004] Acquire a groundwater basic data set containing pollution monitoring information, wherein the groundwater basic data set includes pollutant distribution information at different spatial locations and corresponding time identifiers;

[0005] Extracting spatiotemporal features from the groundwater basic data set using a spatiotemporal association learning model to generate a composite feature set reflecting the spatiotemporal evolution of pollution;

[0006] Performing spatial mapping processing on the composite feature set using a three-dimensional visualization generation model to generate a three-dimensional feature field of pollution distribution having a corresponding relationship of spatial coordinates;

[0007] Performing dynamic trend enhancement processing on the three-dimensional characteristic field of pollution distribution to generate a dynamic three-dimensional visualization result containing pollution diffusion path information;

[0008] The dynamic three-dimensional visualization result is output to a display device to present the spatial distribution and evolution process of groundwater pollution.

[0009] On the other hand, the present invention also provides an artificial intelligence-based three-dimensional visualization system for groundwater pollution, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0010] Based on the above aspects, the present invention obtains a groundwater basic data set containing pollution monitoring information and uses a spatiotemporal correlation learning model to extract spatiotemporal features, which can accurately capture the spatiotemporal evolution of pollution and generate a composite feature set that reflects the dynamic changes in pollution. Furthermore, the composite feature set is mapped into a three-dimensional feature field of pollution distribution with a spatial coordinate correspondence through a three-dimensional visualization generation model, realizing a three-dimensional visualization display of pollution distribution, making the spatial distribution of pollution clear at a glance. At the same time, dynamic trend enhancement processing is performed on the three-dimensional feature field of pollution distribution, which can clearly present information on the pollution diffusion path, significantly improving the accuracy and intuitiveness of groundwater pollution monitoring and analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a schematic diagram of the execution flow of the artificial intelligence-based three-dimensional visualization method for groundwater pollution provided by an embodiment of the present invention.

[0012] Figure 2 Schematic diagram of exemplary hardware and software components of an artificial intelligence-based three-dimensional visualization system for groundwater pollution provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 This is a flow chart of an artificial intelligence-based three-dimensional visualization method for groundwater pollution provided by an embodiment of the present invention. The following is a detailed introduction to the artificial intelligence-based three-dimensional visualization method for groundwater pollution.

[0014] Step S110: Acquire a groundwater basic data set containing pollution monitoring information, wherein the groundwater basic data set contains pollutant distribution information at different spatial locations and corresponding time identifiers.

[0015] This embodiment can collect a groundwater basic data set containing pollution monitoring information from multiple data sources. The data sources mainly include groundwater level monitoring wells, infiltration reaction wall monitoring points and natural spring monitoring points. Groundwater level monitoring wells are usually wells drilled vertically into the ground, with various types of sensors installed inside, which can monitor the groundwater level and the concentration of different pollutants in the water in real time. The sensors can automatically collect data at set time intervals and transmit the data to the data storage center. The infiltration reaction wall monitoring point is located near the infiltration reaction wall. The infiltration reaction wall is a facility used to treat groundwater pollution. Monitoring points are set before and after it to obtain changes in the distribution of pollutants in the groundwater before and after treatment. The natural spring monitoring point monitors naturally gushing springs, which can reflect the pollution status of the groundwater in the natural state of the area.

[0016] When collecting basic groundwater data, it is necessary to ensure that the data for each monitoring point includes the spatial coordinates, pollutant type, pollutant concentration, and monitoring time. Spatial coordinates are obtained through geographic information systems (GIS). GIS uses satellite positioning technology and geographic databases to accurately determine the three-dimensional coordinates of each monitoring point on the Earth's surface, such as longitude, latitude, and altitude. Monitoring time is recorded through a unified time synchronization system, such as the time signal of the Global Positioning System (GPS), to ensure that data from all monitoring points are consistent and comparable.

[0017] Next, the collected pollution monitoring data records are screened for validity. Pollution monitoring data may be affected by various factors during collection and transmission, resulting in inaccurate or incomplete data. For example, sensor failure may cause abnormal pollutant concentration measurements; communication line interruptions may result in partial data loss. To eliminate this invalid data, a series of screening rules are required. For example, a reasonable range of pollutant concentrations can be set. If the pollutant concentration at a monitoring point exceeds this range, the data is considered invalid. Data integrity can also be checked to eliminate data that lacks key information (such as spatial location coordinates or monitoring time).

[0018] Finally, the filtered pollution monitoring data records are integrated to generate a basic groundwater data set. The integration process involves organizing valid data from different monitoring points and at different times according to a predefined format, forming a set containing multiple data items. For example, the data from each monitoring point can be sorted by monitoring time, with each data item containing information such as spatial location coordinates, pollutant type, pollutant concentration, and monitoring time.

[0019] Step S120: extracting spatiotemporal features from the groundwater basic data set through a spatiotemporal association learning model to generate a composite feature set that reflects the spatiotemporal evolution of pollution.

[0020] In this embodiment, the spatiotemporal association learning model is an artificial intelligence model used to mine spatiotemporal features from groundwater basic data sets. This model primarily consists of a bidirectional attention module and a feature extraction layer. The bidirectional attention module calculates the interaction weights between the temporal and spatial dimensions, while the feature extraction layer filters out redundant information and extracts core features.

[0021] Step S121: extracting pollutant distribution information at the same spatial location under different time stamps from the groundwater basic data set, and constructing a pollution evolution sequence in the time dimension.

[0022] In this embodiment, taking groundwater level monitoring well A as an example, over a period of time, the monitoring well records pollutant distribution information at multiple time points. Assume that at time points t1, t2, t3...tn, monitoring well A records the concentrations of different pollutants. For a specific pollutant P, the concentration values ​​of pollutant P corresponding to the above time points are arranged in sequence to construct a pollution evolution sequence of pollutant P at this spatial location (monitoring well A) in the time dimension. The pollution evolution sequence can be represented as a vector, where each element of the vector corresponds to a pollutant concentration value at a time point. For example, the sequence C = [C1, C2, C3, ..., Cn], where Ci represents the concentration of pollutant P at time point ti.

[0023] Step S122: extracting pollutant distribution information at different spatial locations under the same time stamp from the groundwater basic data set, and constructing a pollution distribution matrix of spatial dimension.

[0024] In this embodiment, it is assumed that at time point t, multiple monitoring points (such as groundwater level monitoring well A, infiltration reaction wall monitoring point B, natural spring monitoring point C, etc.) record pollutant distribution information. For each pollutant, the pollutant concentration values ​​corresponding to these monitoring points are arranged according to their spatial positions to construct a pollution distribution matrix for the spatial dimension of the pollutant at time point t. For example, for pollutant P, assuming there are m monitoring points, then the pollution distribution matrix M can be represented as an m-row, 1-column matrix, where each row of the matrix corresponds to the concentration value of pollutant P at a monitoring point. That is, M=[M1, M2, M3, ..., Mm]T, where Mi represents the concentration of pollutant P at the i-th monitoring point at time point t.

[0025] Step S123: Input the pollution evolution sequence of the time dimension and the pollution distribution matrix of the space dimension into the bidirectional attention module of the spatiotemporal association learning model, and calculate the influence weight of the time dimension on the spatial dimension and the influence weight of the spatial dimension on the time dimension respectively through the bidirectional attention module.

[0026] In this embodiment, the bidirectional attention module consists of a time-to-space attention submodule and a space-to-time attention submodule.

[0027] Step S1231: In the time-to-space attention submodule of the bidirectional attention module, the degree of influence of the pollution evolution characteristics of each time point on the pollution distribution of all spatial positions is calculated to generate a time-to-space attention weight matrix.

[0028] In this embodiment, the role of the time-to-space attention submodule is to analyze how the pollution evolution characteristics in the time dimension affect the pollution distribution in the spatial dimension. Taking pollutant P as an example, for the pollution evolution sequence C in the time dimension and the pollution distribution matrix M in the spatial dimension, the time-to-space attention submodule will calculate the influence weight of the pollutant concentration at each time point on the pollutant concentration at all spatial locations. The specific calculation process is as follows: First, the pollution evolution sequence C in the time dimension and the pollution distribution matrix M in the spatial dimension are characterized to obtain their eigenvectors. Then, through a series of nonlinear transformations and calculations, the correlation score between the eigenvector of each time point and the eigenvector of all spatial locations is obtained. After these correlation scores are normalized, a time-to-space attention weight matrix is ​​generated. For example, assuming that there are n time points in the time dimension and m monitoring points in the spatial dimension, then the time-to-space attention weight matrix Wt-s is a matrix with n rows and m columns, where Wt-s(i, j) represents the influence weight of the pollution evolution characteristics at time point i on the pollution distribution at spatial position j.

[0029] Step S1232: In the space-to-time attention submodule of the bidirectional attention module, the degree of influence of the pollution distribution characteristics of each spatial position on the pollution evolution of all time points is calculated to generate a space-to-time attention weight matrix.

[0030] In this embodiment, the role of the space-to-time attention submodule is opposite to that of the time-to-space attention submodule. It analyzes how the pollution distribution characteristics in the spatial dimension affect the pollution evolution in the time dimension. Taking pollutant P as an example, for the pollution evolution sequence C in the time dimension and the pollution distribution matrix M in the spatial dimension, the space-to-time attention submodule will calculate the influence weight of the pollutant concentration at each spatial position on the pollutant concentration at all time points. The specific calculation process is similar to that of the time-to-space attention submodule. First, the characteristics of the time dimension and the spatial dimension are represented, and then the correlation score is calculated and normalized to finally obtain the space-to-time attention weight matrix. For example, the space-to-time attention weight matrix Ws-t is a matrix with m rows and n columns, where Ws-t(i, j) represents the influence weight of the pollution distribution characteristics of spatial position i on the pollution evolution at time point j.

[0031] Step S1233: Normalize the time-to-space attention weight matrix and the space-to-time attention weight matrix so that each weight value represents the relative influence strength between corresponding dimensions.

[0032] In this embodiment, the purpose of normalization is to ensure that each weight value in the attention weight matrix is ​​within a reasonable range and can accurately represent the relative influence strength between corresponding dimensions. Different normalization methods are used for the time-to-space attention weight matrix Wt-s and the space-to-time attention weight matrix Ws-t. For example, the softmax function can be used to normalize each row of the matrix so that the sum of the weight values ​​of each row is 1. After normalization, each weight value represents the relative importance between the corresponding dimensions.

[0033] Step S1234: Input the normalized time-to-space attention weight matrix and the space-to-time attention weight matrix into the fusion unit of the bidirectional attention module, and generate a comprehensive attention weight tensor that simultaneously includes the time-space bidirectional influence relationship through matrix dot multiplication operation.

[0034] In this embodiment, the role of the fusion unit is to fuse the time-to-space and space-to-time attention weight matrices to obtain a comprehensive attention weight tensor. The specific operation is to perform matrix dot multiplication on the normalized time-to-space attention weight matrix Wt-s and the space-to-time attention weight matrix Ws-t. Matrix dot multiplication is to multiply the elements of the corresponding positions of the two matrices to obtain a comprehensive attention weight tensor, which also includes the two-way influence relationship between the time dimension and the space dimension. For example, the comprehensive attention weight tensor T = Wt-s⊙Ws-t, where ⊙ represents the matrix dot multiplication operation.

[0035] Step S124: performing weighted fusion processing on the pollution evolution sequence in the time dimension and the pollution distribution matrix in the space dimension based on the influence weight, and generating an associated feature set including both time evolution features and space distribution features.

[0036] In this embodiment, the comprehensive attention weight tensor T is used to perform weighted fusion of the pollution evolution sequence C of the time dimension and the pollution distribution matrix M of the space dimension. For example, the specific method is to perform weighted summation of the comprehensive attention weight tensor T with the eigenvector of the time dimension and the eigenvector of the space dimension respectively. For example, for the eigenvector of the time dimension, each column of the comprehensive attention weight tensor T is weightedly summed with the pollution evolution sequence C of the time dimension to obtain a new eigenvector of the time dimension. For the eigenvector of the space dimension, each row of the comprehensive attention weight tensor T is weightedly summed with the pollution distribution matrix M of the space dimension to obtain a new eigenvector of the space dimension. Then, the new eigenvector of the time dimension and the new eigenvector of the space dimension are spliced ​​to generate an associated feature set that contains both the time evolution feature and the spatial distribution feature.

[0037] Step S125: performing redundant information filtering on the associated feature set through the feature extraction layer of the spatiotemporal association learning model, and retaining the composite feature set that reflects the evolution law of the pollution core.

[0038] In this embodiment, the feature extraction layer is used to remove redundant information from the associated feature set and extract features that reflect the evolution of the pollution core. The feature extraction layer can use a deep learning model such as a convolutional neural network (CNN) or an autoencoder. Taking a convolutional neural network as an example, the associated feature set is input into the input layer of the convolutional neural network, and the features are extracted and reduced in dimension through multiple convolutional layers and pooling layers. The convolution layer slides the convolution kernel across the feature map to extract local features; the pooling layer downsamples the feature map to reduce the dimensionality of the features. After multiple convolution and pooling operations, the features are finally mapped to a low-dimensional space through a fully connected layer, resulting in a composite feature set that reflects the evolution of the pollution core.

[0039] Step S130: performing spatial mapping processing on the composite feature set using a three-dimensional visualization generation model to generate a three-dimensional feature field of pollution distribution having a corresponding relationship of spatial coordinates.

[0040] In this embodiment, a 3D visualization generation model is used to map a composite feature set into 3D space, generating a 3D feature field of pollution distribution with spatial coordinate correspondence. The model primarily consists of a spatial encoding unit, an interpolation prediction unit, a feature enhancement unit, and a spatial coordinate calibration unit.

[0041] Step S131: inputting the composite feature set into the spatial coding unit of the three-dimensional visualization generation model, and constructing a three-dimensional grid coordinate system based on the spatial position information in the groundwater basic data set through the spatial coding unit.

[0042] In this embodiment, the role of the spatial coding unit is to construct a three-dimensional grid coordinate system based on the spatial position information in the groundwater basic data set. First, the spatial position coordinates of all monitoring points, including longitude, latitude and altitude, are extracted from the groundwater basic data set. Then, the maximum and minimum values ​​of these coordinates are determined to determine the range boundary of the three-dimensional space. Next, the grid units are divided according to the set resolution within the spatial range to form a regularly arranged three-dimensional grid point set. Each grid point has a corresponding spatial coordinate, and the eigenvalues ​​in the composite feature set are mapped to these grid points through the spatial coding unit, completing the spatial coding process. For example, for a certain composite eigenvalue, it can be assigned to the corresponding grid point in the three-dimensional grid coordinate system according to its corresponding spatial position information.

[0043] Step S132: In the three-dimensional grid coordinate system, the interpolation prediction unit of the three-dimensional visualization generation model is used to infer pollutant distribution information for spatial locations that are not directly monitored, and to generate continuous pollution distribution data covering the entire monitoring area.

[0044] In this embodiment, the purpose of the interpolation prediction unit is to infer pollutant distribution information for spatial locations that are not directly monitored in the three-dimensional grid coordinate system, so as to generate continuous pollution distribution data covering the entire monitoring area.

[0045] Step S1321: Determine the spatial range boundary of the three-dimensional grid coordinate system, where the spatial range boundary is determined by the spatial position extreme values ​​of all monitoring points in the groundwater basic data set.

[0046] In this embodiment, the spatial coordinates of all monitoring points in the groundwater basic data set are analyzed to find the maximum and minimum values ​​of longitude, latitude, and altitude. These maximum and minimum values ​​determine the spatial range boundaries of the three-dimensional grid coordinate system. For example, the minimum value of longitude is xmin and the maximum value is xmax; the minimum value of latitude is ymin and the maximum value is ymax; the minimum value of altitude is zmin and the maximum value is zmax. Therefore, the spatial range boundary of the three-dimensional grid coordinate system is a rectangular space defined by (xmin, ymin, zmin) and (xmax, ymax, zmax).

[0047] Step S1322: Divide the spatial range into grid units according to a preset spatial resolution to generate a regularly arranged three-dimensional grid point set.

[0048] In this embodiment, the preset spatial resolution determines the size of the three-dimensional grid cell. For example, assuming the spatial resolution is Δx, Δy, and Δz, then in the longitude direction, a grid cell is divided every Δx; in the latitude direction, a grid cell is divided every Δy; and in the altitude direction, a grid cell is divided every Δz. In this way, a regularly arranged set of three-dimensional grid points is formed within the three-dimensional space. Each grid point has a unique coordinate, which can be expressed as (xi, yj, zk), where i, j, and k represent the grid index in the longitude, latitude, and altitude directions, respectively.

[0049] Step S1323: For each three-dimensional grid point that is not directly monitored, extract the pollutant distribution information and spatial location information of the monitored points within a preset neighborhood around the three-dimensional grid point.

[0050] In this embodiment, for grid points in the three-dimensional grid point set that are not directly monitored, it is necessary to extract information about monitored points within a preset neighborhood around them. The preset neighborhood can be a cubic area centered on the grid point, for example, a cubic neighborhood with a radius of r centered on the grid point (xi, yj, zk). Within this neighborhood, the spatial coordinates of all monitored points and the corresponding pollutant distribution information are found. For example, assuming there are n monitored points in the neighborhood, the information of each monitored point can be represented as a vector containing information such as the spatial coordinates and the pollutant concentration.

[0051] Step S1324: Call the weighted average model of the interpolation prediction unit to calculate the pollutant distribution information of the monitored points in the surrounding preset neighborhood. The weighted average model generates a weight coefficient based on the spatial distance between the grid point and the monitored point. The closer the distance to the monitored point, the larger the corresponding weight coefficient.

[0052] In this embodiment, the weighted average model is the core algorithm of the interpolation prediction unit, which is used to predict the pollutant distribution information of the grid points that are not directly monitored based on the information of the monitored points. The specific method is to generate a weight coefficient based on the spatial distance between the grid point and the monitored point. The spatial distance can be calculated by the Euclidean distance formula, that is, for the grid point (xi, yj, zk) and the monitored point (xm, yn, zp), the spatial distance between them is d=√[(xi-xm)²+(yj-yn)²+(zk-zp)²]. Then, a weight coefficient is generated according to the spatial distance, for example, the weight coefficient w=1 / (d+ε), where ε is a very small positive number used to avoid the denominator being zero. The closer the monitored point is, the greater its weight coefficient is.

[0053] Step S1325: Based on the weight coefficient, the pollutant distribution information of the monitored points in the surrounding preset neighborhood is weighted and summed to obtain the predicted value of the pollutant distribution information of the three-dimensional grid points that are not directly monitored, and generate the continuous pollution distribution data covering the entire monitoring area.

[0054] In this embodiment, the pollutant distribution information for monitored points within a preset neighborhood is multiplied by the corresponding weight coefficient and then summed to obtain the predicted pollutant distribution information for the three-dimensional grid points that are not directly monitored. For example, for a grid point that is not directly monitored, there are n monitored points in its surrounding neighborhood, each with a pollutant concentration of Cm and a corresponding weight coefficient of wm. Then, the predicted pollutant concentration for that grid point, P, is calculated as P = ∑(wm*Cm), where m ranges from 1 to n. By performing this calculation for all grid points that are not directly monitored, continuous pollution distribution data covering the entire monitoring area is generated.

[0055] Step S133: inputting the continuous pollution distribution data into the feature enhancement unit of the three-dimensional visualization generation model, enhancing the transition boundary information of areas with different pollution levels through boundary detection operations, and generating an intermediate three-dimensional feature field containing concentration gradient features.

[0056] In this embodiment, the main function of the feature enhancement unit is to enhance the transition boundary information of areas with different pollution levels in the continuous pollution distribution data. The continuous pollution distribution data is processed using a boundary detection algorithm, such as the Canny edge detection algorithm. The Canny edge detection algorithm first performs Gaussian smoothing on the data to remove noise; then calculates the gradient amplitude and direction; then performs non-maximum suppression to remove non-edge points; and finally, through double threshold processing, determines the true edge points. After the boundary detection operation, the transition boundary information of areas with different pollution levels is enhanced, generating an intermediate three-dimensional feature field containing concentration gradient features. This intermediate three-dimensional feature field can more clearly display the changes in pollutant concentration.

[0057] Step S134: performing spatial coordinate calibration processing on the intermediate three-dimensional feature field to ensure that the spatial coordinates of each position in the intermediate three-dimensional feature field correspond one-to-one with the geographic coordinates of the actual monitoring area, thereby generating the pollution distribution three-dimensional feature field.

[0058] In this embodiment, the spatial coordinate calibration process is to ensure that the spatial coordinates in the intermediate three-dimensional feature field are consistent with the geographic coordinates of the actual monitoring area.

[0059] Step S1341: selecting a plurality of reference points with known actual geographic coordinates in the intermediate three-dimensional feature field, wherein the reference points are spatial position points directly monitored in the groundwater basic data set.

[0060] In this embodiment, multiple reference points with known actual geographic coordinates are selected from the intermediate 3D feature field. These reference points are spatial locations directly monitored in the groundwater basic data set. For example, groundwater level monitoring wells, permeability reaction wall monitoring points, and natural spring monitoring points are selected as reference points. The actual geographic coordinates of these reference points are known, and they also have corresponding locations in the intermediate 3D feature field.

[0061] Step S1342: extracting the model space coordinates and the corresponding actual geographic coordinates of the reference point in the intermediate three-dimensional feature field.

[0062] In this embodiment, for the selected reference points, it is necessary to extract their model space coordinates in the intermediate three-dimensional feature field and the corresponding actual geographic coordinates. Taking a certain reference point as an example, its model space coordinates in the intermediate three-dimensional feature field can be determined by the grid index or relative position relationship of the point in the intermediate three-dimensional feature field. For example, in the intermediate three-dimensional feature field, the reference point may be located in a specific grid cell, and its model space coordinates can be expressed as the center coordinates of the grid cell. Its actual geographic coordinates are obtained through positioning using a geographic information system (GIS), including information such as longitude, latitude, and altitude. The model space coordinates and actual geographic coordinates of all reference points are sorted to form two sets of coordinate data sets, one set is a model space coordinate set, and the other set is an actual geographic coordinate set, which facilitates subsequent coordinate conversion calculations.

[0063] Step S1343: establishing a conversion relationship between the model space coordinates and the actual geographic coordinates through a coordinate conversion algorithm, wherein the coordinate conversion algorithm includes a combination of translation transformation, rotation transformation and scaling transformation.

[0064] In this embodiment, the coordinate transformation algorithm aims to find a mathematical relationship that allows the model space coordinates in the intermediate three-dimensional feature field to be accurately converted to actual geographic coordinates. A translation transformation adjusts the coordinate system's origin, moving the model space coordinate origin to a position corresponding to the actual geographic coordinate origin. For example, assuming the model space coordinate origin is at one point and the actual geographic coordinate origin is at another, the model space coordinates can be translated by calculating the offset between the two origins. A rotation transformation adjusts the coordinate system's orientation so that the axis directions of the model space coordinates align with those of the actual geographic coordinates. This may require calculating the rotation angle based on the actual situation and rotating the model space coordinates using a rotation matrix. A scaling transformation adjusts the coordinate system's scale to ensure that the model space coordinates and actual geographic coordinates are consistent in all directions. The model space coordinates are scaled so that the unit length on each axis matches the unit length of the actual geographic coordinates. Combining the translation, rotation, and scaling transformations in a predetermined order creates a complete coordinate transformation algorithm. Through this coordinate transformation algorithm, the model space coordinates of the reference point are transformed and compared and adjusted with the actual geographic coordinates, and the transformation parameters are continuously optimized until the most suitable transformation relationship is found.

[0065] Step S1344: using the conversion relationship to convert the model space coordinates of all positions in the intermediate three-dimensional feature field to generate calibration space coordinates corresponding to the actual geographic coordinates.

[0066] In this embodiment, after obtaining the conversion relationship between the model space coordinates and the actual geographic coordinates, the conversion relationship is applied to all positions in the intermediate three-dimensional feature field. For each position in the intermediate three-dimensional feature field, its model space coordinates can be converted using the above-mentioned coordinate conversion algorithm. For example, for the model space coordinates (x_model, y_model, z_model) of a certain position, it is converted into actual geographic coordinates (x_actual, y_actual, z_actual) through operations such as translation, rotation, and scaling. The model space coordinates of all positions in the intermediate three-dimensional feature field are sequentially subjected to such conversion processing to generate a set of calibration space coordinates corresponding to the actual geographic coordinates. Each coordinate in the calibration space coordinate set accurately corresponds to a geographic location in the actual monitoring area.

[0067] Step S1345: verifying the matching accuracy between the calibration space coordinates and the actual geographic coordinates. When the matching accuracy meets the preset requirements, the calibrated intermediate three-dimensional feature field is determined as the pollution distribution three-dimensional feature field.

[0068] In this embodiment, in order to ensure the accuracy of the calibration space coordinates and the actual geographic coordinates, their matching accuracy needs to be verified. Some additional verification points can be selected, and the actual geographic coordinates of these verification points are known. The model space coordinates of these verification points in the intermediate three-dimensional feature field are converted into calibration space coordinates through a transformation relationship, and then the converted calibration space coordinates are compared with the actual geographic coordinates. The error between them is calculated, for example, by calculating the absolute value of the coordinate difference or the mean square error to measure the matching accuracy. The preset requirement is an error threshold set according to the actual application scenario and accuracy requirements. If the error between the calibration space coordinates and the actual geographic coordinates of the verification point is within the preset requirement range, it means that the matching accuracy meets the requirements. At this time, the calibrated intermediate three-dimensional feature field is determined as the pollution distribution three-dimensional feature field. The spatial coordinates of each position in the pollution distribution three-dimensional feature field correspond one-to-one to the geographic coordinates of the actual monitoring area, accurately reflecting the distribution of groundwater pollution in three-dimensional space.

[0069] Step S140: performing dynamic trend enhancement processing on the three-dimensional characteristic field of pollution distribution to generate a dynamic three-dimensional visualization result including pollution diffusion path information.

[0070] In this embodiment, the purpose of dynamic trend enhancement processing is to mine the diffusion trends and paths of pollutants in the three-dimensional characteristic field of pollution distribution and clearly display this information through visualization. This process mainly includes the steps of constructing a time series pollution distribution data set, mining pollution evolution patterns, determining pollution diffusion parameters, and performing visualization coding.

[0071] Step S141: extracting pollution distribution data at different time marks in the three-dimensional pollution distribution feature field, and constructing a time series pollution distribution data set.

[0072] In this embodiment, the three-dimensional characteristic field of pollution distribution contains pollutant distribution information at different time marks. The pollution distribution data corresponding to each time mark is extracted from the three-dimensional characteristic field of pollution distribution in chronological order. For example, at different time points t1, t2, t3, etc., the pollution distribution data at that time point is extracted respectively. These data include information such as the pollutant concentration at each spatial position. The pollution distribution data at these different time points are arranged in sequence to construct a time series pollution distribution data set. The pollution distribution data set can be regarded as a three-dimensional array, in which one dimension represents time and the other two dimensions represent spatial positions. Each time slice corresponds to the pollution distribution at a specific time point. Through the pollution distribution data set of the time series, the changes in the spatial distribution of pollutants at different times can be observed.

[0073] Step S142: performing evolution law mining on the pollution distribution data set of the time series, and identifying the migration trajectory of pollutants in three-dimensional space through a motion trajectory tracking algorithm.

[0074] In this embodiment, the motion trajectory tracking algorithm is used to identify the migration trajectory of pollutants in three-dimensional space from a time series pollution distribution data set.

[0075] Step S1421: selecting first pollution distribution data and second pollution distribution data with adjacent time tags in the pollution distribution data set of the time series.

[0076] In this embodiment, two sets of pollution distribution data at adjacent time stamps are selected from a time series pollution distribution data set. These are referred to as the first pollution distribution data and the second pollution distribution data. For example, the pollution distribution data at time point t1 is selected as the first pollution distribution data, and the pollution distribution data at time point t2 is selected as the second pollution distribution data, where t2 is immediately after t1. These two sets of data reflect the spatial distribution of pollutants at adjacent time points. By comparing the differences between them, the migration of pollutants can be analyzed.

[0077] Step S1422: identifying characteristic regions where pollutant concentration changes significantly in the first pollution distribution data, wherein the characteristic regions are continuous spatial regions where the pollutant concentration gradient exceeds a preset threshold.

[0078] In this embodiment, characteristic areas where pollutant concentrations change significantly are searched for in the first pollution distribution data. The pollutant concentration gradient represents the rate of change of pollutant concentrations in space. By calculating the pollutant concentration gradient at each spatial location, it is possible to determine which areas have larger concentration changes. The preset threshold is a concentration gradient critical value set based on actual conditions. When the pollutant concentration gradients in a certain continuous spatial area exceed the preset threshold, the area is identified as a characteristic area. For example, in the first pollution distribution data, there may be a local area where the pollutant concentration rises rapidly from a lower value to a higher value. The concentration gradient in this area is large and exceeds the preset threshold, so it is identified as a characteristic area.

[0079] Step S1423: searching the second pollution distribution data for a matching region that is most similar to the concentration distribution pattern of the characteristic region, wherein the search range of the matching region is limited to an adjacent space range around the characteristic region.

[0080] In this embodiment, a matching area that is most similar to the concentration distribution pattern of the characteristic area in the first pollution distribution data is searched in the second pollution distribution data. In order to improve the search efficiency and accuracy, the search range is limited to the adjacent space around the characteristic area. This is because the migration of pollutants is usually a continuous process. At adjacent time points, it is unlikely that the pollutants will suddenly jump to a very far place. By comparing the concentration distribution patterns of the characteristic area and each area within the adjacent space, the most similar area is found as the matching area. Some similarity measurement methods, such as correlation analysis or pattern matching algorithms, can be used to determine the degree of similarity of the concentration distribution patterns.

[0081] Step S1424: Calculate the spatial position offset between the feature area and the matching area, where the spatial position offset includes a horizontal offset distance and a vertical offset distance.

[0082] In this embodiment, the center position or representative position of the feature area and the matching area is determined, and then the spatial position offset between them is calculated. The spatial position offset includes a horizontal offset distance and a vertical offset distance. The horizontal offset distance can be obtained by calculating the coordinate difference between the feature area and the matching area in the longitude and latitude directions, and the vertical offset distance can be obtained by calculating the coordinate difference between them in the altitude direction. For example, the center coordinates of the feature area are (x1, y1, z1) and the center coordinates of the matching area are (x2, y2, z2). The horizontal offset distance can be expressed as a combination of the coordinate difference in the x and y directions, and the vertical offset distance is the coordinate difference in the z direction.

[0083] Step S1425: determining the migration direction and migration distance of the feature region between adjacent time markers according to the spatial position offset, and generating a single-step migration trajectory of the feature region.

[0084] In this embodiment, the migration direction and migration distance of the feature area between adjacent time markers can be determined based on the calculated spatial position offset. The migration direction can be determined by the direction of the horizontal offset distance and the vertical offset distance. For example, if the horizontal offset distance has a certain value in the positive direction of the x-axis and the vertical offset distance has a certain value in the positive direction of the z-axis, then the migration direction is the composite direction of the positive direction of the x-axis and the positive direction of the z-axis. The migration distance can be obtained by calculating the modulus of the spatial position offset. For example, the square root of the sum of the squares of the horizontal offset distance and the vertical offset distance is calculated using the Euclidean distance formula. Based on the migration direction and migration distance, a single-step migration trajectory of the feature area between adjacent time markers is generated. The single-step migration trajectory can be represented by a vector, where the direction of the vector represents the migration direction and the length of the vector represents the migration distance.

[0085] Step S1426: Continuously splicing the single-step migration trajectories of the characteristic region between multiple adjacent time markers to generate a migration trajectory of the pollutant in three-dimensional space that reflects the overall migration process of the characteristic region.

[0086] In this embodiment, the single-step migration trajectories of the characteristic region between multiple adjacent time markers are continuously spliced. For example, the characteristic region has a single-step migration trajectory between time points t1 and t2, and another single-step migration trajectory between time points t2 and t3. These two single-step migration trajectories are connected in chronological order. Starting from the end point of the first single-step migration trajectory, the starting point of the second single-step migration trajectory is connected, and so on. The single-step migration trajectories between all adjacent time markers are connected to generate a migration trajectory of the pollutant in three-dimensional space that reflects the overall migration process of the characteristic region. This migration trajectory can intuitively show the movement path and trend of the pollutant over a period of time.

[0087] Step S143: determining the diffusion starting position, diffusion direction and diffusion speed parameters of the pollutants according to the migration trajectory, and generating a trajectory feature set reflecting the pollution diffusion process.

[0088] In this embodiment, based on the generated pollutant migration trajectory, the diffusion starting position, diffusion direction and diffusion speed parameters of the pollutants can be determined. The diffusion starting position can be determined by the starting point of the migration trajectory, that is, the position of the characteristic area at the initial time point. The diffusion direction can be determined by the overall direction of the migration trajectory. For example, the directions of each single-step migration trajectory on the migration trajectory are comprehensively analyzed to obtain an overall diffusion direction. The diffusion speed parameter can be obtained by calculating the ratio of the migration distance and the time interval between adjacent time points on the migration trajectory. For example, within the time interval Δt, the distance the pollutant migrates is Δs, then the diffusion speed v=Δs / Δt. These diffusion starting positions, diffusion directions and diffusion speed parameters are sorted to form a trajectory feature set, which contains key information reflecting the pollution diffusion process.

[0089] Step S144: superimposing and fusing the trajectory feature set with the pollution distribution three-dimensional feature field, performing differential display of pollutant concentration, diffusion direction and diffusion speed through visualization coding rules, and generating the dynamic three-dimensional visualization result.

[0090] In this embodiment, the trajectory feature set is superimposed and fused with the three-dimensional pollution distribution feature field to combine pollutant diffusion information with the spatial information of the pollution distribution. Specifically, the diffusion starting position, diffusion direction, and diffusion speed information in the trajectory feature set are associated with the pollutant concentration information at each spatial location in the three-dimensional pollution distribution feature field. This information is then displayed differently using visualization coding rules.

[0091] Step S1441: Establish a mapping rule between pollutant concentration values ​​and color brightness, set values ​​of different concentration levels to correspond to colors of different brightness, and generate a concentration-brightness mapping table, in which values ​​with lower concentration levels correspond to colors with lower brightness, and values ​​with higher concentration levels correspond to colors with higher brightness.

[0092] In this embodiment, in order to intuitively display the levels of pollutant concentration, a mapping rule between pollutant concentration values ​​and color brightness is established. According to the range of pollutant concentration values, it is divided into different concentration level intervals. For each concentration level interval, a corresponding color brightness value is set. For example, a lower concentration level interval corresponds to a lower brightness color, such as dark blue; a higher concentration level interval corresponds to a higher brightness color, such as bright yellow. These concentration level intervals and corresponding color brightness values ​​are sorted to generate a concentration-brightness mapping table. In this concentration-brightness mapping table, each concentration value has a unique corresponding color brightness value, which facilitates subsequent visualization.

[0093] Step S1442: establishing a mapping rule between diffusion direction and arrow direction, converting the direction component of the migration trajectory into the direction angle of the arrow symbol in the three-dimensional space, and generating a direction-direction mapping table.

[0094] In this embodiment, in order to clearly show the diffusion direction of pollutants, a mapping rule between the diffusion direction and the arrow direction is established. The direction component of the migration trajectory contains the direction information in three-dimensional space. Through relevant mathematical calculations, these direction components are converted into the pointing angle of the arrow symbol in three-dimensional space. For example, based on the components of the migration trajectory in the directions of the three coordinate axes x, y, and z, the inclination angle and rotation angle of the arrow in three-dimensional space are calculated. The different diffusion directions and the corresponding arrow pointing angles are sorted out to generate a direction-pointing mapping table. In the direction-pointing mapping table, each diffusion direction has a corresponding arrow pointing angle, so that the diffusion direction of the pollutant can be accurately represented by the arrow symbol during visualization.

[0095] Step S1443: Establish a mapping rule between diffusion speed and display transparency, set the values ​​of different speed levels to correspond to different transparencies, and generate a speed-transparency mapping table, in which the lower the speed level, the higher the transparency, and the higher the speed level, the lower the transparency.

[0096] In this embodiment, in order to display the diffusion speed of pollutants, a mapping rule between diffusion speed and display transparency is established. According to the value range of the diffusion speed, it is divided into different speed level intervals. For each speed level interval, a corresponding display transparency value is set. For example, a lower speed level interval corresponds to a higher transparency, which makes the display content look blurry, indicating that the diffusion speed is slower; a higher speed level interval corresponds to a lower transparency, which makes the display content look clearer, indicating that the diffusion speed is faster. These speed level intervals and the corresponding display transparency values ​​are sorted to generate a speed-transparency mapping table. In the speed-transparency mapping table, each diffusion speed value has a corresponding display transparency value, which is convenient for setting during visual display.

[0097] Step S1444: For each spatial position in the three-dimensional characteristic field of the pollution distribution, query the concentration-brightness mapping table according to the pollutant concentration value at the spatial position to determine the brightness parameter of the displayed color, query the direction-pointing mapping table according to the diffusion direction of the spatial position to determine the pointing parameter of the arrow symbol, and query the speed-transparency mapping table according to the diffusion speed of the spatial position to determine the transparency parameter of the displayed content.

[0098] In this embodiment, for each spatial location in the three-dimensional pollution distribution feature field, display parameters are determined by querying the corresponding mapping table based on information such as the pollutant concentration, diffusion direction, and diffusion velocity. For example, for a spatial location with a pollutant concentration of C, the corresponding color brightness value L is found by querying the concentration-brightness mapping table; the corresponding arrow pointing angle A is found by querying the direction-pointing mapping table; and the corresponding display transparency value T is found by querying the velocity-transparency mapping table for a diffusion velocity of V. In this way, display parameters such as color brightness, arrow direction, and display transparency are determined for each spatial location.

[0099] Step S1445: Integrate the brightness parameters, orientation parameters, and transparency parameters of all spatial positions to generate the dynamic three-dimensional visualization result including the superposition of multi-dimensional information.

[0100] In this embodiment, the brightness parameters, direction parameters, and transparency parameters of all spatial locations in the three-dimensional characteristic field of pollution distribution are integrated. For each spatial location, its display color is set according to the determined color brightness value, an arrow symbol is drawn according to the arrow pointing angle to indicate the diffusion direction, and the transparency of the displayed content is set according to the display transparency value. These display settings of all spatial locations are comprehensively processed to generate a dynamic three-dimensional visualization result containing multi-dimensional information superposition. This dynamic three-dimensional visualization result can intuitively display information such as the spatial distribution of groundwater pollution, the diffusion path of pollutants, and the diffusion speed, helping relevant personnel better understand the situation and evolution process of groundwater pollution.

[0101] Step S150: outputting the dynamic three-dimensional visualization result to a display device to present the spatial distribution and evolution process of groundwater pollution.

[0102] In this embodiment, the generated dynamic 3D visualization results are output to a display device. The display device can be a computer monitor, a virtual reality (VR) device, or a large-screen projector. The data format of the dynamic 3D visualization results is converted into a format that the display device can recognize, such as a common image file format or video file format. The data is then transmitted to the display device for display via the corresponding software or driver. During the display process, users can interact with the dynamic 3D visualization results through the display device's control interface, such as zooming, rotating, and switching the display between different time points, to more comprehensively observe the spatial distribution and evolution of groundwater contamination. For example, users can zoom to more clearly view the contamination situation in a local area; rotate to observe the three-dimensional structure of the pollution distribution from different angles; and switch the display between different time points to observe the diffusion process of pollutants over time. This allows relevant personnel to conduct in-depth analysis and make decisions based on the displayed dynamic 3D visualization results.

[0103] Figure 2 A schematic diagram illustrates exemplary hardware and software components of an artificial intelligence-based 3D visualization system for groundwater contamination 100, which can implement the concepts of the present application, according to some embodiments of the present application. For example, a processor 120 can be used in the artificial intelligence-based 3D visualization system for groundwater contamination 100 to perform the functions described in the present application.

[0104] The artificial intelligence-based 3D visualization system for groundwater contamination 100 can be a general-purpose server or a special-purpose server, both of which can be used to implement the artificial intelligence-based 3D visualization method for groundwater contamination described herein. Although only one server is shown in this application, for convenience, the functions described herein can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0105] For example, the artificial intelligence-based three-dimensional visualization system 100 for groundwater pollution may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and storage media 140 in different forms, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the artificial intelligence-based three-dimensional visualization system 100 for groundwater pollution may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The artificial intelligence-based three-dimensional visualization system 100 for groundwater pollution also includes an I / O interface 150 between the computer and other input and output devices.

[0106] For ease of explanation, only one processor is described in the artificial intelligence-based three-dimensional visualization system 100 for groundwater pollution. However, it should be noted that the artificial intelligence-based three-dimensional visualization system 100 for groundwater pollution in this application may also include multiple processors, so the steps performed by one processor described in this application may also be performed jointly or individually by multiple processors. For example, if the processor of the artificial intelligence-based three-dimensional visualization system 100 for groundwater pollution executes step A and step B, it should be understood that step A and step B may also be executed jointly by two different processors or individually in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor execute steps A and B together.

[0107] In addition, an embodiment of the present invention also provides a readable storage medium, in which computer-executable instructions are preset. When a processor executes the computer-executable instructions, the above-mentioned artificial intelligence-based three-dimensional visualization method for groundwater pollution is implemented.

[0108] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, multiple features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A three-dimensional visualization method for groundwater pollution based on artificial intelligence, characterized in that: The method comprises: Acquire a groundwater basic data set containing pollution monitoring information, wherein the groundwater basic data set includes pollutant distribution information at different spatial locations and corresponding time identifiers; Extracting spatiotemporal features from the groundwater basic data set using a spatiotemporal association learning model to generate a composite feature set reflecting the spatiotemporal evolution of pollution; Performing spatial mapping processing on the composite feature set using a three-dimensional visualization generation model to generate a three-dimensional feature field of pollution distribution having a corresponding relationship of spatial coordinates; Performing dynamic trend enhancement processing on the three-dimensional characteristic field of pollution distribution to generate a dynamic three-dimensional visualization result containing pollution diffusion path information; Outputting the dynamic three-dimensional visualization result to a display device to present the spatial distribution and evolution process of groundwater pollution; The spatiotemporal feature extraction of the groundwater basic data set by the spatiotemporal association learning model is performed to generate a composite feature set reflecting the spatiotemporal evolution law of pollution, including: Extracting pollutant distribution information at different time marks at the same spatial location from the groundwater basic data set, and constructing a pollution evolution sequence in the time dimension; Extracting pollutant distribution information at different spatial locations under the same time stamp from the groundwater basic data set, and constructing a pollution distribution matrix in spatial dimension; Inputting the pollution evolution sequence of the time dimension and the pollution distribution matrix of the space dimension into the bidirectional attention module of the spatiotemporal association learning model, and calculating the influence weight of the time dimension on the spatial dimension and the influence weight of the spatial dimension on the time dimension respectively through the bidirectional attention module; Performing weighted fusion processing on the pollution evolution sequence in the time dimension and the pollution distribution matrix in the space dimension based on the impact weight to generate a correlation feature set that includes both time evolution features and spatial distribution features; The associated feature set is subjected to redundant information filtering processing by the feature extraction layer of the spatiotemporal association learning model, and the composite feature set reflecting the evolution law of the pollution core is retained.

2. The method for three-dimensional visualization of groundwater pollution based on artificial intelligence according to claim 1, characterized in that: The step of inputting the pollution evolution sequence of the time dimension and the pollution distribution matrix of the space dimension into the bidirectional attention module of the spatiotemporal association learning model, and calculating the influence weight of the time dimension on the spatial dimension and the influence weight of the spatial dimension on the time dimension respectively through the bidirectional attention module, includes: In the time-to-space attention submodule of the bidirectional attention module, the influence of the pollution evolution characteristics of each time point on the pollution distribution of all spatial positions is calculated to generate a time-to-space attention weight matrix; In the space-to-time attention submodule of the bidirectional attention module, the influence of the pollution distribution characteristics of each spatial position on the pollution evolution at all time points is calculated to generate a space-to-time attention weight matrix; Normalizing the time-to-space attention weight matrix and the space-to-time attention weight matrix so that each weight value represents the relative influence strength between corresponding dimensions; The normalized time-to-space attention weight matrix and the space-to-time attention weight matrix are input into the fusion unit of the bidirectional attention module, and a comprehensive attention weight tensor that simultaneously includes the time-space bidirectional influence relationship is generated through matrix dot multiplication operation.

3. The method for three-dimensional visualization of groundwater pollution based on artificial intelligence according to claim 1, characterized in that: The method of performing spatial mapping processing on the composite feature set by using a three-dimensional visualization generation model to generate a three-dimensional feature field of pollution distribution having a spatial coordinate correspondence relationship includes: Inputting the composite feature set into a spatial coding unit of the three-dimensional visualization generation model, and constructing a three-dimensional grid coordinate system based on the spatial position information in the groundwater basic data set through the spatial coding unit; In the three-dimensional grid coordinate system, the interpolation prediction unit of the three-dimensional visualization generation model is used to infer pollutant distribution information for spatial locations that are not directly monitored, thereby generating continuous pollution distribution data covering the entire monitoring area; Inputting the continuous pollution distribution data into the feature enhancement unit of the three-dimensional visualization generation model, enhancing the transition boundary information of areas with different pollution levels through boundary detection operations, and generating an intermediate three-dimensional feature field containing concentration gradient features; The intermediate three-dimensional feature field is subjected to spatial coordinate calibration processing to ensure that the spatial coordinates of each position in the intermediate three-dimensional feature field correspond one-to-one with the geographical coordinates of the actual monitoring area, thereby generating the pollution distribution three-dimensional feature field.

4. The method for three-dimensional visualization of groundwater pollution based on artificial intelligence according to claim 3, characterized in that: In the three-dimensional grid coordinate system, pollutant distribution information is inferred for spatial locations that are not directly monitored by the interpolation prediction unit of the three-dimensional visualization generation model to generate continuous pollution distribution data covering the entire monitoring area, including: Determining a spatial range boundary of the three-dimensional grid coordinate system, wherein the spatial range boundary is determined by the spatial position extreme values ​​of all monitoring points in the groundwater basic data set; Dividing the space into grid cells according to a preset spatial resolution within the spatial range boundary to generate a regularly arranged three-dimensional grid point set; For each 3D grid point that is not directly monitored, extract the pollutant distribution information and spatial location information of the monitored points within the preset neighborhood around the 3D grid point; The weighted average model of the interpolation prediction unit is called to calculate the pollutant distribution information of the monitored points in the preset neighborhood. The weighted average model generates a weight coefficient according to the spatial distance between the grid point and the monitored point. The closer the distance to the monitored point, the larger the corresponding weight coefficient; Based on the weight coefficient, the pollutant distribution information of the monitored points in the surrounding preset neighborhood is weighted and summed to obtain the predicted value of the pollutant distribution information of the three-dimensional grid points that are not directly monitored, and generate the continuous pollution distribution data covering the complete monitoring area.

5. The method for three-dimensional visualization of groundwater pollution based on artificial intelligence according to claim 3, characterized in that: The performing of spatial coordinate calibration processing on the intermediate three-dimensional feature field to ensure that the spatial coordinates of each position in the intermediate three-dimensional feature field correspond one-to-one with the geographic coordinates of the actual monitoring area, and generating the pollution distribution three-dimensional feature field, includes: Selecting a plurality of reference points with known actual geographic coordinates in the intermediate three-dimensional feature field, wherein the reference points are spatial location points directly monitored in the groundwater basic data set; Extracting the model space coordinates and the corresponding actual geographic coordinates of the reference point in the intermediate three-dimensional feature field; Establishing a conversion relationship between the model space coordinates and the actual geographic coordinates through a coordinate conversion algorithm, wherein the coordinate conversion algorithm includes a combination of translation transformation, rotation transformation and scaling transformation; Using the conversion relationship, the model space coordinates of all positions in the intermediate three-dimensional feature field are converted to generate calibration space coordinates corresponding to the actual geographic coordinates; The matching accuracy between the calibration space coordinates and the actual geographic coordinates is verified, and when the matching accuracy meets the preset requirements, the calibrated intermediate three-dimensional feature field is determined as the pollution distribution three-dimensional feature field.

6. The method for three-dimensional visualization of groundwater pollution based on artificial intelligence according to claim 1, characterized in that: The dynamic trend enhancement processing is performed on the pollution distribution three-dimensional feature field to generate a dynamic three-dimensional visualization result containing pollution diffusion path information, including: Extracting pollution distribution data at different time marks in the three-dimensional characteristic field of pollution distribution, and constructing a time series pollution distribution data set; Performing evolution law mining on the pollution distribution data set of the time series, and identifying the migration trajectory of pollutants in three-dimensional space through a motion trajectory tracking algorithm; Determining the diffusion starting position, diffusion direction, and diffusion speed parameters of the pollutants based on the migration trajectory, and generating a trajectory feature set reflecting the pollution diffusion process; The trajectory feature set is superimposed and fused with the three-dimensional feature field of the pollution distribution, and the pollutant concentration, diffusion direction and diffusion speed are differentially displayed through visualization coding rules to generate the dynamic three-dimensional visualization result, wherein the visualization coding rules include the mapping relationship between concentration value and color brightness, the mapping relationship between diffusion direction and arrow direction, and the mapping relationship between diffusion speed and display transparency.

7. The method for three-dimensional visualization of groundwater pollution based on artificial intelligence according to claim 6, characterized in that: The step of performing evolution law mining on the pollution distribution data set of the time series and identifying the migration trajectory of pollutants in three-dimensional space by using a motion trajectory tracking algorithm includes: Selecting first pollution distribution data and second pollution distribution data with adjacent time tags in the pollution distribution data set of the time series; Identifying characteristic regions with significant changes in pollutant concentration in the first pollution distribution data, wherein the characteristic regions are continuous spatial regions where the pollutant concentration gradient exceeds a preset threshold; Searching the second pollution distribution data for a matching region that is most similar to the concentration distribution pattern of the characteristic region, wherein the search range of the matching region is limited to an adjacent space around the characteristic region; Calculating a spatial position offset between the feature area and the matching area, wherein the spatial position offset includes a horizontal offset distance and a vertical offset distance; determining the migration direction and migration distance of the feature region between adjacent time markers according to the spatial position offset, and generating a single-step migration trajectory of the feature region; The single-step migration trajectories of the characteristic region between a plurality of adjacent time markers are continuously spliced ​​together to generate a migration trajectory of the pollutant in three-dimensional space that reflects the overall migration process of the characteristic region.

8. The method for three-dimensional visualization of groundwater pollution based on artificial intelligence according to claim 6, characterized in that: The differential display of pollutant concentration, diffusion direction and diffusion speed by using visualization coding rules to generate the dynamic three-dimensional visualization results includes: Establish a mapping rule between pollutant concentration values ​​and color brightness, set values ​​of different concentration levels to correspond to colors of different brightness, and generate a concentration-brightness mapping table, where lower concentration values ​​correspond to lower brightness colors, and higher concentration values ​​correspond to higher brightness colors; Establish a mapping rule between diffusion direction and arrow orientation, convert the directional component of the migration trajectory into the directional angle of the arrow symbol in three-dimensional space, and generate a direction-orientation mapping table; Establish a mapping rule between diffusion speed and display transparency, set different speed level values ​​to correspond to different transparency levels, and generate a speed-transparency mapping table, where lower speed level values ​​correspond to higher transparency, and higher speed level values ​​correspond to lower transparency. For each spatial position in the three-dimensional characteristic field of pollution distribution, the concentration-brightness mapping table is queried based on the pollutant concentration value at the spatial position to determine the brightness parameter of the display color, the direction-pointing mapping table is queried based on the diffusion direction of the spatial position to determine the direction parameter of the arrow symbol, and the speed-transparency mapping table is queried based on the diffusion speed of the spatial position to determine the transparency parameter of the display content; The brightness parameters, orientation parameters and transparency parameters of all spatial positions are integrated to generate the dynamic three-dimensional visualization result including the superposition of multi-dimensional information.

9. A three-dimensional visualization system for groundwater pollution based on artificial intelligence, characterized in that: It includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the artificial intelligence-based three-dimensional visualization method for groundwater pollution as described in any one of claims 1 to 8.

Images