Method and system for identifying micro-plastics in watershed water body

Abstract

The invention relates to the technical field of micro-plastic identification, in particular to a drainage basin water body micro-plastic identification method and system, which comprises a data acquisition and preprocessing module, a principal component analysis dimension reduction module, a K-means clustering optimization module, a micro-plastic classification and identification module and a result visualization module, the dimension reduction module is used for performing dimension reduction on a high-dimensional spectral data matrix and retaining key spectral features; the K-means clustering optimization module is used for carrying out clustering optimization based on the data after dimension reduction; according to the method, dimension reduction and noise and redundant information removal are carried out by means of principal component analysis to reduce calculation complexity, then efficient clustering optimization is carried out based on data after dimension reduction to reveal an internal structure of micro-plastic spectrum data, finally, classification and identification are carried out on various micro-plastics based on a clustering result, and the structure is visually displayed. The observation convenience of workers is greatly improved, and the micro-plastic identification efficiency and accuracy are also remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of microplastic identification technology, and in particular to a method and system for identifying microplastics in river basin water bodies. Background Art

[0002] Currently, microplastics are ubiquitous in the water environment. Microplastics refer to plastic particles with a diameter of less than 5 mm. These particles can be formed by the gradual breakup of large plastic waste, or they can be tiny particles formed during the production process. Microplastics have the characteristics of stable chemical structure and difficult to degrade. They can exist in the environment for a long time, resulting in microplastics causing serious harm to the environment. Therefore, it is necessary to use a spectrometer to identify microplastics in river basin water bodies, so as to assess the degree of pollution and distribution of microplastics in specific areas, which will help understand the source, migration path and impact range of microplastics, and facilitate relevant departments to take countermeasures.

[0003] However, in the aforementioned existing technologies, there are usually multiple types of microplastics in the water bodies of the basin, and the traditional spectrometer identification method can only identify microplastics, but cannot classify them efficiently and accurately, which affects the observation of microplastics by later staff. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for identifying microplastics in river water bodies, so as to solve the problem in the prior art that there are usually multiple types of microplastics in river water bodies, while the traditional spectrometer identification method can only identify microplastics but cannot classify them efficiently and accurately, which affects the observation of microplastics by later staff.

[0005] To achieve the above objectives, the present invention provides a watershed water microplastic identification system, comprising a data acquisition and preprocessing module, a principal component analysis dimensionality reduction module, a K-means clustering optimization module, a microplastic classification and identification module, and a result visualization module, wherein the data acquisition and preprocessing module, the principal component analysis dimensionality reduction module, the K-means clustering optimization module, the microplastic classification and identification module, and the result visualization module are connected in sequence;

[0006] The data acquisition and preprocessing module is used to collect spectral data of microplastic samples and perform data preprocessing to obtain a spectral data matrix;

[0007] The principal component analysis dimensionality reduction module is used to reduce the dimension of the high-dimensional spectral data matrix to retain key spectral features;

[0008] The K-means clustering optimization module is used to perform clustering optimization based on the data after dimensionality reduction;

[0009] The microplastic classification and identification module is used to classify and identify various microplastics based on clustering results;

[0010] The result visualization module is used to visualize the microplastic identification results in the watershed water body.

[0011] Wherein, the data acquisition and preprocessing module includes a spectral data acquisition unit, a denoising unit and a normalization unit, and the spectral data acquisition unit, the denoising unit and the normalization unit are connected in sequence;

[0012] The spectral data acquisition unit is used to collect spectral data of microplastic samples using a Raman spectrometer, covering 400-4000cm -1 Band;

[0013] The denoising unit is used to eliminate high-frequency noise using wavelet transform;

[0014] The normalization unit is used to standardize the spectrum intensity and eliminate baseline drift.

[0015] The normalization unit includes a normalization calculation subunit, a window selection subunit, a moving average calculation subunit, and a baseline subtraction subunit, and the normalization calculation subunit, the window selection subunit, the moving average calculation subunit, and the baseline subtraction subunit are connected in sequence;

[0016] The standardization calculation subunit is used to calculate the mean μ and standard deviation S of the original data set, and then substitute each original data point X into the Z-score standardization formula to calculate the corresponding Z score to obtain standardized spectral data;

[0017] The window selection subunit is used to select a suitable window size according to the sampling rate of the standardized spectral data and the change speed of the baseline drift;

[0018] The moving average calculation subunit is used to calculate the average value of the standardized spectrum data within a selected window as the baseline value within the window;

[0019] The baseline subtraction subunit is used to subtract the calculated baseline value from the standardized spectral data, thereby smoothing the spectral signal and eliminating the baseline drift, and obtaining a spectral data matrix.

[0020] Wherein, the principal component analysis dimensionality reduction module includes a calculation unit and a feature extraction unit, and the calculation unit and the feature extraction unit are connected;

[0021] The calculation unit is used to calculate the covariance matrix based on the spectral data matrix and extract eigenvalues and eigenvectors;

[0022] The feature extraction unit is used to select the principal component with a cumulative contribution rate of ≥95% and retain the key spectral features.

[0023] The K-means clustering optimization module includes a cluster number determination unit, an initial center optimization unit, and a dynamic iteration unit, and the cluster number determination unit, the initial center optimization unit, and the dynamic iteration unit are connected in sequence;

[0024] The cluster number determination unit is used to optimize the cluster number k using the silhouette coefficient;

[0025] The initial center optimization unit is used to initialize the cluster center using the K-means++ algorithm to avoid falling into a local optimum;

[0026] The dynamic iteration unit is used to iteratively update the cluster center based on the Euclidean distance until convergence.

[0027] Wherein, the microplastic classification and identification module includes a feature library establishment unit and a matching strategy unit, and the feature library establishment unit and the matching strategy unit are connected;

[0028] The feature library establishment unit is used to store the cluster centers of known types of microplastics into a database;

[0029] The matching strategy unit is used to calculate the distance between the sample to be tested and the center of each category by using the nearest neighbor method, and determine its final category.

[0030] The present invention also provides a method for identifying microplastics in water bodies in a river basin, which uses the above-mentioned water body microplastic identification system and includes the following steps:

[0031] Use Raman spectrometer to collect spectral data of microplastic samples;

[0032] performing denoising and normalization processing on the spectral data to obtain a spectral data matrix;

[0033] The pre-processed spectral data matrix is subjected to dimensionality reduction, key spectral features are retained, and dimensionality-reduced data is obtained;

[0034] Performing clustering optimization based on the dimensionality reduction data to obtain a clustering result;

[0035] Classifying and identifying a variety of microplastics based on the clustering results to obtain microplastic classification and identification results;

[0036] The microplastic classification and identification results of the water bodies in the basin are displayed visually.

[0037] The present invention provides a method and system for identifying microplastics in water bodies in a river basin. The data acquisition and preprocessing module is used to collect spectral data of microplastic samples and perform data preprocessing to obtain a spectral data matrix; the principal component analysis dimensionality reduction module is used to reduce the dimensionality of the high-dimensional spectral data matrix to retain key spectral features; the K-means clustering optimization module is used to perform clustering optimization based on the reduced dimensionality data; the microplastic classification and identification module is used to classify and identify multiple microplastics based on the clustering results; and the result visualization module is used to classify and identify multiple microplastics based on the clustering results.

[0038] Therefore, principal component analysis is relied upon to reduce dimensionality, remove noise and redundant information, and reduce computational complexity. Then, efficient clustering optimization is performed based on the reduced dimensionality data to reveal the intrinsic structure of the microplastic spectral data. Finally, based on the clustering results, various microplastics are classified and identified, and the structure is visualized. Compared with traditional identification methods, this greatly improves the convenience of observation for staff, and the efficiency and accuracy of identifying microplastics are also significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present application 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.

[0040] Figure 1 It is a schematic diagram of the principle of the watershed water microplastic identification system of the present invention.

[0041] Figure 2 It is a schematic diagram of the data acquisition and preprocessing module of the present invention.

[0042] Figure 3 It is a schematic diagram of the normalization unit of the present invention.

[0043] Figure 4 It is a schematic diagram of the principal component analysis dimensionality reduction module of the present invention.

[0044] Figure 5 It is a schematic diagram of the K-means clustering optimization module of the present invention.

[0045] Figure 6 This is a schematic diagram of the microplastic classification and identification module of the present invention.

[0046] Figure 7 It is a flowchart of the steps of the method for identifying microplastics in water bodies of a river basin of the present invention.

[0047] 1-Data acquisition and preprocessing module, 101-Spectral data acquisition unit, 102-Noise removal unit, 103-Normalization unit, 1031-Standardization calculation subunit, 1032-Window selection subunit, 1033-Calculation of moving average subunit, 1034-Baseline subtraction subunit, 2-Principal component analysis dimensionality reduction module, 201-Calculation unit, 202-Feature extraction unit, 3-K-means clustering optimization module, 301-Cluster number determination unit, 302-Initial center optimization unit, 303-Dynamic iteration unit, 4-Microplastic classification and identification module, 401-Establishment of feature library unit, 402-Matching strategy unit, 5-Result visualization module. DETAILED DESCRIPTION

[0048] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, but should not be understood as limiting the present invention.

[0049] See also Figures 1 to 6 The present invention provides a watershed water microplastic identification system, which specifically includes:

[0050] The data acquisition and preprocessing module 1 is used to collect spectral data of microplastic samples and perform data preprocessing to obtain a spectral data matrix;

[0051] Specifically include:

[0052] The spectral data acquisition unit 101 is used to collect spectral data of microplastic samples using a Raman spectrometer, covering 400-4000cm -1 Band;

[0053] Raman spectroscopy is a molecular fingerprint spectroscopy technology that can provide information on molecular vibration and rotational energy level transitions, thereby reflecting the chemical composition and structural characteristics of microplastics. Due to the small size of microplastic particles, ordinary portable Raman laser spots may not be able to collect the signal of individual microplastics. However, in the range of 400 to 4000 cm -1 Within this wavelength range, high-resolution Raman spectrometers, such as portable scientific Raman spectrometers with micron-scale laser spot sizes, can accurately identify even the smallest microplastic particles. Furthermore, the area scanning mode of a microscopic Raman spectrometer can automatically collect signals point by point across the sample area, significantly improving detection efficiency and accuracy.

[0054] The denoising unit 102 is used to eliminate high-frequency noise using wavelet transform;

[0055] The wavelet transform has excellent time-frequency localization properties and can decompose signals into wavelet coefficients of different frequency components. In high-frequency bands, noise typically appears as random, irregular wavelet coefficients, while signals appear as wavelet coefficients with a certain regularity and structure. The wavelet transform can effectively distinguish signals from noise, providing a foundation for subsequent denoising.

[0056] The normalization unit 103 is used to standardize the spectrum intensity and eliminate baseline drift.

[0057] Specifically include:

[0058] The standardization calculation subunit 1031 is used to calculate the mean μ and standard deviation S of the original data set, and then substitute each original data point X into the Z-score standardization formula to calculate the corresponding Z score to obtain the standardized spectral data;

[0059] The formula for Z-score normalization is: Z = (X - μ) / S, where Z represents the normalized data (also known as the Z score or standard score). X represents the original data point. μ represents the mean of the original dataset. S represents the standard deviation of the original dataset. After Z-score normalization, the mean of the new dataset is 0 and the variance is 1. This gives the data a statistically uniform scale, facilitating subsequent analysis and processing. Z-score normalization eliminates differences in units and magnitude of the data, making comparisons between different features or datasets more fair and meaningful.

[0060] The window selection subunit 1032 is used to select a suitable window size according to the sampling rate of the standardized spectral data and the change speed of the baseline drift;

[0061] By observing the trends and fluctuations in the data, we can understand the approximate situation of the sampling rate and baseline drift of the normalized spectral data. If the baseline drift changes rapidly, we need to choose a smaller window to capture the changes more accurately; if the baseline drift changes slowly, we can choose a larger window to smooth the signal.

[0062] The moving average calculation subunit 1033 is used to calculate the average value of the standardized spectrum data in the selected window as the baseline value in the window;

[0063] According to the selected window size, continuous data points are selected from the normalized spectral data, and then the selected data points are averaged to obtain the baseline value within the window. Finally, the window is moved to the right by one data point, and the above steps are repeated until all the data are processed.

[0064] The baseline subtraction subunit 1034 is used to subtract the calculated baseline value from the standardized spectral data, thereby smoothing the spectral signal and eliminating baseline drift, and obtaining a spectral data matrix.

[0065] Each data point is matched with its corresponding baseline value, and then its corresponding baseline value is subtracted from each data point to obtain smoothed spectral data. Finally, the smoothed spectral data is checked to ensure that the baseline drift has been effectively eliminated while retaining the main characteristics of the spectral signal.

[0066] The principal component analysis dimensionality reduction module 2 is used to reduce the dimension of the high-dimensional spectral data matrix to retain key spectral features;

[0067] Specifically include:

[0068] The calculation unit 201 is used to calculate the covariance matrix based on the spectral data matrix and extract eigenvalues and eigenvectors;

[0069] Let the spectral data matrix be X. The covariance matrix is used to measure the linear relationship between the features in the spectral data matrix. Assuming that the spectral data matrix X has n samples and d features, the steps are as follows:

[0070] First, center the data: subtract the mean of each feature to make the mean 0. The formula is X centered =X-μ, where u is the mean vector of each feature.

[0071] Then, calculate the covariance matrix: The covariance matrix Σ is a d×d symmetric matrix, and the element Σij represents the covariance of features i and j. The formula is

[0072] Secondly, the covariance matrix Σ is eigendecomposed to obtain the eigenvalue λ and the corresponding eigenvector ui. The formula is Σui=λiui, where the eigenvalue λi represents the variance of the principal component and the eigenvector ui represents the direction of the principal component. Finally, the eigenvalues are sorted from large to small, and the top k largest eigenvalues and their corresponding eigenvectors are selected as the principal components after dimensionality reduction.

[0073] The feature extraction unit 202 is used to select the principal components with cumulative contribution rate ≥ 95% and retain the key spectral features.

[0074] First, calculate the contribution rate of each principal component: the eigenvalue λi represents the variance of each principal component, the contribution rate is the ratio of a single eigenvalue to the total eigenvalue, and the cumulative contribution rate is the sum of the contribution rates of the first k principal components. The formula is contribution rate Among them, λ i is the i-th eigenvalue, d is the original number of features;

[0075] Secondly, calculate the cumulative contribution rate: Cumulative contribution rate Contribution rate i, where the cumulative contribution rate represents the amount of original data information retained by the first k principal components.

[0076] After the calculation is complete, the eigenvalues are sorted from largest to smallest, and the corresponding eigenvectors are recorded. Starting from the first principal component, the contribution rate is gradually accumulated until the cumulative contribution rate is ≥95%. The minimum k value that satisfies the cumulative contribution rate is found. Typically, k is reduced to 10-20 dimensions. After selecting the first k principal components, the centralized data is projected onto these principal components to obtain the reduced-dimensional data, which are the key spectral features. The reduced-dimensional data is Y(n×k dimensions), where n is the number of samples and k is the number of selected principal components. This removes noise and redundant information from the spectral data, reduces computational complexity, and provides low-dimensional data for subsequent clustering and visualization.

[0077] The K-means clustering optimization module 3 is used to perform clustering optimization based on the data after dimensionality reduction;

[0078] Specifically include:

[0079] The cluster number determination unit 301 is used to optimize the cluster number k using the silhouette coefficient;

[0080] K is a parameter that needs to be set in advance, which indicates how many clusters (or categories) you want to divide the data into. First, run the K-means algorithm for different k values, then calculate the silhouette coefficient of each sample and the average silhouette coefficient, and finally select the k value that maximizes the average silhouette coefficient to determine the optimal number of clusters.

[0081] The initial center optimization unit 302 is used to initialize the cluster center using the K-means++ algorithm to avoid falling into a local optimum;

[0082] Randomly select the first cluster center from the data Y after PCA dimensionality reduction, and then calculate the closest distance D(x) between each sample and the selected center. Select the next center, and finally, repeat the above steps until k initial centers are selected.

[0083] The dynamic iteration unit 303 is used to iteratively update the cluster center based on the Euclidean distance until convergence.

[0084] First, each sample is assigned to the nearest cluster center. The specific formula is: Among them, x is the sample after PCA dimensionality reduction, μi is the i-th cluster center; secondly, recalculate the center of each cluster, the specific formula is: The algorithm stops when the cluster center no longer changes or the change is less than a preset threshold. The cluster center reveals the intrinsic structure of the data and provides representative characteristics of each category. The samples are divided into several clusters, each cluster represents a potential microplastic category, so as to identify potential microplastic categories and provide preliminary category division for subsequent classification and identification.

[0085] The microplastic classification and identification module 4 is used to classify and identify various microplastics based on the clustering results;

[0086] Specifically include:

[0087] The feature library establishment unit 401 is used to store the cluster centers of known types of microplastics into a database;

[0088] After storing the cluster center in the database, the cluster center information of a specific type of microplastic can be quickly retrieved through the database query function; during operation: through the method of the K-means clustering optimization module 3, K-means clustering is performed on each known type of microplastic to obtain the cluster center of each category, which is specifically classified as follows: clustering PE samples to obtain the cluster center μPE of PE; clustering PP samples to obtain the cluster center μPP of PP; clustering PS samples to obtain the cluster center μPS of PS; thereby, the cluster center of each category is stored in the feature library unit as a reference standard for classification.

[0089] The matching strategy unit 402 is used to calculate the distance between the sample to be tested and the center of each category using the nearest neighbor method, and determine its final category.

[0090] When using the nearest neighbor (1-NN) method: First, calculate the distance. Let the sample to be tested be x, and calculate its distance to the center of each category in the feature library. The specific formula is: dPE = ||x-μPE||2, dPP = ||x-μPP||2, dPS = ||x-μPS||2; calculate the microplastics PE, PP, and PS in turn, and then select the category with the smallest distance to determine the final category. Among them, μi is the cluster center of the i-th category; then select the category with the closest distance as the category of the sample to be tested, thus completing the category determination.

[0091] The result visualization module 5 is used to visualize the microplastic identification results in the watershed water body.

[0092] First, the clustering results are displayed through a 3D scatter plot, and the distribution of samples in the first three principal component spaces can be intuitively observed. Assuming that the data Y after PCA dimensionality reduction has k principal components, the first three principal components are extracted, and then a 3D scatter plot is drawn using Python. Different categories are represented by different colors for staff to observe; secondly, when the system generates a test report: first, the type of sample to be tested determined by the matching strategy unit 402 is recorded; then, for the concentration of microplastics in the watershed water, if the data contains concentration information, it can be output directly; otherwise, it can be indirectly represented by the number of samples or spectral intensity; finally, the confidence is calculated using distance or weight and recorded in the report; all type, concentration and confidence information are organized into tables to provide researchers or decision makers with detailed test data to support subsequent actions.

[0093] See also Figure 7 The present invention also provides a method for identifying microplastics in water bodies in a river basin, comprising the following steps:

[0094] S1: Collect spectral data of microplastic samples using Raman spectrometer;

[0095] S2: performing denoising and normalization processing on the spectral data to obtain a spectral data matrix;

[0096] S3: performing dimension reduction on the pre-processed spectral data matrix to retain key spectral features and obtain dimension-reduced data;

[0097] S4: performing clustering optimization based on the dimensionality reduction data to obtain a clustering result;

[0098] S5: Classifying and identifying the various microplastics based on the clustering results to obtain microplastic classification and identification results;

[0099] S6: Visually display the microplastic classification and identification results of the water body in the basin.

[0100] Among them, a Raman spectrometer is used to collect spectral data of microplastic samples; the spectral data is denoised and normalized to obtain a spectral data matrix; the pre-processed spectral data matrix is subjected to dimensionality reduction to retain key spectral features to obtain reduced-dimensional data; clustering optimization is performed based on the reduced-dimensional data to obtain clustering results; based on the clustering results, various microplastics are classified and identified to obtain microplastic classification and identification results; and the microplastic classification and identification results of the water body in the basin are visualized.

[0101] The above disclosure is merely one or more preferred embodiments of the present application and is not intended to limit the scope of the present application. A person skilled in the art will understand that all or part of the processes of the above embodiments and equivalent changes made in accordance with the claims of the present application are still within the scope of the present application.

Claims

1. A watershed water microplastic identification system, characterized in that: It includes a data acquisition and preprocessing module, a principal component analysis and dimensionality reduction module, a K-means clustering optimization module, a microplastic classification and identification module, and a result visualization module, wherein the data acquisition and preprocessing module, the principal component analysis and dimensionality reduction module, the K-means clustering optimization module, the microplastic classification and identification module, and the result visualization module are connected in sequence; The data acquisition and preprocessing module is used to collect spectral data of microplastic samples and perform data preprocessing to obtain a spectral data matrix; The principal component analysis dimensionality reduction module is used to reduce the dimension of the high-dimensional spectral data matrix to retain key spectral features; The K-means clustering optimization module is used to perform clustering optimization based on the data after dimensionality reduction; The microplastic classification and identification module is used to classify and identify various microplastics based on clustering results; The result visualization module is used to visualize the microplastic identification results in the watershed water body.

2. The watershed water microplastic identification system according to claim 1, characterized in that: The data acquisition and preprocessing module includes a spectral data acquisition unit, a denoising unit and a normalization unit, and the spectral data acquisition unit, the denoising unit and the normalization unit are connected in sequence; The spectral data acquisition unit is used to collect spectral data of microplastic samples using a Raman spectrometer, covering 400-4000cm -1 Band; The denoising unit is used to eliminate high-frequency noise using wavelet transform; The normalization unit is used to standardize the spectrum intensity and eliminate baseline drift.

3. The watershed water microplastic identification system according to claim 2, characterized in that: The normalization unit includes a normalization calculation subunit, a window selection subunit, a moving average calculation subunit, and a baseline subtraction subunit, wherein the normalization calculation subunit, the window selection subunit, the moving average calculation subunit, and the baseline subtraction subunit are connected in sequence; The standardization calculation subunit is used to calculate the mean μ and standard deviation S of the original data set, and then substitute each original data point X into the Z-score standardization formula to calculate the corresponding Z score to obtain standardized spectral data; The window selection subunit is used to select a suitable window size according to the sampling rate of the standardized spectral data and the change speed of the baseline drift; The moving average calculation subunit is used to calculate the average value of the standardized spectrum data within a selected window as the baseline value within the window; The baseline subtraction subunit is used to subtract the calculated baseline value from the standardized spectral data, thereby smoothing the spectral signal and eliminating the baseline drift, and obtaining a spectral data matrix.

4. The watershed water microplastic identification system according to claim 3, characterized in that: The principal component analysis dimensionality reduction module includes a calculation unit and a feature extraction unit, and the calculation unit and the feature extraction unit are connected; The calculation unit is used to calculate the covariance matrix based on the spectral data matrix and extract eigenvalues and eigenvectors; The feature extraction unit is used to select the principal component with a cumulative contribution rate of ≥95% and retain the key spectral features.

5. The watershed water microplastic identification system according to claim 4, characterized in that: The K-means clustering optimization module includes a cluster number determination unit, an initial center optimization unit, and a dynamic iteration unit, wherein the cluster number determination unit, the initial center optimization unit, and the dynamic iteration unit are connected in sequence; The cluster number determination unit is used to optimize the cluster number k using the silhouette coefficient; The initial center optimization unit is used to initialize the cluster center using the K-means++ algorithm to avoid falling into a local optimum; The dynamic iteration unit is used to iteratively update the cluster center based on the Euclidean distance until convergence.

6. The watershed water microplastic identification system according to claim 5, characterized in that: The microplastic classification and identification module includes a feature library establishment unit and a matching strategy unit, and the feature library establishment unit and the matching strategy unit are connected; The feature library establishment unit is used to store the cluster centers of known types of microplastics into a database; The matching strategy unit is used to calculate the distance between the sample to be tested and the center of each category by using the nearest neighbor method, and determine its final category.

7. A method for identifying microplastics in water bodies in a river basin, using the water body microplastic identification system according to claim 6, characterized in that: The steps include: Use Raman spectrometer to collect spectral data of microplastic samples; performing denoising and normalization processing on the spectral data to obtain a spectral data matrix; The pre-processed spectral data matrix is subjected to dimensionality reduction, key spectral features are retained, and dimensionality-reduced data is obtained; Performing clustering optimization based on the dimensionality reduction data to obtain a clustering result; Classifying and identifying a variety of microplastics based on the clustering results to obtain microplastic classification and identification results; The microplastic classification and identification results of the water bodies in the basin are displayed visually.

Images