A soil microplastic detection method and system

Abstract

The present invention belongs to the technical field of soil detection, and provides a soil microplastic detection method and system, comprising: using a five-point sampling method to uniformly collect soil samples in a designated area, and then using a quartering method to reduce the collected samples to 5g; after mixing the soil sample with an extract, using a centrifuge to perform centrifugal separation to collect microplastic particles; using an infrared spectrometer to perform spectral analysis on the extracted microplastics; using a microscope in combination with dyeing or fluorescent labeling to observe and count the extracted microplastic particles; using a hyperspectral camera to obtain spatial image data of soil samples, and quickly identify and locate microplastics; processing and analyzing the collected spectral data and image data, extracting characteristic information of microplastics, and using a machine learning algorithm to train and predict a classification model to obtain soil microplastic detection data; the present invention extracts more comprehensive microplastic features and improves the intelligent level of soil microplastic detection.

Description

Technical Field

[0001] The present invention belongs to the technical field of soil detection, and specifically relates to a soil microplastic detection method and system. Background Art

[0002] In the current context of environmental protection and sustainable resource utilization, the monitoring and detection of microplastics has become an important research direction. Microplastics refer to plastic particles with a diameter of less than 5 mm, which are widely present in soil, water and air, posing a potential threat to the ecological environment and human health;

[0003] In a Chinese patent with publication number CN118443538A, a soil microplastic detection method and system are mentioned, including: obtaining image information and Fourier transform microscopic infrared spectrum of soil microplastics; constructing a microplastic physical feature detection model and a microplastic chemical composition detection model; performing graphic correction processing on the image information, combining the microplastic physical feature detection model and the image information after correction processing to obtain the physical characteristics of soil microplastics; performing spectral adjustment processing on the Fourier transform microscopic infrared spectrum, combining the microplastic chemical composition detection model and the Fourier transform microscopic infrared spectrum after spectral adjustment processing to obtain the chemical composition of soil microplastics; summarizing the physical characteristics and the chemical composition to obtain the detection data of soil microplastics, reducing the influence of human factors on the soil microplastic detection results and improving the accuracy and efficiency of soil microplastic detection.

[0004] However, it mainly relies on traditional laboratory analysis methods, such as optical microscopy observation, Fourier transform infrared spectroscopy (FTIR) analysis, and gas chromatography-mass spectrometry (GC-MS), which have significant disadvantages. First, traditional techniques usually require complex sample pretreatment, and the sampling and analysis process is time-consuming and cumbersome, resulting in low detection efficiency and difficulty in meeting the needs of large-scale detection. Secondly, laboratory analysis generally relies on professional equipment and experimenters, which increases the cost of detection, and the use of equipment limits the flexibility of analysis. In addition, traditional methods often require a large amount of chemical reagents, which not only increases the environmental burden, but also may pose a risk to the health of experimenters, especially in the case of long-term and high-frequency use; at the same time, the existing technology needs to be improved in the description and identification of microplastic characteristics. Traditional methods mainly rely on physical and chemical analysis methods, which are easily affected by the complexity of samples, making the feature identification of microplastics not comprehensive enough, which may lead to missed detection or false detection, further affecting the accuracy of the detection data. In addition, many traditional detection technologies cannot be effectively automated, and manual intervention introduces human errors, affecting the consistency and repeatability of the results.

[0005] To this end, technicians in this field have proposed a soil microplastic detection method and system, aiming to extract more comprehensive microplastic characteristics and improve the intelligence level of soil microplastic detection. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a soil microplastic detection method and system to solve the problems raised in the background technology.

[0007] According to the first aspect of the present disclosure, a method for detecting soil microplastics is proposed, comprising the following steps:

[0008] S1. Use the five-point sampling method to evenly collect soil samples in the designated area, then use the quartering method to reduce the collected samples to 5g, dry them, and grind and sieve the dried samples;

[0009] S2. Use saturated NaCl solution or ZnCl2 solution as the extracting solution, mix the soil sample with the extracting solution, and then use a centrifuge to perform centrifugal separation. According to the density difference between microplastics and soil particles, adjust the centrifugal speed and time to float or settle the microplastics, and collect the microplastic particles in the upper or lower layer of liquid;

[0010] S3. Use an infrared spectrometer to perform spectral analysis on the extracted microplastics, and identify and quantitatively analyze the types of microplastics by comparing the characteristic peaks of microplastics with the data in the standard library;

[0011] S4. Use a microscope and dye or fluorescent label to observe and count the extracted microplastic particles, and record their quantity and particle size distribution information;

[0012] S5. Use a hyperspectral camera to obtain spatial image data of soil samples, and quickly identify and locate microplastics by analyzing the spectral information in the image;

[0013] S6. Process and analyze the collected spectral data and image data, extract the characteristic information of microplastics, use machine learning algorithms to train and predict the classification model, and obtain the detection data of soil microplastics.

[0014] Preferably, in the step S2, for microplastics with a lower density, a saturated NaCl solution is used, the centrifugal speed is 3000-5000 rpm, and the centrifugal time is 5-10 minutes, so that the microplastics are floated to the upper liquid for collection;

[0015] For microplastics with higher density, ZnCl2 solution was used with a centrifugal speed of 8000-12000rpm and a centrifugal time of 10-15 minutes to allow the microplastics to settle into the lower layer of liquid for collection.

[0016] Preferably, the S3 infrared spectrum analysis operation step includes: S301, placing the microplastic particles obtained by centrifugal separation on a sample table, setting the scanning range of the infrared spectrometer to 4000-650cm -1 , resolution 4cm -1 , the number of scans is 32 times; S302, before testing the sample, clear the sample stage, run a complete scanning process, and record the background spectrum data; S303, place the sample on the sample stage, start the infrared spectrometer for scanning, record the absorption intensity of the sample at different wavelengths, and form a spectrum graph containing multiple data points; S304, process the collected spectral data, first perform baseline correction, then smoothing, and then normalize the spectral data; S305, compare the processed spectral data with the data in the known standard library, find the matching characteristic peaks, and determine the corresponding chemical bond vibration mode; S306, use Beer's law to calculate the content of microplastics, first establish a standard curve, that is, the relationship between the standard solution of known concentration and the intensity of its corresponding characteristic peak, and then use the standard curve to calculate the content of microplastics in the sample according to the intensity and area of ​​the sample characteristic peak, expressed as:

[0017]

[0018] Among them, A is absorbance, which indicates the ability of a substance to absorb light; and Represents the constant term and slope term obtained through linear regression analysis. By measuring the absorbance of standard solutions of known concentrations and drawing a standard curve, we can obtain and The value of; c represents the concentration of the absorbing substance, the unit is mol / L.

[0019] Preferably, the S4 step includes: S401, placing the microplastic particle sample processed by the S3 step on a glass slide, observing the stained sample with a high-power microscope, observing the fluorescent-labeled sample with a fluorescence microscope, and distinguishing different types of microplastic particles through filters; S402, during the observation process, counting the microplastic particles in the field of view one by one, repeatedly observing the same sample multiple times, and taking the average value as the final result, and recording its number; S403, classifying and counting according to the particle size range, and drawing a particle size distribution diagram.

[0020] Preferably, the S5 step includes: S501, using a hyperspectral camera to scan the soil sample, obtain its spatial image data, record the spectral information of each pixel, including reflectivity and absorptivity, and keep the distance and angle between the camera and the sample stable; S502, preprocessing the original spectral data, extracting the characteristic spectral information of microplastics, comparing the characteristic spectral information with the known microplastic spectral library, and quickly identifying and classifying microplastics; S503, after identifying the microplastics, using the position information in the spatial image data to locate the microplastics, and determining their specific positions in the soil sample by calculating the coordinate positions of the microplastics in the image; assuming that the image size obtained by the hyperspectral camera is W×H, the coordinates of each pixel are (x, y), where x represents the number of columns, i.e., the horizontal direction, and y represents the number of rows, i.e., the vertical direction. For each pixel P (x, y) identified as microplastics, its coordinate position is expressed as:

[0021]

[0022] Then, according to the known distance and angle between the camera and the sample and the resolution parameters of the camera, a transformation matrix M is defined to convert the image coordinates into actual spatial coordinates. For the pixel point P (x, y), its actual spatial coordinate S (X, Y, Z) is calculated by the following formula:

[0023]

[0024] Among them, M is a 3×2 transformation matrix, which is used to convert 2D image coordinates into 3D space coordinates; P is a column vector containing pixel coordinates, which calculates the actual position of each pixel identified as microplastic in the soil sample.

[0025] Preferably, in step S6, feature extraction is performed on the spectral data and the image data respectively, wherein the spectral data selects the absorption peak or reflection valley at a specific wavelength as a feature, and calculates the height, width, and position parameters of the absorption peak, or calculates the depth and area parameters of the reflection valley, which are used as input features of the machine learning algorithm; the image data selects the spatial position coordinates of the microplastics obtained in step S5, as well as the shape and texture features as input features;

[0026] The edge detection algorithm is used to extract the edge contour of microplastics. I is set to represent the input image. and Represent the gradients in the horizontal and vertical directions respectively, expressed as: , and set the threshold T1<T2, then the judgment condition of the edge point is or And G is a local maximum;

[0027] The texture analysis algorithm is used to extract the texture pattern of the microplastic surface. I is set to represent the input image. It represents the number of times the pixel pairs with gray values ​​i and j appear at a distance of d and a direction of θ, which is expressed as:

[0028]

[0029] in, Indicates that the image is at position The gray value at , W and H are the width and height of the input image, d represents the distance, θ represents the direction angle, , denote the cosine and sine of the angle θ respectively.

[0030] Preferably, the extracted spectral features are integrated to form a complete spectral feature vector, and the extracted image features are integrated to form a complete image feature vector, and then the spectral feature vector and the image feature vector are merged to form a complete feature vector;

[0031] Select the integrated features, select the most representative feature subset, and set the training data set to ,in represents the feature vector, Represents the label, where C is the number of categories, through the projection vector , so that the projected data satisfies the maximization of inter-class scatter and minimization of intra-class scatter, that is: ,in , denote the between-class scatter matrix and the within-class scatter matrix respectively, Represents the transpose of the projection vector.

[0032] Preferably, after the characteristic information of microplastics is extracted, the classification model is trained and predicted using a machine learning algorithm, the characteristic information is used as input data, propagated through each layer of the network, the output of each layer is calculated, and at the output layer of the network, the loss is calculated based on the predicted results and the true label, expressed as: , where N represents the number of samples and C represents the number of categories. represents the true label, represents the predicted probability, Represents the loss function; according to the gradient of the loss function, it propagates forward layer by layer starting from the output layer and updates the weight of each layer, which is expressed as:

[0033]

[0034] in, represents the weight, represents the learning rate, t represents the number of iterations, Represents partial derivatives; repeat the forward propagation and back propagation process until the loss function converges or reaches the preset number of iterations; then input the new unlabeled sample into the trained model, and obtain the predicted category of the new sample based on the output of the model.

[0035] According to the second aspect of the present disclosure, a soil microplastic detection system is proposed, comprising: an input device, an output device, a processor, and a memory, wherein the input device, the output device, the processor, and the memory are interconnected, and the memory comprises program instructions, and the program instructions are used to execute the steps of the soil microplastic detection method described in the first aspect of the present disclosure.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. The present invention can more comprehensively describe the characteristics of microplastics by extracting features from spectral data and image data and integrating these features. It also selects convolutional neural networks as machine learning algorithms, which can automatically learn and extract high-level features in images, thereby improving the accuracy and efficiency of detection and realizing intelligent detection of soil microplastics.

[0038] 2. Compared with traditional laboratory analysis methods, the detection method based on spectroscopy and image processing reduces the dependence on chemical reagents, thereby reducing the risk of environmental pollution. No disposable consumables are involved in the detection process, which is beneficial to environmental protection and sustainable utilization of resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a flow chart of the soil microplastic detection method of the present invention;

[0040] Figure 2 This is a framework diagram of the soil microplastic detection system of the present invention. DETAILED DESCRIPTION

[0041] The following embodiments of the present invention are described in further detail in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0042] As attached Figure 1 As shown:

[0043] Embodiment 1: The present invention provides a method for detecting soil microplastics, comprising the following steps:

[0044] S1. Use the five-point sampling method to evenly collect soil samples in the designated area, and then use the quartering method to reduce the collected samples to 5g to reduce the sample volume while maintaining representativeness, and then dry the dried samples, grind and sieve them to remove moisture to avoid interference during the experiment;

[0045] S2. Use saturated NaCl solution or ZnCl2 solution as the extracting solution, mix the soil sample with the extracting solution, and then use a centrifuge to centrifuge. According to the density difference between microplastics and soil particles, adjust the centrifugal speed and time to float or settle the microplastics, and collect the microplastic particles in the upper or lower liquid. For microplastics with lower density, use saturated NaCl solution, the centrifugal speed is 3000-5000rpm, and the centrifugal time is 5-10 minutes, so that the microplastics float to the upper liquid for collection; for microplastics with higher density, use ZnCl2 solution, the centrifugal speed is 8000-12000rpm, and the centrifugal time is 10-15 minutes, so that the microplastics settle to the lower liquid for collection. For specific types of microplastics, use oil extraction or pressurized fluid extraction technology to improve the extraction efficiency, in which the oil extraction method is to mix the soil sample with an oil solvent to dissolve the microplastic particles in the oil, and then separate the oil phase containing microplastics by filtering or centrifugation. Finally, wash the oil phase with anhydrous ethanol to remove the residual oil.

[0046] S3. Use an infrared spectrometer to perform spectral analysis on the extracted microplastics, and identify and quantitatively analyze the types of microplastics by comparing the characteristic peaks of the microplastics with the data in the standard library. The infrared spectral analysis operation steps include: S301. Place the microplastic particles obtained by centrifugation on the sample table, and set the scanning range of the infrared spectrometer to 4000-650cm -1 , resolution 4cm -1 , the number of scans is 32 times; S302, before testing the sample, clear the sample table, run a complete scanning process, and record the background spectrum data to eliminate the influence of environmental noise; S303, place the sample on the sample table, start the infrared spectrometer for scanning, record the absorption intensity of the sample at different wavelengths, and form a spectrum graph containing multiple data points; S304, process the collected spectral data, first perform baseline correction to eliminate the influence of background noise and baseline drift, then perform smoothing to reduce the interference of random noise, and then normalize the spectral data to better compare the differences between different samples; S305, compare the processed spectral data with the data in the known standard library, find the matching characteristic peaks, and determine the corresponding chemical bond vibration mode; S306, use Beer's law to calculate the content of microplastics, first establish a standard curve, that is, the relationship between the standard solution of known concentration and the intensity of its corresponding characteristic peak, and then calculate the content of microplastics in the sample based on the intensity and area of ​​the sample characteristic peak using the standard curve, expressed as:

[0047]

[0048] Among them, A is absorbance, which indicates the ability of a substance to absorb light; and Represents the constant term and slope term obtained through linear regression analysis. By measuring the absorbance of standard solutions of known concentrations and drawing a standard curve, we can obtain and The value of; c represents the concentration of the absorbing substance, the unit is mol / L, and then the absorbance of the sample to be tested is substituted into the above linear relationship to calculate the concentration of microplastics in the sample. For complex samples, microscopic infrared technology is used for in-situ monitoring.

[0049] S4. Use a microscope in combination with staining or fluorescent labeling to observe and count the extracted microplastic particles, and record their quantity and particle size distribution information; the specific steps include: S401. Place the microplastic particle sample processed in step S3 on a glass slide, and use a high-power microscope to observe the stained samples. The high-power microscope can more clearly observe the details of the microplastic particles. At the same time, adjust the light source brightness and aperture size to obtain the best observation effect; use a fluorescence microscope to observe the fluorescently labeled samples, and use filters to distinguish different types of microplastic particles; S402. During the observation process, count the microplastic particles in the field of view one by one, observe the same sample repeatedly, and take the average value as the final result, and record its quantity; S403. Use the ruler function of the microscope or image analysis software to measure the particle size of the microplastic particles, record the particle size of each particle, and classify and count according to the particle size range, and finally draw a particle size distribution diagram to intuitively display the particle size distribution of the microplastic particles.

[0050] S5. Use a hyperspectral camera to obtain spatial image data of soil samples, and quickly identify and locate microplastics by analyzing the spectral information in the image; specifically, S501. Use a hyperspectral camera to scan soil samples, obtain their spatial image data, and record the spectral information of each pixel, including reflectivity and absorptivity. These spectral information will be used for subsequent microplastic identification and positioning, and the distance and angle between the camera and the sample are kept stable to avoid distortion or blurred images; S502. Preprocess the original spectral data to extract the characteristic spectral information of microplastics, such as the absorption at a specific wavelength. Peaks or reflection valleys, these characteristic spectral information will be used as the basis for identifying microplastics, and the characteristic spectral information will be compared with the known microplastic spectral library to quickly identify and classify microplastics; S503, after identifying microplastics, the microplastics are located using the position information in the spatial image data, and their specific positions in the soil sample are determined by calculating the coordinate positions of the microplastics in the image. In order to improve the accuracy of positioning, multi-point positioning or regional positioning methods can be used to comprehensively consider the information of multiple pixel points; by assuming that the image size obtained by the hyperspectral camera is W×H, the coordinates of each pixel point are (x, y), where x represents the number of columns, i.e., the horizontal direction, and y represents the number of rows, i.e., the vertical direction. For each pixel point P(x, y) identified as microplastics, its coordinate position is expressed as:

[0051]

[0052] Then, according to the known distance and angle between the camera and the sample and the resolution parameters of the camera, a transformation matrix M is defined to convert the image coordinates into actual spatial coordinates. For the pixel point P (x, y), its actual spatial coordinate S (X, Y, Z) is calculated by the following formula:

[0053]

[0054] Among them, M is a 3×2 transformation matrix, which is used to convert 2D image coordinates into 3D space coordinates; P is a column vector containing pixel coordinates, which calculates the actual position of each pixel identified as microplastic in the soil sample.

[0055] S6. Process and analyze the collected spectral data and image data, extract the characteristic information of microplastics, use machine learning algorithms to train and predict the classification model, and obtain soil microplastic detection data. Feature extraction is performed on the spectral data and image data respectively, where the spectral data selects the absorption peak or reflection valley at a specific wavelength as a feature, and calculates the height, width, and position parameters of the absorption peak, or calculates the depth and area parameters of the reflection valley, and uses it as the input feature of the machine learning algorithm; if the image data contains color information, the color features of the microplastics are extracted, which helps to identify microplastic particles with different colors; the image data selects the spatial position coordinates of the microplastics obtained in step S5, as well as the shape and texture features as input features;

[0056] The edge detection algorithm is used to extract the edge contour of microplastics. I is set to represent the input image. and Represent the gradients in the horizontal and vertical directions respectively, expressed as: , and set the threshold T1<T2, then the judgment condition of the edge point is or And G is a local maximum;

[0057] The texture analysis algorithm is used to extract the texture pattern of the microplastic surface. I is set to represent the input image. It represents the number of times the pixel pairs with gray values ​​i and j appear at a distance of d and a direction of θ, which is expressed as:

[0058]

[0059] in, Indicates that the image is at position The gray value at , W and H are the width and height of the input image, d represents the distance, θ represents the direction angle, , denote the cosine and sine of the angle θ respectively.

[0060] Integrate the extracted spectral features to form a complete spectral feature vector, integrate the extracted image features to form a complete image feature vector, and then merge the spectral feature vector and the image feature vector to form a complete feature vector;

[0061] Select the integrated features, select the most representative feature subset, and set the training data set to ,in represents the feature vector, Represents the label, where C is the number of categories, through the projection vector , so that the projected data satisfies the maximization of inter-class scatter and minimization of intra-class scatter, that is: ,in , denote the between-class scatter matrix and the within-class scatter matrix respectively, Represents the transpose of the projection vector.

[0062] After the feature information of microplastics is extracted, the classification model is trained and predicted using a machine learning algorithm. The feature information is used as input data and propagated through each layer of the network to calculate the output of each layer. At the output layer of the network, the loss is calculated based on the predicted results and the true label, expressed as: , where N represents the number of samples and C represents the number of categories. represents the true label, represents the predicted probability, Represents the loss function; according to the gradient of the loss function, it propagates forward layer by layer starting from the output layer and updates the weight of each layer, which is expressed as:

[0063]

[0064] in, represents the weight, represents the learning rate, t represents the number of iterations, Represents partial derivative, which represents the derivative of a specific variable in a multivariable function; repeat the forward propagation and backpropagation process until the loss function converges or reaches a preset number of iterations; then input the new unlabeled sample into the trained model, and obtain the predicted category of the new sample based on the output of the model.

[0065] As attached Figure 2 As shown:

[0066] Embodiment 2: The present invention also provides a soil microplastic detection system, comprising: an input device, an output device, a processor, and a memory, wherein the input device, the output device, the processor, and the memory are interconnected, and the memory comprises program instructions, and the program instructions are used to execute the steps of the soil microplastic detection method described in Embodiment 1.

[0067] Additionally, in order to provide a concise description of exemplary embodiments, all features of an actual embodiment may not be described (ie, those features that are not relevant to the best mode presently contemplated for carrying out the invention or those that are not relevant to implementing the invention).

[0068] It should be understood that in the development of any actual implementation, as in any engineering or design project, numerous implementation-specific decisions may be made. Such a development effort may be complex and time-consuming, but for those of ordinary skill having the benefit of this disclosure, without undue experimentation, the development effort will be a routine task of design, fabrication, and production.

[0069] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. 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 solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for detecting soil microplastics, characterized in that: The following steps are involved: S1. Use the five-point sampling method to evenly collect soil samples in the designated area, then use the quartering method to reduce the collected samples to 5g, dry them, and grind and sieve the dried samples; S2. Use saturated NaCl solution or ZnCl2 solution as the extracting solution, mix the soil sample with the extracting solution, and then use a centrifuge to perform centrifugal separation. According to the density difference between microplastics and soil particles, adjust the centrifugal speed and time to float or settle the microplastics, and collect the microplastic particles in the upper or lower layer of liquid; S3. Use an infrared spectrometer to perform spectral analysis on the extracted microplastics, and identify and quantitatively analyze the types of microplastics by comparing the characteristic peaks of the microplastics with the data in the standard library; S4. Use a microscope and dye or fluorescent label to observe and count the extracted microplastic particles, and record their quantity and particle size distribution information; S5. Use a hyperspectral camera to obtain spatial image data of soil samples, and quickly identify and locate microplastics by analyzing the spectral information in the image; S6. Process and analyze the collected spectral information and spatial image data, extract the characteristic information of microplastics, use machine learning algorithms to train and predict the classification model, and obtain soil microplastic detection data; In the step S6, feature extraction is performed on the spectral information and the spatial image data respectively, wherein the spectral information selects the absorption peak or reflection valley at a specific wavelength as a feature, and calculates the height, width, and position parameters of the absorption peak, or calculates the depth and area parameters of the reflection valley, which are used as input features of the machine learning algorithm; The spatial image data selects the spatial position coordinates of the microplastics obtained in step S5, as well as the shape and texture features as input features; The edge detection algorithm is used to extract the edge contour of microplastics. I is set to represent the input image. and Represent the gradients in the horizontal and vertical directions respectively, expressed as , and set the threshold value T1<T2, then the judgment condition of the edge point is G>T2 or T1<G≤T2 and G is the local maximum; The texture analysis algorithm is used to extract the texture pattern of the microplastic surface. I is set to represent the input image. It represents the number of times the pixel pairs with gray values ​​i and j appear at a distance of d and a direction of θ, which is expressed as: ; in, Indicates that the image is at position The gray value at , W and H are the width and height of the input image, d represents the distance, θ represents the direction angle, , denote the cosine and sine of the angle θ respectively; The extracted spectral features are integrated to form a complete spectral feature vector, and the extracted image features are integrated to form a complete image feature vector, and then the spectral feature vector and the image feature vector are merged to form a complete feature vector; Select the integrated features, select the most representative feature subset, and set the training data set to ,in represents the feature vector, Represents the label, where C is the number of categories, through the projection vector , so that the projected data satisfies the maximization of inter-class scatter and minimization of intra-class scatter, that is: ,in , denote the between-class scatter matrix and the within-class scatter matrix respectively, represents the transpose of the projection vector; After the feature information of microplastics is extracted, the classification model is trained and predicted using a machine learning algorithm. The feature information is used as input data and propagated through each layer of the network to calculate the output of each layer. At the output layer of the network, the loss is calculated based on the predicted results and the true label, expressed as , where N represents the number of samples and C represents the number of categories. represents the true label, represents the predicted probability, L represents the loss function; according to the gradient of the loss function, it propagates forward layer by layer starting from the output layer, and updates the weight of each layer, which is expressed as: ; in, represents the weight, η represents the learning rate, t represents the number of iterations, Represents partial derivatives; repeat the forward propagation and back propagation process until the loss function converges or reaches the preset number of iterations; then input the new unlabeled sample into the trained model, and obtain the predicted category of the new sample based on the output of the model.

2. A soil microplastic detection method as claimed in claim 1, characterized in that: In the step S2, for microplastics with a lower density, a saturated NaCl solution is used, the centrifugal speed is 3000-5000 rpm, and the centrifugal time is 5-10 minutes, so that the microplastics are floated to the upper liquid for collection; For microplastics with higher density, ZnCl2 solution was used with a centrifugal speed of 8000-12000rpm and a centrifugal time of 10-15 minutes to allow the microplastics to settle into the lower layer of liquid for collection.

3. A soil microplastic detection method as claimed in claim 1, characterized in that: The infrared spectrum analysis operation steps of S3 include: S301, placing the microplastic particles obtained by centrifugal separation on a sample table, and setting the scanning range of the infrared spectrometer to 4000-650cm -1 , resolution 4cm -1 , the number of scans is 32 times; S302, before testing the sample, clear the sample stage, run a complete scanning process, and record the background spectrum data; S303, place the sample on the sample stage, start the infrared spectrometer for scanning, record the absorption intensity of the sample at different wavelengths, and form a spectrum graph containing multiple data points; S304, process the collected spectral data, first perform baseline correction, then smoothing, and then normalize the spectral data; S305, compare the processed spectral data with the data in the known standard library, find the matching characteristic peaks, and determine the corresponding chemical bond vibration mode; S306, use Beer's law to calculate the content of microplastics, first establish a standard curve, that is, the relationship between the standard solution of known concentration and the intensity of its corresponding characteristic peak, and then use the standard curve to calculate the content of microplastics in the sample according to the intensity and area of ​​the sample characteristic peak, expressed as: A=K'c+b' Among them, A is absorbance, which indicates the ability of a substance to absorb light; K' and b' represent the constant term and slope term obtained through linear regression analysis. The values ​​of K' and b' are obtained by measuring the absorbance of standard solutions of known concentrations and drawing a standard curve; c represents the concentration of the absorbing substance in mol / L.

4. A soil microplastic detection method as claimed in claim 1, characterized in that: The S4 step includes: S401, placing the microplastic particle sample processed by the S3 step on a glass slide, observing the stained sample with a high-power microscope, observing the fluorescent-labeled sample with a fluorescence microscope, and distinguishing different types of microplastic particles through filters; S402, during the observation process, counting the microplastic particles in the field of view one by one, repeatedly observing the same sample, and taking the average value as the final result, and recording its number; S403, classifying and counting according to the particle size range, and drawing a particle size distribution diagram.

5. A soil microplastic detection method as claimed in claim 1, characterized in that: The step S5 includes: S501, using a hyperspectral camera to scan the soil sample, obtain its spatial image data, record the spectral information of each pixel, including reflectivity and absorptivity, and keep the distance and angle between the camera and the sample stable; S502, preprocessing the original spectral data, extracting the characteristic spectral information of microplastics, comparing the characteristic spectral information with the known microplastic spectral library, and quickly identifying and classifying the microplastics; S503, after identifying the microplastics, using the position information in the spatial image data to locate the microplastics, and determining their specific positions in the soil sample by calculating the coordinate positions of the microplastics in the image; assuming that the image size obtained by the hyperspectral camera is W×H, the coordinates of each pixel are (x, y), where x represents the number of columns, i.e., the horizontal direction, and y represents the number of rows, i.e., the vertical direction. For each pixel P (x, y) identified as microplastics, its coordinate position is expressed as: P=(x,y) Then, according to the known distance and angle between the camera and the sample and the resolution parameters of the camera, a transformation matrix M is defined to convert the image coordinates into actual spatial coordinates. For the pixel point P (x, y), its actual spatial coordinate S (X, Y, Z) is calculated by the following formula: S=M·P Among them, M is a 3×2 transformation matrix, which is used to convert 2D image coordinates into 3D space coordinates; P is a column vector containing pixel coordinates, which calculates the actual position of each pixel identified as microplastic in the soil sample.

6. A soil microplastic detection system, characterized in that: include: An input device, an output device, a processor, and a memory, wherein the input device, the output device, the processor, and the memory are interconnected, and the memory includes program instructions, and the program instructions are used to execute the steps of the soil microplastic detection method according to any one of claims 1 to 5.

Images