Method for identifying heavy metal binding form on soil particles based on image superposition and machine learning

Abstract

The invention discloses a method for identifying the binding form of heavy metals on soil particles based on image superposition and machine learning. The method comprises the following steps: S1, establishing a method for classifying soil particles by taking composition elements as standards; s2, eliminating background noise of an element distribution situation graph obtained by the electron microscope energy spectrum; s3, the types of soil particles under the scanning electron microscope are determined, a plurality of soil particles are contained in an image of the scanning electron microscope, boundaries among the particles are obvious, and the soil particles are generally analyzed one by one in sequence; and S4, calculating the distribution condition of the heavy metals among the soil particle components. According to the method, the change of the biotoxicity of the Cu in the soil after the exogenous microorganisms are added can be indicated by a lossless and in-situ method, and the adsorption condition of the Cu in the heavy metal contaminated soil on each component after the exogenous microorganisms are added can be analyzed, so that the influence generated by the Cu can be obtained; the method has great potential for reducing the heavy metal biotoxicity of the polluted soil, and has great significance for follow-up guide practice.

Description

Technical Field

[0001] The present invention relates to the technical field of ecological risk assessment, and in particular to a method for identifying the binding morphology of heavy metals on soil particles based on image superposition and machine learning. Background Art

[0002] With the process of human modernization, the degree of industrialization has been continuously improving, resulting in increasingly serious soil pollution. The types of polluted soils are numerous, widely distributed, and large in quantity. Therefore, soil pollution is a problem that needs to be solved urgently.

[0003] Heavy metal pollution is a common phenomenon of soil pollution. Heavy metals combine with various solids in the soil to form complexes of different forms with different biological toxicities. Heavy metals in the soil will change their distribution, content and binding state due to changes in the physical and chemical properties of the soil and the activity of microorganisms due to environmental changes. That is, the heavy metals undergo a migration and transformation process.

[0004] Different forms of heavy metals cause great differences in biological toxicity, so it is crucial to study the existence forms of heavy metals in the soil. Previous studies have mostly used traditional chemical extraction methods to study heavy metal pollution and its forms. The defect of this experimental method is that it will destroy the heterogeneity of the soil itself and cannot reflect its original structure. It is questionable whether the morphological information obtained from the test can reflect the true situation of the soil. Therefore, the traditional method of using chemical reagents to treat the soil and then detect it has great defects, and it is urgently needed to improve or create new methods. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings existing in the prior art, such as: the traditional chemical extraction method for studying heavy metal pollution and its morphology. The defect of this experimental method is that it will destroy the heterogeneity of the soil itself and cannot reflect its original structure. It is questionable whether the morphological information obtained by detection can reflect the true situation of the soil. Therefore, the traditional method of using chemical reagents to treat the soil and then detect it has great defects, and it is urgently needed to improve or generate new methods. A method for identifying the binding morphology of heavy metals on soil particles based on image superposition and machine learning is proposed.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: A method for identifying the binding morphology of heavy metals on soil particles based on image overlay and machine learning, comprising the following steps: S1. Establish a method to classify soil particles based on their constituent elements; S2, eliminating the background noise of the element distribution diagram obtained by electron microscope energy spectrum; S3. Determine the types of soil particles under a scanning electron microscope. The image of a scanning electron microscope contains several soil particles with obvious boundaries between particles. Generally, they are analyzed one by one in order. S4. Calculate the distribution of heavy metals among various soil particle components.

[0007] Preferably, the soil contains a variety of different types of particles, mainly silicate minerals, iron oxides, organic matter, carbonates, etc., which cannot be distinguished by observing the morphology presented by an electron microscope, and need to be distinguished in combination with the element information presented by the electron microscope energy spectrum; the principle of using element information for distinction is that the elemental composition of different particles is quite different. The present invention classifies different particles according to the elements contained: silicate minerals - Si, Al, O, iron oxides - Fe, O, organic matter - C, O, carbonates - C, O, Ca; wherein silicate minerals are subdivided, and common silicate minerals in soil include kaolinite (1:1 type) and montmorillonite (2:1 type). Although the main constituent elements of both are Si, Al, O, the Si / Al ratios are different; by measuring standard samples of kaolinite and montmorillonite, a large number of Si / Al ratios of the two types of standard samples are obtained, and machine learning cluster analysis is used to find the Si / Al boundary value to distinguish the two types of silicate minerals.

[0008] Preferably, the electron microscope energy spectrum can measure the distribution of multiple elements at the same time, and generate corresponding images according to different elements, which show the distribution and content of elements in the form of pixels. The presence of bright pixels indicates that the element is distributed at this position, and the more pixels and the brighter the pixels are, the higher the content of the element is. However, the electron microscope energy spectrum itself has large instrument noise, which will seriously affect the presentation of the true situation of the elements. The main reason is that when the electron microscope energy spectrum is measured, the noise (background noise) generated by the substrate on which the sample is placed is too large. In order to eliminate the noise generated, the machine learning median method is used to determine the brightness threshold between the real pixel and the background pixel, and the pixels below the threshold are removed, and the image obtained after removal is regarded as the true image of the element distribution and content.

[0009] Preferably, the position of the analyzed particle is determined under the scanning electron microscope image, and its corresponding position in the electron microscope energy spectrum image is found. After removing the noise according to the second step, the element distribution at this position is observed. If Si, Al, and O are distributed in large quantities at this position, the particle is determined to be a silicate particle. If Fe and O are distributed in large quantities at this position, the particle is determined to be an iron oxide.... Among various types of soil particles, the properties of iron oxides are relatively active and tend to combine with other soil particle components to form a complex. Therefore, when a particle contains Si, Al, O, and Fe at the same time, it is determined to be a silicate-iron oxide complex, and the same applies to organic matter-iron oxide complexes.

[0010] Preferably, after the second step of noise reduction processing is performed on the desired element image, the obtained element distribution image is superimposed with the scanning electron microscope image of each particle type that has been determined after the third step, and the proportion of pixel points of the desired element falling on each component is calculated. The specific formula is (taking the organic matter-bound state as an example): organic matter-bound heavy metals = (heavy metal signal points falling on the organic matter component / total heavy metal signal points) * 100%.

[0011] Compared with the prior art, the beneficial effects of the present invention are: it can indicate the change of Cu biotoxicity in soil after adding exogenous microorganisms by a non-destructive, in-situ method, and can analyze the adsorption of Cu on various components in heavy metal contaminated soil after adding exogenous microorganisms, so as to obtain the impact that can be produced, which has great potential for reducing the biotoxicity of heavy metals in contaminated soil and is of great significance for guiding subsequent practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A scanning electron microscope image of a sample measured by a method for identifying the binding morphology of heavy metals on soil particles based on image superposition and machine learning proposed by the present invention; Figure 2 The electron microscope energy spectrum of the sample measured by the method for identifying the binding morphology of heavy metals on soil particles based on image superposition and machine learning proposed by the present invention; Figure 3 A diagram showing the dynamic changes in Cu adsorption in site soil after adding high concentration of Bacillus subtilis, according to a method for identifying the binding morphology of heavy metals on soil particles based on image overlay and machine learning proposed by the present invention; Figure 4 A diagram showing the dynamic changes in Cu adsorption in site soil after adding high concentrations of Escherichia coli, based on a method for identifying the binding morphology of heavy metals on soil particles based on image overlay and machine learning proposed by the present invention; Figure 5 A diagram showing the dynamic changes in Cu adsorption in site soil after adding low concentration of Bacillus subtilis, according to a method for identifying the binding morphology of heavy metals on soil particles based on image overlay and machine learning proposed by the present invention; Figure 6 This is a diagram of the dynamic changes in Cu adsorption in site soil after low concentration of Escherichia coli is added, based on a method proposed in the present invention for identifying the binding forms of heavy metals on soil particles based on image overlay and machine learning. DETAILED DESCRIPTION

[0013] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0014] Reference Figure 1-Figure 6 , a method for identifying the binding morphology of heavy metals on soil particles based on image superposition and machine learning, comprising the following steps: S1. Establish a method to classify soil particles based on their constituent elements; S2, eliminating the background noise of the element distribution diagram obtained from the electron microscope energy spectrum; S3. Determine the types of soil particles under a scanning electron microscope. The image of a scanning electron microscope contains several soil particles with obvious boundaries between particles. Generally, they are analyzed one by one in order. S4. Calculate the distribution of heavy metals among various soil particle components; Soil contains many different types of particles, mainly silicate minerals, iron oxides, organic matter, carbonates, etc. They cannot be distinguished by observing the morphology presented by an electron microscope, and need to be distinguished in combination with the element information presented by the electron microscope energy spectrum; the principle of distinguishing by element information is that the elemental composition of different particles is quite different. The present invention classifies different particles according to the elements contained: silicate minerals - Si, Al, O, iron oxides - Fe, O, organic matter - C, O, carbonates - C, O, Ca; silicate minerals are subdivided, and common silicate minerals in soil are kaolinite (1:1 type) and montmorillonite (2:1 type). Although the main components of these two are Si, Al, O, the Si / Al ratios are different; by measuring standard samples of kaolinite and montmorillonite, a large number of Si / Al of the two types of standard samples are obtained, and machine learning cluster analysis is used to find the Si / Al boundary value to distinguish these two types of silicate minerals; The electron microscope energy spectrum can measure the distribution of multiple elements at the same time, and generate corresponding images according to different elements. It shows the distribution and content of elements in the form of pixels. The existence of bright pixels indicates that the element is distributed at this position, and the more pixels and the brighter the pixels are, the higher the content of the element is. However, the electron microscope energy spectrum itself has a large instrument noise, which will seriously affect the presentation of the true situation of the elements. The main reason is that when measuring the electron microscope energy spectrum, the noise (background noise) generated by the substrate on which the sample is placed is too large. In order to eliminate the noise generated, the machine learning median method is used to determine the brightness threshold between the real pixel and the background pixel, and the pixels below the threshold are removed. The image obtained after removal is regarded as the true image of the element distribution and content. Determine the position of the analyzed particle under the scanning electron microscope image, find its corresponding position in the electron microscope energy spectrum image, and observe the element distribution at this position after removing the noise according to the second step. If Si, Al, and O are distributed in large quantities at this position, the particle is determined to be a silicate particle. If Fe and O are distributed in large quantities at this position, the particle is determined to be an iron oxide...; among various soil particles, the properties of iron oxides are relatively active and tend to combine with other soil particle components to form a complex; therefore, when a particle contains Si, Al, O, and Fe at the same time, it is determined to be a silicate-iron oxide complex, and the same is true for organic matter-iron oxide complex; After the second step of noise reduction processing on the required element image, the obtained element distribution image is superimposed with the scanning electron microscope image of each particle type that has been determined after the third step, and the proportion of pixel points of the required element falling on each component is calculated. The specific formula is (taking the organic matter-bound state as an example): Organic matter-bound heavy metals = (heavy metal signal points falling on the organic matter component / total heavy metal signal points) * 100%.

[0015] In the present invention, the preliminary preparation and measurement work are firstly carried out: the soil is first naturally air-dried, evenly ground, and passed through a 100-mesh sieve, then 1g of soil sample is taken and added to a 25ml centrifuge tube, 10ml of deionized water is added, and the soil suspension is fully vortexed and dispersed; then the soil suspension is ultrasonically separated and treated, the ultrasonic frequency is 40kHz, and the time is 5min, so that the soil particles are fully dispersed, and then the centrifuge tube is centrifuged at 4000r / min for 15min, the supernatant is discarded, and the soil is placed in a 60-degree Celsius oven for drying, and after drying, it is fully ground for standby use, and then a trace amount (<0.1g) of soil sample is taken, dispersed with anhydrous ethanol, placed on a pure molybdenum sheet (substrate), fixed with glue, and subjected to gold spraying treatment, and then after the treatment is completed, the SEM-EDS measurement is carried out on the machine, and the magnification is set to 30,000 times, and the measurement energy is 20keV; Secondly, the classification method of silicate minerals was established: first, a large number of kaolinite and montmorillonite standard samples were measured using the above measurement method, and the images of scanning electron microscope and electron microscope energy spectrum were respectively imported into image J image processing software, and the region selection tool was used to select the soil single particle in the scanning electron microscope image, and the corresponding region of the energy spectrum image was found. The size of the region and the intensity of Si and Al elements in the region were measured to obtain the size of the particle and calculate the Si / Al of the particle. Then, this process was performed for all particles in each image, and the size and Si / Al data of all particles were integrated. This data set was imported into Matlab software for machine learning cluster analysis. Two cluster centers were set, with Si / Al as the X-axis and particle size as the Y-axis. Then, the machine learning algorithm could be used to classify the two types of silicate minerals into their respective cluster centers in this coordinate system to complete the classification process. In this way, kaolinite and montmorillonite can be classified according to the difference in Si / Al by machine learning. Then, the Si / Al of a silicate mineral particle is input into this cluster analysis system to determine whether it belongs to kaolinite or montmorillonite. Then, the background noise of the electron microscope energy spectrum element distribution image is eliminated: Matlab is used to read a single electron microscope element distribution map (tiff format), and the element distribution map is binarized using the rgb2gray function to convert the element distribution map from a color scatter plot to a black and white scatter plot. Then, the medfilt2 function is used to perform median filtering and denoising on the black and white scatter plot to remove the background noise. After the denoising is completed, the threshold screening function is used to screen the remaining scatter points, and finally the remaining pixel points are obtained as the true distribution of the elements. This process is performed on all element images to obtain element images for subsequent experimental processes. Finally, the measured images are analyzed: the scanning electron microscope and energy spectrum images of the soil samples are observed and measured, and the distribution of each element presented in the electron microscope energy spectrum (after noise reduction) is focused on; first, the Si, Al, and O elements are focused on, and the overlapping parts of the three-element electron microscope images are found, and the soil particles in these parts are qualitatively identified as silicate minerals. The Si and Al element intensities of these silicate mineral particles are measured, and their Si / Al is calculated. The calculated data are input into the established cluster analysis system in turn to distinguish whether each silicate mineral belongs to kaolinite or montmorillonite; the second step is to focus on the electron microscope energy spectrum images of the Fe and O elements, find the overlapping parts of the two elements, and qualitatively identify the soil particles in these parts. The first step is to focus on the images of C and O, find the overlapping part of the elements in the two images, and there is no Si and Al distribution in this part, and characterize this part as organic matter; the fourth step is to focus on the electron microscope energy spectrum images of the three elements of Ca, C, and O, find the overlapping part of the three elements, and characterize the soil particles in this part as carbonate; so far, the properties of all particles have been identified, and the last step is to consider the composite situation between the particles. If it is found that iron oxide particles are distributed on silicate mineral particles at the same time, then the whole is characterized as an iron oxide-silicate mineral complex. If it is found that iron oxide particles are distributed on organic matter particles at the same time, then the whole is characterized as an iron oxide-organic matter complex. After completely determining the types of soil particles in the measured area, we finally focus on the electron microscope energy spectrum image of Cu. Because the Cu content in the soil is low, the signal points of Cu in the energy spectrum image are weaker and fewer. We focus on the distribution of Cu signal points on different types of soil particles. If the signal point falls on a silicate mineral, it is classified as Cu adsorbed by a silicate mineral. If the signal point falls on an organic particle, it is classified as Cu adsorbed by organic matter. If the signal point falls on an iron oxide-silicate complex, it is classified as Cu adsorbed by an iron oxide-silicate complex. And so on, we can distinguish which particles Cu is distributed on and which components it is adsorbed by.

[0016] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for identifying the binding morphology of heavy metals on soil particles based on image superposition and machine learning, characterized in that: The following steps are involved: S1. Establish a method to classify soil particles based on their constituent elements; S2, eliminating the background noise of the element distribution diagram obtained by electron microscope energy spectrum; S3. Determine the types of soil particles under a scanning electron microscope. The image of a scanning electron microscope contains several soil particles with obvious boundaries between particles. Generally, they are analyzed one by one in order. S4. Calculate the distribution of heavy metals among various soil particle components.

2. The method for identifying the binding morphology of heavy metals on soil particles based on image superposition and machine learning according to claim 1, characterized in that: Soil contains many different types of particles, mainly silicate minerals, iron oxides, organic matter, carbonates, etc. They cannot be distinguished by observing the morphology presented by an electron microscope, and need to be distinguished in combination with the element information presented by the electron microscope energy spectrum; the principle of using element information for distinction is that the elemental composition of different particles is quite different. The present invention classifies different particles according to the elements contained: silicate minerals - Si, Al, O, iron oxides - Fe, O, organic matter - C, O, carbonates - C, O, Ca; wherein silicate minerals are subdivided, and common silicate minerals in soil include kaolinite (1:1 type) and montmorillonite (2:1 type). Although the main constituent elements of both are Si, Al, O, the Si / Al ratios are different; by measuring standard samples of kaolinite and montmorillonite, a large number of Si / Al of the two types of standard samples are obtained, and machine learning cluster analysis is used to find the Si / Al boundary value to distinguish the two types of silicate minerals.

3. The method for identifying the binding morphology of heavy metals on soil particles based on image superposition and machine learning according to claim 1, characterized in that: The electron microscope energy spectrum can measure the distribution of multiple elements at the same time, and generate corresponding images according to different elements. It shows the distribution and content of elements in the form of pixels. The presence of bright pixels indicates that the element is distributed at this position, and the more pixels and the brighter the pixels are, the higher the content of the element is. However, the electron microscope energy spectrum itself has large instrument noise, which will seriously affect the presentation of the true situation of the elements. The main reason is that when measuring the electron microscope energy spectrum, the noise (background noise) generated by the substrate on which the sample is placed is too large. In order to eliminate the noise generated, the machine learning median method is used to determine the brightness threshold between the real pixel and the background pixel, and the pixels below the threshold are removed. The image obtained after removal is regarded as the true image of the element distribution and content.

4. The method for identifying the binding morphology of heavy metals on soil particles based on image superposition and machine learning according to claim 1, characterized in that: Determine the position of the analyzed particle under the scanning electron microscope image, find its corresponding position in the electron microscope energy spectrum image, and after removing the noise according to the second step, observe the element distribution at this position. If Si, Al, and O are distributed in large quantities at this position, the particle is determined to be a silicate particle. If Fe and O are distributed in large quantities at this position, the particle is determined to be an iron oxide...; among various types of soil particles, iron oxides are more active and tend to combine with other soil particle components to form a complex; therefore, when a particle contains Si, Al, O, and Fe at the same time, it is determined to be a silicate-iron oxide complex, and the same is true for organic matter-iron oxide complexes.

5. The method for identifying the binding morphology of heavy metals on soil particles based on image superposition and machine learning according to claim 1, characterized in that: After the second step of noise reduction processing on the required element image, the obtained element distribution image is superimposed with the scanning electron microscope image of each particle type that has been determined after the third step, and the proportion of pixel points of the required element falling on each component is calculated. The specific formula is (taking the organic matter-bound state as an example): Organic matter-bound heavy metals = (heavy metal signal points falling on the organic matter component / total heavy metal signal points) * 100%.

Images