Single-particle source intelligent analysis method and device based on scanning electron microscope energy spectrum analysis

Abstract

The application discloses a single-particle source intelligent analysis method and device based on scanning electron microscope energy spectrum analysis, and relates to the field of environmental protection. The method is characterized in that: a computer-controlled scanning electron microscope is applied to analyze collected samples of a set pollution source, to obtain images, element data and particle size parameters of single particles in the pollution source samples, and to construct a single-particle data set of the pollution source; a residual network model is used to mine the morphological features of the single-particle images, to obtain a preliminary probability value of the single particle belonging to the set pollution source; the preliminary probability value of the single particle is combined with the particle size parameters and the element data of the single particle, and is input into an extreme gradient boosting model, to obtain a probability value of the single particle belonging to the set pollution source; the probability value of the single particle is weighted according to the mass or quantity of the single particle, and is subjected to source-class summation calculation processing, to obtain the mass or quantity contribution / occupancy ratio of atmospheric particulate matters emitted by the set pollution source. The application can fully mine the morphological, particle size and component features of the particulate matters, identify the sources of the particulate matters, and quantitatively determine the contributions of different pollution sources.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of atmospheric particulate matter source analysis, and in particular to a single particle source intelligent analysis method and device based on scanning electron microscope energy spectrum analysis. BACKGROUND

[0002] Atmospheric particulate matter pollution has a significant impact on climate, ecosystems and human health, and has been widely concerned by the community in recent years. Particulate matter source analysis is an important tool for developing particulate matter pollution reduction policies.

[0003] Although there are many source analysis methods now, most of the source analysis methods are based on total sample chemical components, and there are still problems of high uncertainty of source analysis results and low level of source class identification refinement. In order to further improve the accuracy and refinement of atmospheric particulate matter source analysis results and better serve the national atmospheric environmental management needs, it is urgent to dig more particulate matter information and develop new source analysis methods. Electron microscope analysis shows that the particles emitted by different pollution sources have differences in morphology, particle size and composition, which provides valuable information for further accurate identification of particulate matter sources. The development of computer-controlled scanning electron microscope realizes the rapid analysis of a large number of particles (thousands of particles / hour), but there is no source analysis method that can obtain quantitative source analysis results based on a large number of single particle morphology, particle size and composition information.

[0004] Therefore, how to invent a source analysis method that can obtain quantitative source analysis results based on a large number of single particle morphology, particle size and composition information has become a problem to be solved. SUMMARY

[0005] Therefore, the present application provides a single particle source intelligent analysis method and device based on scanning electron microscope energy spectrum analysis. This method can fully exploit the morphology, particle size and composition characteristics of particulate matter, identify the source of particulate matter and quantify the contribution of different pollution sources, and has more obvious advantages in identifying and quantifying collinear source classes.

[0006] In order to achieve the above purpose, the present application provides the following technical scheme: a single particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis, comprising:

[0007] By applying a computer-controlled scanning electron microscope to analyze the collected sample of a set pollution source, the image, element data and particle size parameters of a single particle in the sample of the set pollution source are obtained, and a single particle dataset of the pollution source sample is constructed;

[0008] The image features of the single particle are extracted by a residual network model, and the source of the single particle is preliminarily predicted according to the image features of the single particle to obtain a preliminary probability value of the single particle belonging to the set pollution source;

[0009] The preliminary probability value of the single particle belonging to the set pollution source is combined with the particle size parameter and element data of the single particle, and is input into an extreme gradient boosting model, and a probability value of the single particle belonging to the set pollution source is obtained by prediction of the extreme gradient boosting model.

[0010] The probability value of the single particle belonging to the set pollution source obtained is weighted by the mass or number of the single particle, and a source class summation calculation processing is performed, to obtain a mass or number contribution / occupancy ratio of atmospheric particulate matter emitted by the set pollution source.

[0011] As a preferred scheme of the single particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis, in the process of analyzing the collected sample of the set pollution source by applying a computer-controlled scanning electron microscope, the set pollution source includes soil dust, road dust, building dust, coal smoke dust, biomass combustion dust and steel smelting dust.

[0012] As a preferred scheme of the single particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis, in the process of extracting the image features of the single particle by the residual network model, the initial parameters of the residual network model are determined by training a large ImageNet data set; in the training process of the residual network model, the performance of the residual network model is optimized by an Adam optimizer and a negative log-likelihood loss function.

[0013] As a preferred scheme of the single particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis, in the process of obtaining the probability value of the single particle belonging to the set pollution source by prediction of the extreme gradient boosting model, the hyperparameters meeting the set requirements are selected by a grid search strategy, and the performance of the extreme gradient boosting model is evaluated by a five-fold cross-validation strategy.

[0014] As a preferred scheme of the single particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis, the mass occupancy ratio calculation formula of the atmospheric particulate matter emitted by the set pollution source is:

[0015]

[0016] In the formula, M i is the mass contribution occupancy ratio of the particulate matter emitted by the i-th pollution source; p ij is the preliminary probability of the particle j belonging to the pollution source i; m j is the mass of the particle j; k is the total number of the set pollution sources; and n is the total number of particles.

[0017] The number occupancy ratio calculation formula of the atmospheric particulate matter emitted by the set pollution source is:

[0018]

[0019] In the formula, N i is the proportion of the number of particles emitted by the i-th pollution source.

[0020] The application also provides a single-particle source intelligent analysis device based on scanning electron microscope energy spectrum analysis, which adopts the single-particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis described above and comprises the following modules:

[0021] A single-particle data set acquisition module is configured to analyze the collected sample of a set pollution source by using a computer-controlled scanning electron microscope, obtain the image, element data and particle size parameter of a single particle in the sample of the set pollution source, and construct a single-particle data set of the pollution source sample.

[0022] A single-particle pollution source preliminary prediction module is configured to extract the image feature of the single particle by using a residual network model, preliminarily predict the source of the single particle according to the image feature of the single particle, and obtain a preliminary probability value of the single particle belonging to the set pollution source.

[0023] A single-particle pollution source deep prediction module is configured to combine the preliminary probability value of the single particle belonging to the set pollution source with the particle size parameter and element data of the single particle, input the combination into an extreme gradient boosting model, and obtain a probability value of the single particle belonging to the set pollution source by using the extreme gradient boosting model.

[0024] An atmospheric particulate matter proportion data acquisition module is configured to perform weighted and source-class summation calculation processing on the probability value of the single particle belonging to the set pollution source by using the mass or number of the single particle, and obtain the mass or number contribution / proportion of atmospheric particulate matter emitted by the set pollution source.

[0025] As a preferred scheme of the single-particle source intelligent analysis device based on scanning electron microscope energy spectrum analysis, in the process of analyzing the collected sample of the set pollution source by using the computer-controlled scanning electron microscope, the set pollution source comprises soil dust, road dust, building dust, coal smoke dust, biomass combustion dust and steel smelting dust.

[0026] As a preferred scheme of the single-particle source intelligent analysis device based on scanning electron microscope energy spectrum analysis, in the process of extracting the image feature of the single particle by using the residual network model, the initial parameters of the residual network model are determined by training a large ImageNet data set; and in the training process of the residual network model, the performance of the residual network model is optimized by using an Adam optimizer and a negative log-likelihood loss function.

[0027] As a preferred scheme of the single-particle source intelligent analysis device based on scanning electron microscope energy spectrum analysis, in the process of obtaining the probability value of the single particle belonging to the set pollution source by the extreme gradient boosting model prediction, the super parameter meeting the set requirement is selected by the grid search strategy, and the performance of the extreme gradient boosting model is evaluated by adopting the five-fold cross-validation strategy.

[0028] As a preferred scheme of the single-particle source intelligent analysis device based on scanning electron microscope energy spectrum analysis, in the process of obtaining the probability value of the single particle belonging to the set pollution source by the extreme gradient boosting model prediction, the super parameter meeting the set requirement is selected by the grid search strategy, and the performance of the extreme gradient boosting model is evaluated by adopting the five-fold cross-validation strategy.

[0029]

[0030] In the formula, M is the mass contribution ratio of the particles emitted by the i-th pollution source; p is the preliminary probability of the particle j belonging to the pollution source i; m is the mass of the particle j; k is the total number of the set pollution sources; and n is the total number of the particles. i ij j

[0031] The number contribution ratio of the atmospheric particles emitted by the set pollution source is calculated by the following formula:

[0032]

[0033] In the formula, N is the number contribution ratio of the particles emitted by the i-th pollution source. i

[0034] ​​​​The present application has the following advantages: by applying a computer-controlled scanning electron microscope to analyze the collected sample of the set pollution source, the image, element data and particle size parameter of the single particle in the sample of the set pollution source are obtained, the single particle dataset of the pollution source sample is constructed, the image features of the single particle are extracted through a residual network model, and the source of the single particle is preliminarily predicted according to the image features of the single particle to obtain a preliminary probability value of the single particle belonging to the set pollution source; the preliminary probability value of the single particle belonging to the set pollution source is combined with the particle size parameter and element data of the single particle, and is input into an extreme gradient boosting model, and the probability value of the single particle belonging to the set pollution source is obtained through prediction of the extreme gradient boosting model; the probability value of the single particle belonging to the set pollution source obtained is weighted, and a source class summation calculation processing is performed, so that the mass or quantity contribution / percentage of the atmospheric particulate matter emitted by the set pollution source is obtained. The present application can fully exploit the morphology, particle size and component characteristics of the particulate matter, identify the source of the particulate matter and quantitatively determine the contribution of different pollution sources. At the same time, combined with the morphology and particle size characteristics, the present application has more obvious advantages in identifying and quantifying the collinear source class (the particulate matter emitted by multiple pollution sources has similar chemical components). BRIEF DESCRIPTION OF DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the following embodiment or prior art description will be briefly introduced. Obviously, the drawings in the following description are only exemplary, and those skilled in the art can also obtain other implementation drawings according to the provided drawings without any creative labor.

[0036] The structures, proportions, sizes, etc. shown in the specification are only used to cooperate with the content disclosed in the specification, to be understood and read by those skilled in the art, and do not define the limiting conditions for the implementation of the present application, so they do not have technical significance. Any modification of structure, change of proportion relationship or adjustment of size, which does not affect the effects and purposes that can be achieved by the present application, should still fall within the scope of the technical content disclosed by the present application.

[0037] Figure 1 The single particle source intelligent analysis method flowchart based on scanning electron microscope energy spectrum analysis provided for embodiment 1 of the present application is shown in the figure;

[0038] Figure 2 The single particle source intelligent analysis method flowchart based on scanning electron microscope energy spectrum analysis provided for embodiment 1 of the present application is shown in the figure;

[0039] Figure 3An error diagram of particle mass and quantity source analysis results calculated by the method and true values in the single particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis provided for embodiment 1 of the present application;

[0040] Figure 4 An error diagram of particle mass source analysis results of different particle size sections calculated by the method in the single particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis provided for embodiment 1 of the present application;

[0041] Figure 5 An error diagram of particle quantity source analysis results of different particle size sections calculated by the method in the single particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis provided for embodiment 1 of the present application;

[0042] Figure 6 An architecture diagram of the single particle source intelligent analysis device based on scanning electron microscope energy spectrum analysis provided for embodiment 2 of the present application. DETAILED DESCRIPTION

[0043] The embodiments of the present application will be described in detail by specific embodiments, and those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in the specification. Obviously, the described embodiments are part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0044] Embodiment 1

[0045] Referring to Figure 1 and Figure 2 , the single particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis provided by embodiment 1 of the present application comprises:

[0046] S1, analyzing the collected sample of the set pollution source by a scanning electron microscope controlled by a computer, obtaining the image, element data and particle size parameter of the single particle in the sample of the set pollution source, and constructing the single particle data set of the pollution source sample;

[0047] S2, extracting the image features of the single particle by a residual network model, and preliminarily predicting the source of the single particle according to the image features of the single particle, and obtaining the preliminary probability value of the single particle belonging to the set pollution source;

[0048] S3, merging the preliminary probability value of the single particle belonging to the set pollution source with the particle size parameter and element data of the single particle, and inputting into an extreme gradient boosting model, and obtaining the probability value of the single particle belonging to the set pollution source by predicting through the extreme gradient boosting model;

[0049] S4, weighting the probability value of the single particle belonging to the set pollution source by the mass or number of the single particle, and performing a source class summation calculation processing to obtain the mass or number proportion of the atmospheric particulate matter emitted by the set pollution source.

[0050] In the embodiment, in step S1, in the process of analyzing the collected sample of the set pollution source by applying a computer-controlled scanning electron microscope, the set pollution source includes soil dust, road dust, building dust, coal smoke dust, biomass burning dust and steel smelting dust.

[0051] Specifically, in the embodiment, six types of pollution source samples, including soil dust, road dust, building dust, coal smoke dust, biomass burning dust and steel smelting dust, are collected. The scanning electron microscope energy spectrum analysis (SEM-EDS) method is used to analyze the pollution source samples, and a typical pollution source single particle database is constructed. In particular, the computer-controlled scanning electron microscope energy spectrum analysis (CCSEM-EDS) can realize automatic and rapid analysis of particulate matter samples (thousands of particles / hour), and the CCSEM-EDS is recommended for sample analysis. At least 10,000 particles of each source sample are analyzed, and at least 2,000 particles of each sample are analyzed.

[0052] In the embodiment, in step S1, the single particles in the single particle data set are analyzed by the scanning electron microscope energy spectrum analysis strategy to obtain the image, element data and particle size parameters of the single particles.

[0053] Specifically, the data collected by SEM-EDS includes a microscopic picture of particulate matter, particle size parameters and element data. The volume of the particle is calculated based on the particle size parameters, the density is calculated based on the chemical composition, and the mass is the product of the volume and the density, which are all automatically calculated by the CCSEM-EDS system.

[0054] In the embodiment, in step S2, in the process of extracting the image features of the single particles by the residual network model, the initial parameters of the residual network model are determined by training a large ImageNet data set; in the training process of the residual network model, the performance of the residual network model is optimized by the Adam optimizer and the negative log-likelihood loss function.

[0055] Specifically, data augmentation techniques are applied to enhance the diversity of the dataset. Pictures are randomly rotated, flipped and scaled, and image brightness, contrast, saturation and hue are randomly adjusted to reduce the differences in different sample microscopic images caused by differences in SEM analysis conditions. ResNet model based on transfer learning is used to extract image features and predict the source of particulate matter according to the image features. The initial parameters of the model are trained based on the large ImageNet dataset. During the training process, the Adam optimizer is used, the learning rate is set to 1e-2, the step is set to 7, the decay rate is set to 0.1, and the negative log-likelihood loss function is used to optimize the performance of the model. Finally, the probability value of each particle from different sources is recorded.

[0056] In this embodiment, in step S3, in the process of obtaining the probability value of the single particle belonging to the set pollution source by the extreme gradient boosting model prediction, the hyperparameters meeting the set requirements are selected by the grid search strategy, and the performance of the extreme gradient boosting model is evaluated by using the five-fold cross-validation strategy.

[0057] Specifically, the preliminary probability value of the single particle belonging to the set pollution source predicted by the residual network model is combined with the particle size parameter and element data of the single particle as the input of the extreme gradient boosting (XGBoost) model, and the particle source label is the dependent variable. The extreme gradient boosting model is trained, and the probability of each particle belonging to different pollution sources calculated by the extreme gradient boosting model is output. In the optimization process of the extreme gradient boosting model, the most suitable hyperparameters are selected by using the grid search, and the performance of the XGBoost model is evaluated by using the five-fold cross-validation.

[0058] In this embodiment, in step S4, the probability value of the single particle belonging to the set pollution source obtained is weighted by the mass or number of the single particle, and a source class summation calculation process is performed to obtain the mass or number contribution / percentage of atmospheric particulate matter emitted by the set pollution source.

[0059] Specifically, the mass percentage calculation formula of atmospheric particulate matter emitted by the set pollution source is:

[0060]

[0061] In the formula, M i is the mass contribution percentage of particulate matter emitted by the i-th pollution source; p ij is the preliminary probability of particle j belonging to pollution source i; m j is the mass of particle j; k is the total number of set pollution sources; and n is the total number of particles.

[0062] The number percentage calculation formula of atmospheric particulate matter emitted by the set pollution source is:

[0063]

[0064] In the formula, N i is the contribution ratio of the number of particulate matter emitted by the i-th pollution source.

[0065] In this embodiment, the single-particle data set obtained in step S1 is also divided into two subsets. 80% of the single-particle data in each pollution source is randomly extracted as a first data set, which is mainly used for training the model in the present application; and the remaining 20% of the single-particle data in each pollution source is a second data set, which is used to verify the accuracy of the analysis results of the present application.

[0066] The analysis results of the present method are verified by the second data set. The real source analysis results and the results calculated by the present application are shown in Figure 3 . The calculation results of the present application are very close to the real source contribution, and the average source contribution errors of particulate matter mass and number are 0.11% and 0.15%, respectively. The error calculation formulas are as follows:

[0067]

[0068]

[0069] In the formula, M and N are the average errors of the mass and number contributions of all pollution sources; M i,true and N i,true are the real mass contribution and number contribution of the pollution source i; M i and N i are the mass contribution and number contribution of the pollution source i calculated by the present application.

[0070] The errors of the mass and number source analysis results of different particle size segments calculated by the big data artificial intelligence particulate matter source analysis method are shown in Figure 4 and Figure 5 , respectively. The accuracy of the source analysis results of the present application for particulate matter of different particle sizes has reached a very high level.

[0071] In summary, the present application analyzes the collected sample of the set pollution source by using a computer-controlled scanning electron microscope, obtains the image, element data and particle size parameter of a single particle in the sample of the set pollution source, and constructs a single particle dataset of the pollution source sample; extracts the image features of the single particle through a residual network model, and preliminarily predicts the source of the single particle according to the image features of the single particle to obtain a preliminary probability value of the single particle belonging to the set pollution source; combines the preliminary probability value of the single particle belonging to the set pollution source with the particle size parameter and element data of the single particle, and inputs the combined data into an extreme gradient boosting model to obtain a probability value of the single particle belonging to the set pollution source through the prediction of the extreme gradient boosting model; and weights the obtained probability value of the single particle belonging to the set pollution source, and performs source class summation calculation processing to obtain the mass or quantity contribution / percentage of the atmospheric particulate matter emitted by the set pollution source. The present application can fully mine the morphology, particle size and component characteristics of particulate matter, identify the source of the particulate matter and quantitatively determine the contribution of different pollution sources. At the same time, the present application has more obvious advantages in identifying and quantifying collinear source classes (the chemical components of the particulate matter emitted by multiple pollution sources are similar) by combining the morphology and particle size characteristics.

[0072] It should be noted that the method of the embodiments of the present disclosure can be performed by a single device, such as a computer or a server. The method of the embodiments can also be applied to a distributed scenario, and be completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiments of the present disclosure, and the multiple devices can interact with each other to complete the method.

[0073] It should be noted that some embodiments of the present disclosure have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in a different order than the order described above and still achieve desirable results. In addition, the processes depicted in the figures do not necessarily require the particular order shown, or sequential order, to achieve the desired results. In certain implementations, multitasking and parallel processing can be advantageous.

[0074] Embodiment 2

[0075] Referring to Figure 6 Embodiment 2 of the present application provides a single particle source intelligent analysis device based on scanning electron microscope energy spectrum analysis, comprising:

[0076] The single-particle dataset acquisition module 001 is configured to analyze the collected sample of the set pollution source by using a computer-controlled scanning electron microscope, to obtain an image, element data and particle size parameter of a single particle in the sample of the set pollution source, and to construct a single-particle dataset of the pollution source sample.

[0077] The single-particle pollution source preliminary prediction module 002 is configured to extract image features of the single particle by using a residual network model, and to preliminarily predict a source of the single particle according to the image features of the single particle, to obtain a preliminary probability value of the single particle belonging to the set pollution source.

[0078] The single-particle pollution source deep prediction module 003 is configured to combine the preliminary probability value of the single particle belonging to the set pollution source with the particle size parameter and the element data of the single particle, and to input the combined data into an extreme gradient boosting model, to obtain a probability value of the single particle belonging to the set pollution source by using the extreme gradient boosting model.

[0079] The atmospheric particulate matter proportion data acquisition module 004 is configured to perform weighted processing on the probability value of the single particle belonging to the set pollution source by using a mass or quantity of the single particle, and to perform source-class summation calculation processing, to obtain a mass or quantity contribution / proportion of atmospheric particulate matter emitted by the set pollution source.

[0080] In the single-particle dataset acquisition module 001, the set pollution source includes soil dust, road dust, building dust, coal smoke dust, biomass burning dust and steel smelting dust in the process of analyzing the collected sample of the set pollution source by using a computer-controlled scanning electron microscope.

[0081] In the single-particle pollution source preliminary prediction module 002, initial parameters of the residual network model are determined by training a large ImageNet dataset in the process of extracting image features of the single particle by using the residual network model; and the performance of the residual network model is optimized by using an Adam optimizer and a negative log-likelihood loss function in the training process of the residual network model.

[0082] In the single-particle pollution source deep prediction module 003, hyperparameters meeting set requirements are selected by using a grid search strategy, and the performance of the extreme gradient boosting model is evaluated by using a five-fold cross-validation strategy in the process of obtaining the probability value of the single particle belonging to the set pollution source by using the extreme gradient boosting model.

[0083] In the atmospheric particulate matter proportion data acquisition module 004, the mass proportion calculation formula of atmospheric particulate matter emitted by the set pollution source is:

[0084]

[0085] In the formula, M i is the mass contribution ratio of particles emitted by the i-th pollution source; p ij is the preliminary probability that particle j belongs to pollution source i; m j is the mass of particle j; k is the total number of set pollution sources; and n is the total number of particles.

[0086] The number contribution ratio of atmospheric particles emitted by the set pollution source is calculated according to the following formula:

[0087]

[0088] In the formula, N i is the number contribution ratio of particles emitted by the i-th pollution source.

[0089] It should be noted that the information interaction and execution process between the modules / units of the system described above are based on the same concept as the method embodiments in Embodiment 1 of the present application, and the technical effects brought by them are the same as those of the method embodiments of the present application. For specific content, please refer to the description of the method embodiments in the foregoing description of the method embodiments of the present application. Here, no further description is given.

[0090] Embodiment 3

[0091] Embodiment 3 of the present application provides a non-transitory computer-readable storage medium, which stores a program code of a single-particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis. The program code includes instructions for executing the single-particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis of Embodiment 1 or any possible implementation manner thereof.

[0092] The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state disk (SSD)), etc.

[0093] Embodiment 4

[0094] Embodiment 4 of the present application provides an electronic device, which includes a memory and a processor.

[0095] The processor and the memory complete mutual communication through a bus; the memory stores program instructions executable by the processor, and the processor calling the program instructions can execute the single-particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis of Embodiment 1 or any possible implementation manner thereof.

[0096] Specifically, the processor can be implemented by hardware or software. When implemented by hardware, the processor can be a logic circuit, an integrated circuit, etc. When implemented by software, the processor can be a general-purpose processor, which implements the software code stored in a memory. The memory can be integrated in the processor or located outside the processor.

[0097] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) means.

[0098] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present application can be implemented by general computing devices, which can be concentrated on a single computing device or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by program codes executable by computing devices, so that they can be stored in storage devices and executed by computing devices, and in some cases, the steps shown or described can be executed in different order, or they can be manufactured into individual integrated circuit modules, or multiple modules or steps can be manufactured into a single integrated circuit module. Thus, the present application is not limited to any specific combination of hardware and software.

[0099] Although the present application has been described in detail above with general description and specific embodiments, some modifications or improvements can be made on the basis of the present application, which is obvious to those skilled in the art. Therefore, these modifications or improvements made on the basis of not deviating from the spirit of the present application are within the scope of the present application.

Claims

1. A single particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis, characterized in that, The application comprises the following steps: By applying a computer-controlled scanning electron microscope to analyze the collected sample of a set pollution source, the image, element data and particle size parameters of single particles in the sample of the set pollution source are obtained, and a single particle data set of the pollution source sample is constructed; By a residual network model, the image features of the single particles are extracted, and the source of the single particles is preliminarily predicted according to the image features of the single particles, to obtain a preliminary probability value of the single particles belonging to the set pollution source; The preliminary probability value of the single particles belonging to the set pollution source is combined with the particle size parameters and element data of the single particles, and is input into an extreme gradient boosting model, and the probability value of the single particles belonging to the set pollution source is obtained by prediction of the extreme gradient boosting model; The probability value of the single particles belonging to the set pollution source is weighted and summed according to the mass or number of the single particles, to obtain the mass or number proportion of atmospheric particulate matter emitted by the set pollution source.

2. The single particle origin intelligent analysis method based on scanning electron microscope energy spectrum analysis according to claim 1, characterized in that, In the process of analyzing the collected sample of the set pollution source by applying a computer-controlled scanning electron microscope, the set pollution source includes soil dust, road dust, building dust, coal smoke dust, biomass burning dust and steel smelting dust. 3.The single particle origin intelligent analysis method based on SEM-EDX according to claim 2, characterized in that, In the process of extracting the image features of the single particles by the residual network model, the initial parameters of the residual network model are determined by training a large ImageNet data set; in the training process of the residual network model, the performance of the residual network model is optimized by an Adam optimizer and a negative log likelihood loss function.

4. The single particle origin intelligent analysis method based on scanning electron microscope energy spectrum analysis according to claim 3, characterized in that, In the process of obtaining the probability value of the single particles belonging to the set pollution source by prediction of the extreme gradient boosting model, the hyperparameters meeting the set requirements are selected by a grid search strategy, and the performance of the extreme gradient boosting model is evaluated by a five-fold cross-validation strategy.

5. The single particle origin intelligent analysis method based on scanning electron microscope energy spectrum analysis according to claim 4, characterized in that, The mass proportion calculation formula of atmospheric particulate matter emitted by the set pollution source is: In the formula, M i is the proportion of the mass of particulate matter emitted by the i th pollution source; p ij is the probability that the particle j belongs to the pollution source i; m j is the mass of the particle j; k is the total number of set pollution sources; and n is the total number of particles. The number proportion calculation formula of atmospheric particulate matter emitted by the set pollution source is: In the formula, N i is the proportion of the number of particulate matter discharged by the i-th pollution source.

6. The single particle source intelligent analysis device based on scanning electron microscope energy spectrum analysis, using the single particle source intelligent analysis method based on scanning electron microscope energy spectrum analysis according to any one of claims 1-5, characterized in that, The application comprises the following steps: A single particle data set acquisition module is configured to analyze a sample of a set pollution source collected by applying a computer-controlled scanning electron microscope, to obtain the image, element data and particle size parameters of single particles in the sample of the set pollution source, and to construct a single particle data set of the pollution source sample; A single particle pollution source preliminary prediction module is configured to extract the image features of the single particles by a residual network model, and to preliminarily predict the source of the single particles according to the image features of the single particles, to obtain a preliminary probability value of the single particles belonging to the set pollution source; A single particle pollution source deep prediction module is configured to combine the preliminary probability value of the single particles belonging to the set pollution source with the particle size parameters and element data of the single particles, and to input them into an extreme gradient boosting model, to obtain the probability value of the single particles belonging to the set pollution source by prediction of the extreme gradient boosting model. The atmospheric particulate matter proportion data acquisition module is configured to weight the probability value of the single particle belonging to the set pollution source by the mass or number of the single particle, and perform source-class summation calculation processing to obtain the mass proportion or number proportion of the atmospheric particulate matter emitted by the set pollution source.

7. The single particle origin intelligent analysis apparatus based on SEM-EDX according to claim 6, characterized in that, In the single particle dataset acquisition module, the set pollution source includes soil dust, road dust, building dust, coal smoke dust, biomass burning dust and steel smelting dust during analysis of the collected sample of the set pollution source by the computer-controlled scanning electron microscope.

8. The single particle origin intelligent analysis apparatus based on SEM-EDX according to claim 7, characterized in that, In the single particle pollution source preliminary prediction module, the initial parameters of the residual network model are determined by training a large ImageNet dataset during extraction of the image features of the single particle by the residual network model; and the performance of the residual network model is optimized by an Adam optimizer and a negative log-likelihood loss function during training of the residual network model.

9. The single particle origin intelligent analysis apparatus based on SEM-EDX according to claim 8, characterized in that, In the single particle pollution source deep prediction module, the hyperparameters meeting the set requirements are selected by a grid search strategy during prediction by the extreme gradient boosting model to obtain the probability value of the single particle belonging to the set pollution source, and a five-fold cross-validation strategy is used to evaluate the performance of the extreme gradient boosting model. 10.The single particle origin intelligent analysis device based on scanning electron microscope energy spectrum analysis according to claim 9, characterized in that, In the atmospheric particulate matter proportion data acquisition module, the mass proportion calculation formula of the atmospheric particulate matter emitted by the set pollution source is: In the formula, M i is the proportion of the mass of particulate matter emitted by the i-th pollution source; p ij is the probability that the particle j belongs to the pollution source i; m j is the mass of the particle j; k is the total number of set pollution sources; and n is the total number of particles. In the atmospheric particulate matter proportion data acquisition module, the mass proportion calculation formula of the atmospheric particulate matter emitted by the set pollution source is: In the formula, N i is the proportion of the number of particulate matter discharged by the i-th pollution source.

Images