Mining area heavy metal monitoring method and system based on spatio-temporal evolution

Abstract

The invention discloses a mining area heavy metal monitoring method and system based on spatio-temporal evolution, relates to the technical field of environmental pollution monitoring, and solves the technical problem of inaccurate monitoring result caused by incomplete data and single analysis method in the prior art. According to the method, time sequence hyperspectral data and time sequence geochemical data of a target area are collected, space-time alignment is carried out, and space-time consistency of the data is ensured; the hyperspectral data and the geochemical data are processed through a first model and a second model which are constructed based on an artificial intelligence model, a heavy metal inversion result and a prediction result are obtained, the weight of the heavy metal inversion result and the weight of the prediction result are dynamically adjusted through a weight adjustment function set in a third model, and a heavy metal monitoring result is comprehensively obtained; according to the method, the reliability and correlation of different data sources are fully considered, the actual concentration and distribution condition of the heavy metals in the mining area can be reflected more accurately, and a reliable basis is provided for environmental governance and risk assessment of the mining area.

Description

Technical Field

[0001] The present invention belongs to the field of environmental pollution monitoring, and in particular relates to a mining area heavy metal monitoring method and system based on spatiotemporal evolution. Background Art

[0002] Heavy metal pollution in mining areas is becoming increasingly prominent, posing a serious threat to the ecological environment and human health. Efficient and accurate heavy metal monitoring technology in mining areas has become a key research direction. Existing monitoring technologies have numerous limitations in data acquisition, model building, and spatiotemporal analysis, urgently requiring innovative breakthroughs.

[0003] Currently, heavy metal inversion models based on hyperspectral data and prediction models based on geochemical data suffer from significant deficiencies in accuracy and comprehensiveness. For one thing, the spectral signatures of most heavy metals at concentrations below 4.0 mg / g are weak and easily masked by the spectra of other soil components, significantly compromising the accuracy of hyperspectral inversion models. Furthermore, traditional monitoring models often overlook the impact of environmental factors such as atmosphere, lighting, and topography on hyperspectral data. The combined effects of these environmental factors make it difficult for traditional models to accurately reflect the true concentration and distribution of heavy metals.

[0004] The present invention provides a mining area heavy metal monitoring method and system based on spatiotemporal evolution, which are used to solve the above technical problems. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a mining area heavy metal monitoring method and system based on spatiotemporal evolution, which realizes accurate monitoring and comprehensive evaluation of heavy metals in mining areas through multi-source data fusion and intelligent analysis; in the data acquisition link, time series hyperspectral data and time series geochemical data of the target area are collected, and spatiotemporal alignment is performed to ensure the spatiotemporal consistency of the data; Model 1 and Model 2 constructed based on the artificial intelligence model are used to process the hyperspectral data and geochemical data respectively to obtain heavy metal inversion results and prediction results, and then the weight adjustment function set in Model 3 is used to dynamically adjust the weights of the heavy metal inversion results and the prediction results to comprehensively obtain the heavy metal monitoring results; the present invention fully considers the reliability and correlation of different data sources, overcomes the problem of inaccurate monitoring results due to incomplete data and single analysis method in the prior art, can more accurately reflect the actual concentration and distribution of heavy metals in mining areas, and provides a reliable basis for mining area environmental governance and risk assessment.

[0006] To achieve the above objectives, the first aspect of the present invention provides a mining area heavy metal monitoring method based on spatiotemporal evolution, comprising:

[0007] Collecting time-series hyperspectral data and time-series geochemical data of the target area, performing spatiotemporal alignment on the time-series hyperspectral data and the time-series geochemical data to obtain several spatiotemporal data groups; wherein the spatiotemporal data groups include hyperspectral data and geochemical data of the same spatiotemporal area, and the target area includes the mining area;

[0008] Hyperspectral data from several data sets are input into Model 1 to obtain heavy metal inversion results; geochemical data are input into Model 2 to obtain heavy metal prediction results. Models 1 and 2 are constructed based on artificial intelligence models.

[0009] The heavy metal inversion results and heavy metal prediction results are input into Model 3 to calculate the heavy metal monitoring results; wherein, a weight adjustment function is set in Model 3 to dynamically adjust the weights of the heavy metal inversion results and heavy metal prediction results;

[0010] Extract heavy metal monitoring results corresponding to several data groups; render the heavy metal monitoring results according to the time and space corresponding to the several data groups to obtain a heavy metal spatiotemporal evolution diagram.

[0011] Preferably, the step of acquiring time series hyperspectral data includes:

[0012] Identify remote sensing satellites that cover the target mining area and collect hyperspectral data of the target area through remote sensing satellites; remote sensing satellites include EO-1 Hyperion satellite and Gaofen-5 satellite;

[0013] After preprocessing, the collected hyperspectral data are integrated into time series hyperspectral data according to the acquisition time; the preprocessing includes radiation correction, atmospheric correction, geometric correction and image segmentation.

[0014] Preferably, the step of acquiring time series geochemical data includes:

[0015] Divide the target area into several grids, evenly arrange sampling points in the grids, and collect sample data of the sampling points at set time intervals;

[0016] After preprocessing, the sample data are arranged in chronological order to obtain time series geochemical data.

[0017] Preferably, the heavy metal monitoring results are calculated, including:

[0018] The heavy metal inversion results and heavy metal prediction results were integrated and input into Model 3;

[0019] Model 3 dynamically adjusts the weights according to the heavy metal inversion results and heavy metal prediction results, and performs weighted summation based on the dynamically adjusted weights to obtain the final heavy metal monitoring results.

[0020] Preferably, the weight coefficient is calculated by a weight adjustment function set in model three; wherein the weight adjustment function is constructed based on the sigmoid function.

[0021] Preferably, the weight adjustment function is: Where C is the heavy metal prediction result corresponding to the geochemical data, k is an adjustable parameter used to control the rate of weight change, θ is taken as 0.4, the weight of the heavy metal inversion result is w1 = f(C), and the weight of the heavy metal prediction result is w2 = 1-w1.

[0022] Preferably, the environmental quality index is introduced into the weight adjustment function, and the weight adjustment function is Where α is the environmental quality index influence coefficient, which is used to adjust the influence of the environmental quality index E on the function; the environmental quality index E is used to evaluate the impact of environmental factors on hyperspectral data.

[0023] Preferably, the calculation of environmental quality indicators includes:

[0024] Extract the environmental factors at the corresponding moment of the data set; the environmental factors include aerosol thickness, water vapor content, ozone concentration, solar altitude angle, cloud cover, slope and aspect;

[0025] Set the weight coefficient of each environmental factor, add the standardized results of the environmental factors and the weight coefficient to obtain the environmental quality index.

[0026] Preferably, extracting environmental factors at corresponding moments of the data set includes:

[0027] Collect historical environmental data;

[0028] Environmental factors of several data groups are matched from historical environmental data according to spatiotemporal elements, and the matched environmental factors are associated with the data groups; wherein the spatiotemporal elements include the spatial coordinates and collection time of the data groups.

[0029] The second aspect of the present invention provides a mining area heavy metal monitoring system based on spatiotemporal evolution, comprising a spatiotemporal monitoring module and a data acquisition module connected thereto;

[0030] Data acquisition module: used to collect time-series hyperspectral data and time-series geochemical data of the target area, perform spatiotemporal alignment on the time-series hyperspectral data and time-series geochemical data, and obtain several spatiotemporal data groups; wherein the spatiotemporal data groups include hyperspectral data and geochemical data of the same spatiotemporal data, and the target area includes the mining area;

[0031] Spatiotemporal monitoring module: used to input hyperspectral data from several data sets into Model 1 to obtain heavy metal inversion results; input geochemical data into Model 2 to obtain heavy metal prediction results; wherein, Model 1 and Model 2 are constructed based on artificial intelligence models; and

[0032] It is used to input the heavy metal inversion results and heavy metal prediction results into Model 3 to calculate the heavy metal monitoring results; extract the heavy metal monitoring results corresponding to several data groups; render several heavy metal monitoring results according to the time and space corresponding to several data groups to obtain a heavy metal spatiotemporal evolution diagram; wherein, a weight adjustment function is set in Model 3 for dynamically adjusting the weight.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. The present invention realizes accurate monitoring and comprehensive assessment of heavy metals in mining areas through multi-source data fusion and intelligent analysis; in the data acquisition link, time-series hyperspectral data and time-series geochemical data of the target area are collected, and time-space alignment is performed to ensure the time-space consistency of the data; Model 1 and Model 2 constructed based on the artificial intelligence model are used to process the hyperspectral data and geochemical data respectively to obtain heavy metal inversion results and prediction results, and then the weight adjustment function set in Model 3 is used to dynamically adjust the weights of the heavy metal inversion results and prediction results to comprehensively obtain the heavy metal monitoring results; the present invention fully considers the reliability and correlation of different data sources, overcomes the problem of inaccurate monitoring results due to incomplete data and single analysis method in the existing technology, can more accurately reflect the actual concentration and distribution of heavy metals in the mining area, and provides a reliable basis for mining area environmental governance and risk assessment.

[0035] 2. The present invention can realize the spatiotemporal dynamic monitoring of heavy metals in mining areas and effectively predict their changing trends; by continuously collecting time series data at different time points, combined with the spatial sampling point layout and hyperspectral data coverage, the present invention can comprehensively obtain information on heavy metals in mining areas under different spatiotemporal conditions; extract the heavy metal monitoring results corresponding to several data groups, and render these results according to time and space to generate a heavy metal spatiotemporal evolution map; the heavy metal spatiotemporal evolution map can intuitively display the migration trajectory of heavy metals in the mining area over time and the diffusion in space, helping researchers to clearly observe the development trend of heavy metal pollution; compared with the existing technology that can only perform static or single-point monitoring, the spatiotemporal dynamic monitoring method of the present invention has obvious advantages. It can not only grasp the current status of heavy metal pollution in mining areas in real time, but also make predictions based on historical data and current trends, and give early warning of potential environmental risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0037] Figure 1 Schematic diagram of the method steps of the mining area heavy metal monitoring method of the present invention;

[0038] Figure 2 Schematic diagram of the process of acquiring time series hyperspectral data in the present invention;

[0039] Figure 3 Schematic diagram of the process of acquiring time-series geochemical data in the present invention. DETAILED DESCRIPTION

[0040] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0041] See also Figure 1-Figure 3 The first embodiment of the present invention provides a mining area heavy metal monitoring method based on spatiotemporal evolution, comprising:

[0042] S100: collecting time-series hyperspectral data and time-series geochemical data of a target area, performing spatiotemporal alignment on the time-series hyperspectral data and the time-series geochemical data to obtain a plurality of spatiotemporal data groups; wherein the spatiotemporal data groups include hyperspectral data and geochemical data of the same spatiotemporal time, and the target area includes a mining area;

[0043] S200: Inputting hyperspectral data from several data sets into Model 1 to obtain heavy metal inversion results; inputting geochemical data into Model 2 to obtain heavy metal prediction results; wherein Model 1 and Model 2 are constructed based on artificial intelligence models;

[0044] S300: Inputting the heavy metal inversion results and heavy metal prediction results into Model 3 to obtain heavy metal monitoring results; wherein, a weight adjustment function is set in Model 3 for dynamically adjusting the weight;

[0045] S400: extracting heavy metal monitoring results corresponding to a number of data groups; rendering the heavy metal monitoring results according to the time and space corresponding to the number of data groups to obtain a heavy metal spatiotemporal evolution diagram.

[0046] In S100, the target area refers to a set area including a target mining area, which is a mining area that needs to be monitored for heavy metals. The target area is set around the mining area. While monitoring the heavy metals in the mining area, heavy metal monitoring is also performed on areas around the mining area that may be affected.

[0047] The detailed steps for acquiring time series hyperspectral data of the target area are as follows:

[0048] S111: Determine remote sensing satellites covering the target mining area and collect hyperspectral data of the target area through remote sensing satellites; remote sensing satellites include EO-1 Hyperion satellite and Gaofen-5 satellite;

[0049] S112: After preprocessing the collected hyperspectral data, the data is integrated into time-series hyperspectral data according to the acquisition time; wherein the preprocessing includes radiation correction, atmospheric correction, geometric correction and image segmentation.

[0050] Hyperspectral data collection in the 111 program is primarily through remote sensing satellites, due to their wide coverage and sophisticated data processing capabilities. However, the EO-1 satellite has a five-day revisit period, meaning that theoretically, for a given point, the satellite could potentially collect data every five days. The Gaofen-5 and Gaofen-1 satellites, in a networked configuration, achieve a two-day revisit observation of my country's land areas, typically allowing Gaofen-5 to collect data every two days. Furthermore, remote sensing image acquisition is affected by factors such as satellite mission scheduling, weather conditions, and sensor status.

[0051] Therefore, a single remote sensing satellite cannot meet the time series requirements for hyperspectral data collection. Multiple remote sensing satellites can be used to collect data together. For example, the EO-1 satellite collects data every five days, which does not meet the two-day requirement for time series analysis. Therefore, the Gaofen-5 satellite can be combined with the EO-1 satellite to collect hyperspectral data.

[0052] Furthermore, if multiple remote sensing satellites combined alone cannot meet the hyperspectral data requirements for spatiotemporal analysis, drones equipped with hyperspectral sensors can be used to collect hyperspectral data of the target area. The drones' collection is configured based on the replay period of the remote sensing satellites, leveraging their flexibility to compensate for the lack of hyperspectral data.

[0053] In S112, hyperspectral data is essentially remote sensing satellite imagery. Preprocessing for these data follows the same principles as for remote sensing satellite imagery, primarily including radiometric calibration, atmospheric correction, and geometric correction. Radiometric calibration converts digital quantized values ​​recorded by the sensor into actual radiant brightness values; atmospheric correction eliminates the effects of atmospheric scattering and absorption on the spectrum, restoring the true reflectivity of the ground features; and geometric correction ensures the spatial accuracy of the data, enabling accurate matching of data from different time periods. Each of these preprocessing steps has existing processing algorithms or software. Data preprocessing can be performed by directly invoking built-in modules in software such as ENVI.

[0054] It should be noted that the migration and changes of heavy metals require a certain amount of time, so the set time series analysis interval does not need to be too short. For example, a time analysis can be performed every two days. The time interval can be reasonably set according to the satellite's flight trajectory to ensure that the time series hyperspectral data can invert the heavy metal pollution in the target area and reduce the complexity of hyperspectral data acquisition as much as possible.

[0055] The detailed steps for obtaining time series geochemical data of the target area are as follows:

[0056] S121: Divide the target area into several grids, evenly arrange sampling points in the grids, and collect sample data of the sampling points at set time intervals;

[0057] S122: After pre-processing, the sample data are arranged in chronological order to obtain time series geochemical data.

[0058] In step S121, when dividing the target area into a plurality of grids, the grid division method may be selected based on factors such as the size, shape, and geographical features of the target area. For example, if the target area is a regular rectangle, a square grid may be used; if the area has a complex shape, an irregular polygonal grid may be used to better fit the area boundaries.

[0059] When arranging sampling points, sampling points can be arranged in each grid. If there are a large number of grids, sampling points can be set up in several grids with equal spacing. On this basis, when setting sampling points, it is necessary to combine the geological structure, ore body distribution and historical research data of the mining area. The sampling points should be representative and able to reflect the geochemical characteristics of different areas and different geological conditions in the mining area. At the same time, the sampling points should also have long-term stability so that repeated sampling can be carried out at different times. For example, near faults, at outcrops of mineral veins or in places with large changes in terrain, arranging sampling points in multiple continuous grids can better capture the changing patterns of geochemical elements.

[0060] Sample data types include soil, rock, and water. For soil samples, surface (0-20 cm) and deep (20-100 cm) samples are usually collected to analyze geochemical characteristics at different depths. Rock samples should include representative ore and surrounding rock samples. Water samples are collected from surface water (such as rivers and lakes) and groundwater (such as well water and borehole water) samples based on the hydrogeological conditions of the mining area.

[0061] It's important to note that the time intervals between adjacent moments in the time-series geochemical data should be consistent with those in the time-series hyperspectral data to facilitate subsequent data alignment. The time interval depends on the requirements of spatiotemporal evolution analysis. If heavy metal monitoring is required every two days, a 2-day interval is sufficient.

[0062] The purpose of temporal and spatial alignment of time series hyperspectral data and time series geochemical data is to ensure the temporal and spatial consistency of time series hyperspectral data and time series geochemical data so that they can be effectively integrated to achieve high-precision monitoring of heavy metal pollution.

[0063] Spatiotemporal alignment refers to aligning hyperspectral data and geochemical data in time and space.

[0064] To perform spatial alignment, refer to the following steps:

[0065] 1. Convert the coordinate systems of both the hyperspectral data and the geochemical data to WGS84 or a specific engineering coordinate system. This can be accomplished using the coordinate conversion tools in the Geographic Information System (GIS) software. If the coordinate systems are the same, no conversion is required.

[0066] 2. Select a suitable reference image, such as a high-resolution satellite image or aerial image. Align the hyperspectral data with the reference image. By selecting points with the same name (e.g., distinct and easily identifiable points such as road intersections and river inflection points) on both the hyperspectral and reference images, a registration algorithm is used to calculate the transformation parameters (e.g., translation, rotation, and scaling) of the hyperspectral image relative to the reference image, thereby achieving preliminary spatial positioning and alignment of the hyperspectral data.

[0067] 3. For geochemical data, import the sampling point coordinate file into GIS software and overlay it with the hyperspectral imagery, marking the sampling point coordinates in the same geographic space as the hyperspectral data. Find the locations corresponding to the geochemical sampling points on the hyperspectral imagery. For points that are difficult to accurately match directly, terrain and other features can be used for auxiliary positioning.

[0068] It should be noted that if the spatial resolution of hyperspectral data is inconsistent with the sampling point density of geochemical data, spatial interpolation may be required. For example, if the hyperspectral data has a high resolution but the geochemical sampling points are sparse, methods such as kriging interpolation and inverse distance weighted interpolation can be used to interpolate the geochemical data to obtain a dataset that is more spatially consistent with the hyperspectral data. Conversely, if the hyperspectral data has a low resolution, it can be resampled or fused with other high-resolution data to improve spatial consistency.

[0069] Similarly, since geochemical data are collected through set sampling points, they only cover a part of the target area and do not cover the entire target area. The geochemical data of each sampling point can also be processed by interpolation to obtain geochemical data covering the entire target area.

[0070] Time alignment can refer to the following benchmarks:

[0071] If the time interval between the two data collections is short and the geochemical process changes relatively slowly, a similar time point can be selected as a common reference time, and the geochemical data can be approximately regarded as the data at that time point.

[0072] If the time interval is long and the geochemical processes vary significantly, it is necessary to consider the time series variation of the geochemical data and extrapolate the geochemical data to a time close to the time of hyperspectral data collection through methods such as interpolation, extrapolation, or the establishment of a time series model. For example, if the geochemical element content is known to vary linearly with time, the geochemical element content at the time of hyperspectral data collection can be calculated through linear interpolation based on the time and content data of the two previous samplings.

[0073] In S200 , the hyperspectral data in the plurality of data groups are input into the first model to obtain the heavy metal inversion result; the geochemical data are input into the second model to obtain the heavy metal prediction result.

[0074] The several data groups corresponding to the target area are actually several groups of hyperspectral data and geochemical data in the same space and time (same location, same time); the hyperspectral data in the same data group are controlled to be input into model one to obtain heavy metal inversion results; the geochemical data are input into model two to obtain heavy metal prediction results; the heavy metal inversion results and heavy metal prediction results are then integrated and input into model three. Model three dynamically adjusts the weights of the two, and obtains the final heavy metal monitoring results after weighted summation.

[0075] The training of Model 1 includes:

[0076] Collect hyperspectral data of the mining area, extract spectral data of characteristic bands from the hyperspectral data, and match the heavy metal concentrations corresponding to the spectral data of characteristic bands through measurement and other methods.

[0077] Characteristic bands refer to bands that are closely related to heavy metal concentrations. The correlation coefficient between each band and heavy metal concentration can be calculated using correlation analysis and other methods, and the characteristic bands can be determined based on the correlation coefficient.

[0078] The heavy metal concentration may include multiple heavy metals and concentrations. The heavy metal concentration is determined by laboratory measurement and is used as an output true value for model training.

[0079] Determine the location corresponding to the characteristic band spectral data and match the heavy metal concentration at that location. It is also necessary to determine the medium type at that location. Medium types include rock, soil, water, etc. The medium type is encoded for input into the model for identification.

[0080] After modeling the media type code and the spectral data of the characteristic band, the data is combined with the corresponding heavy metal concentrations that have undergone modeling to form a single piece of training data. This process can generate multiple pieces of training data. Modeling refers to processing the data into a data type that the model can recognize, such as normalization.

[0081] The training data is divided into training, test, and validation sets according to a set ratio. The constructed AI model is trained and, when the training results meet the requirements, the AI ​​model is labeled and saved. The set ratios include 7:2:1 or 5:3:2. AI algorithms include partial least squares regression (PLSR), artificial neural networks (ANN), and support vector machines (SVM), which are well-suited to mapping the relationship between hyperspectral data and heavy metal concentrations.

[0082] After preprocessing the hyperspectral data and media encoding at each moment in the time series hyperspectral data, the data is fed into Model 1 to invert the heavy metal concentrations at each moment. The preprocessing here is the same as that used for model training, including the extraction of characteristic band spectral data and model adaptation.

[0083] Training of model 2:

[0084] Collect a large amount of geochemical data related to heavy metal concentrations in the target mining area or similar mining areas, such as soil pH, redox potential, organic matter content, and the content of various mineral components. Accurately measure the corresponding heavy metal concentration data as labels for training models.

[0085] Accurate measurement of heavy metal concentration data can be carried out in the laboratory. Commonly used analytical methods include inductively coupled plasma mass spectrometry (ICP-MS), atomic absorption spectrometry (AAS), X-ray fluorescence spectrometry (XRF), etc. These methods can accurately determine the content of multiple heavy metal elements in samples.

[0086] Perform feature extraction and modeling on geochemical data, and integrate the processed geochemical data with heavy metal concentration data that has also been modeled into a piece of training data; a number of training data can be obtained based on a large amount of geochemical data;

[0087] The training data is divided into a training set, a test set, and a validation set according to a set ratio. The constructed AI model is trained. When the training results meet the requirements, the AI ​​model is labeled as Model 2 and saved. The set ratios include 7:2:1 or 5:3:2. AI algorithms include decision trees, random forests, support vector machines, etc.

[0088] In S300, the input of model three is the heavy metal inversion result and the heavy metal prediction result. The weights of the two are calculated through the weight adjustment function, and the weighted sum is used to obtain the corresponding heavy metal monitoring result.

[0089] Considering that the spectral characteristics of most heavy metals can only be effectively identified when their contamination concentration exceeds 4.0 mg / g, and that their response characteristics are easily masked by the spectra of other soil components below this concentration, it is necessary to introduce additional weight distribution factors for Models 1 and 2, and set a weight adjustment function in Model 3.

[0090] Considering that the prediction results of geochemical data will not be affected by the actual heavy metal concentration, a weight adjustment function is set based on geochemical data. The weight adjustment function can be constructed based on the sigmoid function. The weight adjustment function is In this formula, C is the heavy metal prediction result corresponding to the geochemical data, k is an adjustable parameter, and k is greater than 0. The default value is 1, which is used to control the weight change rate. The value of θ is 0.4. The weight of the heavy metal inversion result is w1=f(C), and the weight of the heavy metal prediction result is w2=1-w1.

[0091] Considering that hyperspectral data is also affected by various environmental factors, such as atmospheric conditions, lighting conditions, etc., environmental factors are taken into account and a weight adjustment function is set.

[0092] Identify various environmental factors that affect hyperspectral data, such as:

[0093] Atmospheric factors include aerosol thickness, water vapor content, and ozone concentration. Aerosols scatter and absorb light, affecting the radiative transmission of hyperspectral data. Water vapor has a strong absorption effect in the near-infrared band. Ozone primarily affects the ultraviolet band spectrum. All of these factors can interfere with the accuracy of heavy metal concentration inversion using hyperspectral data. Atmospheric factor data can be obtained through ground-based meteorological stations and satellite remote sensing (such as MODIS satellites, which obtain data on aerosol optical thickness and water vapor content). After obtaining this data, it is necessary to align it with the heavy metal inversion and prediction results in both time and space.

[0094] Illumination factors include solar altitude and cloud cover. Solar altitude affects the intensity and angle of illumination on objects, while clouds block sunlight, changing the path and intensity of light, affecting the quality of spectral data. Solar altitude can be calculated based on date and geographic location, while cloud cover can be obtained from meteorological satellite imagery or ground observations. After acquiring this data, it is necessary to align it with the heavy metal inversion and prediction results in both time and space.

[0095] Topographic factors include slope, aspect, and relief. Different slopes and aspects result in varying lighting conditions for features, and relief can cause multiple reflections and scattering of light, affecting spectral characteristics. Topographic factor data can be calculated using a digital elevation model (DEM). After acquiring this data, it is necessary to align it with the heavy metal inversion and prediction results in both time and space.

[0096] Each numerical data in the environmental factors is normalized and converted to the [0,1] interval. Taking the solar altitude angle as an example, first determine its maximum value (max) and minimum value (min) in the study area and study period, and then use the normalization formula Perform standardization.

[0097] For data that is difficult to quantify, such as cloud coverage, a graded score can be applied. The cloud coverage is divided into four levels: no clouds (0-10%), few clouds (10-30%), cloudy (30-70%), and overcast (70-100%), and assigned scores of 0, 0.3, 0.6, and 1 respectively.

[0098] After standardizing or grading the environmental factors, weights are assigned to each environmental factor. Methods for assigning weights include expert experience, principal component analysis, and entropy weighting. These methods are already known in the art, and the calculation process will not be described in detail here.

[0099] Assume that atmospheric factors include aerosol optical thickness AOT, water vapor content WV, ozone concentration O3, illumination factors include solar altitude angle SHA, cloud cover CC, terrain factors include slope Slope, slope aspect Aspect, terrain relief TR, and soil factors include soil moisture SM. The weights corresponding to each environmental factor are calculated by the above method, respectively: w AOT 、w WV 、 w SHA 、w CC 、w Slope 、w Aspect 、w TR and w SM .

[0100] Environmental quality indicators:

[0101] Where norm() represents the function that normalizes the corresponding factors and maps the values ​​of each factor to the interval [0, 1]. score() represents the function that grades the cloud cover and assigns corresponding scores according to different ranges of cloud cover.

[0102] Introducing environmental quality indicators into the weight adjustment function After adjustment, the weight adjustment function is Where α is the environmental quality index influence coefficient, and α is greater than 0 and defaults to 1. It is used to adjust the influence of the environmental quality index E on the function. The larger α is, the more significant the impact of environmental factors on the weight adjustment. When the environmental quality index E increases (the environmental conditions become worse relative to the hyperspectral data), the value of f(C, E) will change. Specifically, the weight w2 = 1-f(C, E) of the geochemical data inversion result increases, and the weight w1 = f(C, E) of the heavy metal inversion result decreases.

[0103] It is worth noting that the weight adjustment function can replace the traditional activation function. For example, in neural network structures such as multi-layer perceptrons (MLPs) and convolutional neural networks (CNNs), activation functions are usually used after the linear combination of neurons to introduce nonlinearity. Traditional activation functions (such as ReLU, tanh, etc.) can be replaced with the above-mentioned weight adjustment function. Of course, the weight adjustment function can also be used simultaneously with other activation functions. In some neural network architectures, different activation functions can be used in different layers. First use the ReLU function to quickly converge and alleviate the gradient vanishing problem. In subsequent layers, use the weighted layer sigmoid function to perform more refined nonlinear transformations on the output to better fit the data.

[0104] The parameters k, θ, and α in the weight adjustment function have a key impact on the final monitoring results. Due to differences in geological conditions, climate conditions, mining activities, and other factors, different mining areas require different parameter combinations to achieve optimal monitoring results. These parameters can be dynamically adjusted based on statistical analysis of historical data: historical monitoring data for the mining area is collected, including hyperspectral data, geochemical data, and corresponding environmental factor data. Through statistical analysis of this data, a relationship model between parameters and monitoring indicators is established. For example, methods such as linear regression and polynomial regression can be used to analyze the accuracy and stability of monitoring results under different parameter combinations, thereby determining the most appropriate initial parameter values ​​for the current mining environment.

[0105] Of course, adjustments can also be made through an adaptive adjustment strategy based on error feedback: During actual monitoring, the error between the model predictions and the actual measurements is continuously compared. When the error exceeds a certain threshold, the parameters are adjusted accordingly based on the direction and magnitude of the error. For example, if the predictions are too high, the value of parameter k can be appropriately reduced to reduce the influence of the geochemical data predictions in the weight calculation.

[0106] After calculating w1 and w2 through the weight adjustment function, the formula C final =w1×C hyperspectral +w2×C geochemical Calculate heavy metal monitoring results C final Where, C hyperspetral is the heavy metal inversion result obtained by model 1, C geohemical This is the heavy metal prediction result obtained by model 2.

[0107] In S400 , heavy metal monitoring results corresponding to a plurality of data groups are extracted; the heavy metal monitoring results are rendered according to the time and space corresponding to the plurality of data groups to obtain a heavy metal spatiotemporal evolution diagram.

[0108] After obtaining heavy metal monitoring results from several datasets, the data can be imported into a geographic information platform to render a spatiotemporal evolution map of heavy metals corresponding to the target area. Furthermore, since the heavy metal monitoring results include multiple heavy metals, the spatiotemporal evolution characteristics of a specific heavy metal can also be selected through the geographic information platform.

[0109] The second embodiment of the present invention provides a mining area heavy metal monitoring system based on spatiotemporal evolution, including a spatiotemporal monitoring module and a data acquisition module connected thereto.

[0110] Data acquisition module: used to collect time-series hyperspectral data and time-series geochemical data of the target area, perform spatiotemporal alignment on the time-series hyperspectral data and time-series geochemical data, and obtain several spatiotemporal data groups; wherein the spatiotemporal data groups include hyperspectral data and geochemical data of the same spatiotemporal data, and the target area includes the mining area;

[0111] Spatiotemporal monitoring module: used to input hyperspectral data from several data sets into Model 1 to obtain heavy metal inversion results; input geochemical data into Model 2 to obtain heavy metal prediction results; wherein, Model 1 and Model 2 are constructed based on artificial intelligence models; and

[0112] It is used to input the heavy metal inversion results and heavy metal prediction results into Model 3 to calculate the heavy metal monitoring results; extract the heavy metal monitoring results corresponding to several data groups; render several heavy metal monitoring results according to the time and space corresponding to several data groups to obtain a heavy metal spatiotemporal evolution diagram; wherein, a weight adjustment function is set in Model 3 for dynamically adjusting the weight.

[0113] The data acquisition module communicates with the spatiotemporal monitoring module, exchanging data between them. The module connects to databases, meteorological platforms, remote sensing data platforms, and other platforms to obtain data for heavy metal monitoring in the target area. The spatiotemporal monitoring module trains and updates Models 1, 2, and 3 to ensure their reliability. It also uses the trained models to analyze data collected by the data acquisition module and renders the spatiotemporal monitoring results to generate a spatiotemporal evolution map of heavy metals in the mining area.

[0114] It should be noted that, in addition to connecting with the database and platform to collect data, the data acquisition module can also collect relevant data by setting up data sensors, and the set data sensors are communicatively connected to the data acquisition module.

[0115] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A mining area heavy metal monitoring method based on spatiotemporal evolution, characterized in that: include: Collecting time-series hyperspectral data and time-series geochemical data of a target area, performing spatiotemporal alignment on the time-series hyperspectral data and the time-series geochemical data to obtain a plurality of spatiotemporal data groups; wherein the spatiotemporal data groups include hyperspectral data and geochemical data in the same spatiotemporal area, and the target area includes a mining area; Inputting the hyperspectral data from the data sets into Model 1 to obtain heavy metal inversion results; inputting the geochemical data into Model 2 to obtain heavy metal prediction results; wherein Model 1 and Model 2 are constructed based on artificial intelligence models; The heavy metal inversion results and heavy metal prediction results are input into Model 3 to calculate the heavy metal monitoring results; wherein a weight adjustment function is provided in Model 3 for dynamically adjusting the weights of the heavy metal inversion results and the heavy metal prediction results; Extracting the heavy metal monitoring results corresponding to the data groups; rendering the heavy metal monitoring results according to the time and space corresponding to the data groups to obtain a heavy metal spatiotemporal evolution diagram.

2. The method for monitoring heavy metals in mining areas based on spatiotemporal evolution according to claim 1, characterized in that: The step of acquiring the time series hyperspectral data includes: Identify remote sensing satellites that cover the target mining area and collect hyperspectral data of the target area through remote sensing satellites; remote sensing satellites include EO-1 Hyperion satellite and Gaofen-5 satellite; After preprocessing, the collected hyperspectral data are integrated into time series hyperspectral data according to the acquisition time; the preprocessing includes radiation correction, atmospheric correction, geometric correction and image segmentation.

3. The method for monitoring heavy metals in mining areas based on spatiotemporal evolution according to claim 1, characterized in that: The steps of acquiring the temporal geochemical data include: Divide the target area into several grids, evenly arrange sampling points in the grids, and collect sample data of the sampling points at set time intervals; After preprocessing, the sample data are arranged in chronological order to obtain time series geochemical data.

4. The method for monitoring heavy metals in mining areas based on spatiotemporal evolution according to claim 1, characterized in that: Calculate the heavy metal monitoring results, including: The heavy metal inversion results and heavy metal prediction results were integrated and input into Model 3; Model 3 dynamically adjusts the weights according to the heavy metal inversion results and heavy metal prediction results, and performs weighted summation based on the dynamically adjusted weights to obtain the final heavy metal monitoring results.

5. The method for monitoring heavy metals in mining areas based on spatiotemporal evolution according to claim 4, characterized in that: The weight coefficient is calculated by the weight adjustment function set in the model three; wherein the weight adjustment function is constructed based on the sigmoid function.

6. The method for monitoring heavy metals in mining areas based on spatiotemporal evolution according to claim 5, characterized in that: The weight adjustment function is: Where C is the heavy metal prediction result corresponding to the geochemical data, k is an adjustable parameter used to control the rate of weight change, θ is taken as 0.4, the weight of the heavy metal inversion result is w1 = f(C), and the weight of the heavy metal prediction result is w2 = 1-w1.

7. The method for monitoring heavy metals in mining areas based on spatiotemporal evolution according to claim 6, characterized in that: Introducing environmental quality indicators into the weight adjustment function, the weight adjustment function is Where α is the environmental quality index influence coefficient, which is used to adjust the influence of the environmental quality index E on the function; the environmental quality index E is used to evaluate the impact of environmental factors on hyperspectral data.

8. The method for monitoring heavy metals in mining areas based on spatiotemporal evolution according to claim 7, characterized in that: The calculation of the environmental quality indicators includes: Extracting environmental factors at a corresponding moment in the data set; wherein the environmental factors include aerosol thickness, water vapor content, ozone concentration, solar altitude angle, cloud cover, slope and aspect; Set the weight coefficient of each environmental factor, add the standardized results of the environmental factors and the weight coefficient to obtain the environmental quality index.

9. The method for monitoring heavy metals in mining areas based on spatiotemporal evolution according to claim 8, characterized in that: Extracting the environmental factors at the corresponding moment of the data group includes: Collect historical environmental data; Environmental factors of several data groups are matched from the historical environmental data according to spatiotemporal elements, and the matched environmental factors are associated with the data groups; wherein the spatiotemporal elements include the spatial coordinates and acquisition time of the data groups.

10. A mining area heavy metal monitoring system based on spatiotemporal evolution, implementing a mining area heavy metal monitoring method based on spatiotemporal evolution according to any one of claims 1 to 9, characterized in that: It includes a time-space monitoring module and a data acquisition module connected thereto; Data acquisition module: used to collect time-series hyperspectral data and time-series geochemical data of the target area, perform spatiotemporal alignment on the time-series hyperspectral data and the time-series geochemical data to obtain multiple spatiotemporal data groups; wherein the spatiotemporal data groups include hyperspectral data and geochemical data in the same spatiotemporal area, and the target area includes the mining area; Spatiotemporal monitoring module: used to input the hyperspectral data from the data sets into Model 1 to obtain heavy metal inversion results; input the geochemical data into Model 2 to obtain heavy metal prediction results; wherein Model 1 and Model 2 are constructed based on artificial intelligence models; and It is used to input the heavy metal inversion results and heavy metal prediction results into Model 3 to calculate the heavy metal monitoring results; extract the heavy metal monitoring results corresponding to several of the data groups; render several of the heavy metal monitoring results according to the time and space corresponding to several of the data groups to obtain a heavy metal spatiotemporal evolution diagram; wherein, a weight adjustment function is set in Model 3 for dynamically adjusting the weight.

Images