Automatic monitoring system and method of pollution sources based on Internet of Things
Through the Internet of Things automatic pollution source monitoring system, using multiple pollutant detection data and feature matrix, the continuity and timeliness of traditional air monitoring methods are solved, real-time monitoring and classification evaluation of air quality are realized, and monitoring efficiency is improved. and accuracy promotes environmental protection and intelligent management.
Patent Information
- Application Number
- CN202410478754.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-20
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-04-20
AI Technical Summary
Traditional air monitoring methods lack continuity and timeliness, and cannot achieve real-time monitoring of air quality, resulting in low monitoring efficiency and accuracy, making it difficult to achieve automated and intelligent management.
Using an automatic pollution source monitoring system based on the Internet of Things, the real-time continuous monitoring and classification evaluation of air quality through a number of pollutant detection data and feature matrix structures, including pollutant data acquisition, processing, fusion and analysis modules, using feature matrix and The classifier determines whether the air quality is normal.
Real-time continuous monitoring of air quality has been achieved, monitoring efficiency and accuracy have been improved, and the automation and intelligent development of environmental protection and urban air quality management have been promoted.
Smart Images

Figure CN118243874B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of Internet of Things, and more specifically, to an automatic monitoring system and method for pollution sources based on the Internet of Things. Background Art
[0002] The pollution source automatic monitoring system is based on Internet of Things technology to achieve real-time and continuous monitoring of air quality, which helps enterprises and environmental regulatory departments to understand the air quality situation in a timely manner, thereby reducing the workload of regulatory departments and improving law enforcement efficiency.
[0003] Air monitoring is the process of sampling and measuring pollutants in the air at fixed points, continuously or at regular intervals. The main monitoring items include sulfur dioxide, nitrogen monoxide, hydrocarbons and floating dust. Traditional air monitoring methods require setting up multiple monitoring points in the city, installing automatic monitoring equipment for continuous monitoring, and then manually retrieving the monitoring results for analysis at regular intervals. This method is time-consuming and labor-intensive, lacks continuity and timeliness, is not conducive to achieving automated and unmanned management, and is not conducive to control personnel understanding urban air quality parameters in real time.
[0004] Therefore, an automatic monitoring system and method for pollution sources based on the Internet of Things is desired, which can solve the limitations of traditional monitoring methods, realize real-time continuous monitoring of air quality, improve monitoring efficiency and accuracy, thereby better protecting the environment and promoting the automation and intelligent development of urban air quality management. Summary of the invention
[0005] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides an automatic monitoring system and method for pollution sources based on the Internet of Things, which realizes real-time continuous monitoring and classification evaluation of air quality through multiple pollutant detection data and feature matrix construction.
[0006] Accordingly, according to one aspect of the present application, there is provided an automatic monitoring system for pollution sources based on the Internet of Things, which includes:
[0007] A pollutant data acquisition module is used to obtain multiple pollutant detection data of the monitored area at multiple predetermined time points, wherein the pollutant detection data includes PM2.5, PM10, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone;
[0008] A pollutant data processing module, used to extract a first pollutant characteristic matrix and a second pollutant characteristic matrix from the plurality of pollutant detection data at the plurality of predetermined time points;
[0009] A pollutant data fusion module, used for constructing an air quality classification feature matrix between the first pollutant feature matrix and the second pollutant feature matrix, and performing expected activation of feature representation on the air quality classification feature matrix to obtain an optimized air quality classification feature matrix;
[0010] The pollutant data analysis module is used to pass the optimized air quality classification feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the air quality of the area to be monitored is normal.
[0011] According to another aspect of the present application, a method for automatic monitoring of pollution sources based on the Internet of Things is provided, which includes:
[0012] Acquire multiple pollutant detection data of the monitored area at multiple predetermined time points, the pollutant detection data including PM2.5, PM10, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone;
[0013] respectively extracting a first pollutant characteristic matrix and a second pollutant characteristic matrix from the plurality of pollutant detection data at the plurality of predetermined time points;
[0014] Constructing an air quality classification feature matrix between the first pollutant feature matrix and the second pollutant feature matrix, and performing expected activation of feature representation on the air quality classification feature matrix to obtain an optimized air quality classification feature matrix;
[0015] The optimized air quality classification feature matrix is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the air quality of the area to be monitored is normal.
[0016] Compared with the prior art, the present application provides an automatic monitoring system and method for pollution sources based on the Internet of Things, which obtains pollutant detection data (including PM2.5, PM10, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone) at multiple predetermined time points, extracts feature matrices and constructs air quality classification feature matrices, and finally uses a classifier to determine whether the air quality is normal. This method realizes real-time and continuous monitoring of air quality, improves monitoring efficiency and accuracy, and contributes to the automation and intelligent development of environmental protection and urban air quality management. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.
[0018] Figure 1 It is a block diagram of an automatic monitoring system for pollution sources based on the Internet of Things according to an embodiment of the present application.
[0019] Figure 2 It is a block diagram of a pollutant data processing module in an automatic monitoring system of pollution sources based on the Internet of Things according to an embodiment of the present application.
[0020] Figure 3 It is a block diagram of the pollutant extraction unit at each time point in the automatic monitoring system of pollution sources based on the Internet of Things according to an embodiment of the present application.
[0021] Figure 4 It is a block diagram of various pollutant extraction units in the automatic monitoring system of pollution sources based on the Internet of Things according to an embodiment of the present application.
[0022] Figure 5 This is a flow chart of an automatic monitoring method for pollution sources based on the Internet of Things according to an embodiment of the present application. DETAILED DESCRIPTION
[0023] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.
[0024] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.
[0025] In addition, in order to better illustrate the present application, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.
[0026] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.
[0027] Figure 1 The figure shows a block diagram of an automatic monitoring system for pollution sources based on the Internet of Things according to an embodiment of the present application. Figure 1As shown, according to the embodiment of the present application, the automatic monitoring system 100 of pollution sources based on the Internet of Things includes: a pollutant data acquisition module 110, which is used to obtain a plurality of pollutant detection data of the monitored area at a plurality of predetermined time points, and the pollutant detection data include PM2.5, PM10, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone; a pollutant data processing module 120, which is used to extract a first pollutant feature matrix and a second pollutant feature matrix from the plurality of pollutant detection data at the plurality of predetermined time points respectively; a pollutant data fusion module 130, which is used to construct an air quality classification feature matrix between the first pollutant feature matrix and the second pollutant feature matrix, and activate the expected activation of the feature representation of the air quality classification feature matrix to obtain an optimized air quality classification feature matrix; a pollutant data analysis module 140, which is used to pass the optimized air quality classification feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the air quality of the monitored area is normal.
[0028] In the embodiment of the present application, the pollutant data acquisition module 110 is used to obtain multiple pollutant detection data of the monitored area at multiple predetermined time points, and the pollutant detection data includes PM2.5, PM10, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone. It should be understood that obtaining multiple pollutant detection data of the monitored area at multiple predetermined time points, including PM2.5, PM10, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone, is to fully understand the air quality status of the area. These pollutants are common air pollutants, and their concentrations and changes can reflect the situation of air quality, and are also closely related to human health and environmental quality. By monitoring the data of these pollutants, air pollution can be discovered in time, and corresponding control and improvement measures can be taken to ensure the health of people and the sustainable development of the environment. Air pollution is not only harmful to human health, but also affects the ecological environment, affecting vegetation growth and ecological balance. By monitoring multiple pollutants, environmental pollution problems can be discovered in time to protect the health of the ecosystem. Monitoring multiple pollutants can provide timely air quality warning information, help the government and relevant departments to take effective management and control measures, and ensure the safety and health of the public. Specifically, the air quality collector includes PM2.5 sensors, PM10 sensors, sulfur dioxide sensors, nitrogen dioxide sensors, carbon monoxide sensors, and ozone sensors. Sensors are usually integrated into air quality monitoring equipment to monitor the concentration of different pollutants in real time and provide data for evaluating air quality conditions. Through the monitoring of these sensors, air quality problems can be discovered in a timely manner and corresponding measures can be taken to protect public health and the environment.
[0029] In an embodiment of the present application, the pollutant data processing module 120 is used to extract the first pollutant feature matrix and the second pollutant feature matrix from the multiple pollutant detection data at the multiple predetermined time points. It should be understood that extracting the first pollutant feature matrix and the second pollutant feature matrix may involve feature extraction and conversion of different pollutant data in order to better characterize the characteristics and attributes of the data set. By extracting the pollutant feature matrix, various analyses, such as correlation analysis, can be performed to understand the correlation and mutual influence between different pollutants. Extracting the feature matrix can help identify the patterns and trends of different pollutants, so as to better understand the changes in air quality and possible sources of pollution. Based on the extracted feature matrix, a prediction model can be established to predict the changing trend of air quality at future time points and help formulate corresponding control measures. Extracting the feature matrix can help simplify the data set and reduce the data dimension, so as to better visualize and understand the data. In general, extracting the first pollutant feature matrix and the second pollutant feature matrix can help us deeply analyze and understand the air quality monitoring data, so as to better manage and improve the air quality.
[0030] Specifically, in one embodiment of the present application, Figure 2 The figure shows a block diagram of a pollutant data processing module in an automatic monitoring system of pollution sources based on the Internet of Things according to an embodiment of the present application. Figure 2 As shown, in the above-mentioned Internet of Things-based pollution source automatic monitoring system 100, the pollutant data processing module 120 includes: a pollutant extraction unit 121 at each time point, which is used to extract the first pollutant characteristic matrix from multiple pollutant detection data at each of the predetermined time points; and a pollutant extraction unit 122, which is used to extract the second pollutant characteristic matrix from the sequence of each of the pollutant detection data at the multiple predetermined time points.
[0031] Specifically, the pollutant extraction unit 121 at each time point is used to extract the first pollutant feature matrix from the multiple pollutant detection data at each predetermined time point. It should be understood that extracting the first pollutant feature matrix allows us to analyze the data of each pollutant at different time points separately, so as to better understand the distribution, concentration changes and possible pollution sources of each pollutant. By extracting the first pollutant feature matrix, the abnormal conditions of a single pollutant at different time points, such as sudden pollution events or abnormal concentration values, can be monitored, which helps to take timely measures to deal with them. Understanding the characteristics of each pollutant at different time points can help formulate targeted pollution prevention and control measures, such as adjusting emission control strategies according to data at different time points. Extracting the first pollutant feature matrix is the basis for establishing a prediction model or a data analysis model, which can be used to predict the concentration change trend of pollutants at future time points. By analyzing the pollutant data at different time points and extracting the feature matrix, it can help optimize the monitoring strategy and determine which time points are more likely to have pollution events or need to strengthen monitoring. Therefore, extracting the first pollutant feature matrix from the multiple pollutant detection data at each predetermined time point helps to gain a deep understanding of the characteristics and changing laws of each pollutant, and provides more useful information and basis for air quality monitoring, management and improvement.
[0032] further, Figure 3 The figure shows a block diagram of the pollutant extraction unit at each time point in the pollution source automatic monitoring system based on the Internet of Things according to an embodiment of the present application. Figure 3 As shown, in the pollutant data processing module 120 of the above-mentioned Internet of Things-based pollution source automatic monitoring system 100, the pollutant extraction unit 121 for each time point includes: a vector extraction sub-unit 1211, which is used to pass the multiple pollutant detection data of each of the predetermined time points through a context encoder including an embedding layer to obtain multiple feature vectors, and cascade the multiple feature vectors to obtain a first pollutant feature vector corresponding to each of the predetermined time points; a vector two-dimensional arrangement sub-unit 1212, which is used to arrange the first pollutant feature vectors of each of the predetermined time points in two dimensions into a first pollutant feature matrix.
[0033] Specifically, the extraction vector subunit 1211 is used to obtain multiple feature vectors through the context encoder containing the embedding layer for each of the multiple pollutant detection data at the predetermined time points, and cascade the multiple feature vectors to obtain the first pollutant feature vector corresponding to each of the predetermined time points. It should be understood that by using the context encoder of the embedding layer, the original multiple pollutant detection data can be converted into a feature vector with more information. This helps to extract key features in the data and capture patterns and changes in the data. After converting the multiple pollutant detection data at each time point into feature vectors, the dimensions of these feature vectors may be consistent, which makes it easier to process and compare in subsequent processing. Cascading multiple feature vectors into a larger feature vector helps to integrate the information at each time point. In this way, data at different time points can be comprehensively considered to obtain more comprehensive pollutant feature information. After cascading these feature vectors into one feature vector, it can be more convenient to use it as the input of the model, for example, for further analysis and prediction in a machine learning model or a deep learning model. Converting data into feature vectors and cascading operations help reduce data dimensions, avoid dimensional disaster problems, and improve the efficiency of data processing and the generalization ability of the model. Therefore, processing pollutant detection data in this way can better extract features, integrate information, reduce dimensions, and provide more effective data representation for subsequent analysis and modeling.
[0034] Correspondingly, the extraction vector sub-unit includes: an embedding vectorization secondary sub-unit, which is used to use the embedding layer of the context encoder to convert the multiple pollutant detection data at each of the predetermined time points into an embedding vector to obtain a sequence of embedded vectors; a context encoding secondary sub-unit, which is used to use the converter-based Bert model of the context encoder to perform global context semantic encoding on the sequence of embedded vectors to obtain the multiple feature vectors; and a cascading secondary sub-unit, which is used to cascade the multiple feature vectors to obtain the first pollutant feature vector corresponding to each of the predetermined time points.
[0035] Further, the embedding vectorization secondary subunit is used to use the embedding layer of the context encoder to convert the multiple pollutant detection data at each predetermined time point into an embedding vector to obtain a sequence of embedded vectors. It should be understood that the embedding layer can convert the original pollutant detection data into dense embedding vectors, which can better express the semantic information of the data. Doing so helps to capture the key features and patterns in the data and improve the representation ability of the data. Converting data into embedded vectors can achieve dimensionality reduction processing of data, converting high-dimensional raw data into low-dimensional vector representations, thereby reducing the complexity of the data and improving computational efficiency. Embedding vectors are a common data representation form that can be widely used in various machine learning and deep learning models. By converting data into a sequence of embedded vectors, a more convenient input form can be provided for subsequent data processing and modeling. The embedding layer can extract important features in the data by learning the intrinsic representation of the data. This helps the system automatically learn abstract features in the data without manually designing feature extractors. The context encoder can consider the contextual information around the data points when generating embedded vectors, so as to better understand the relationship and connection between the data. This helps to improve the representation ability of the data and the performance of the model. Therefore, by converting the pollutant detection data into a sequence of embedded vectors using the embedding layer of the context encoder, we can better characterize the data, extract features, reduce dimensionality, and provide a more effective data representation for subsequent data analysis and modeling.
[0036] Furthermore, the context encoding secondary subunit is used to use the converter-based Bert model of the context encoder to perform global context semantic encoding on the sequence of the embedded vectors to obtain the multiple feature vectors. It should be understood that the Bert model is a pre-trained self-attention model that can understand the contextual relationship of the entire input sequence. By Bert encoding the sequence of embedded vectors, the complex associations and semantic information between the data can be better captured, which helps the system to understand the data more comprehensively. The Bert model can convert the input sequence into semantically rich representations that contain semantic information at the word, phrase, and even sentence levels. Through Bert encoding, higher-level feature vectors can be obtained, which contain deeper semantic information, which helps to improve the accuracy of data processing and analysis. The Bert model can effectively extract features, extract key information, and establish associations between data when processing sequence data. By Bert encoding the sequence of embedded vectors, more representative feature vectors can be obtained, which contain important features in the data and help the system better understand the meaning and structure of the data. Therefore, using the transformer-based BERT model for global context semantic encoding can improve the efficiency and accuracy of data processing, help the system better understand the relationship and meaning between data, and thus provide stronger support for subsequent analysis and decision-making.
[0037] Specifically, the cascaded secondary subunit is used to cascade the multiple feature vectors to obtain the first pollutant feature vector corresponding to each of the predetermined time points. It should be understood that by cascading multiple feature vectors, the information of each time point can be integrated together to construct a more comprehensive representation. This helps to capture the evolution and changes of data in the time dimension and provide a more comprehensive feature vector for subsequent analysis and modeling. Cascading feature vectors can expand the dimension of feature vectors so that feature vectors can more comprehensively represent the characteristics of data. Doing so helps to improve the representation ability of feature vectors so that the model can better learn the patterns and laws of data. In time series data, cascading feature vectors of multiple time points can help the model better understand the continuity and correlation of data in time. This helps to improve the modeling effect of time series data so that the model can more accurately predict future development trends. Cascading feature vectors can achieve feature fusion and combine feature information at different time points. This helps to improve the diversity and richness of features, thereby better describing the characteristics and attributes of data. Therefore, cascading multiple feature vectors to obtain the first pollutant feature vector corresponding to each predetermined time point can integrate information, expand dimensions, support time series modeling, and achieve feature fusion, thereby improving the data characterization capability and model performance.
[0038] Specifically, the vector two-dimensional arrangement subunit 1212 is used to arrange the first pollutant feature vectors of each predetermined time point in two dimensions into a first pollutant feature matrix. It should be understood that in many cases, arranging the feature vectors of each time point into a matrix can better structure the data. The matrix form helps to show the relationship between the data in time and feature dimensions, making the data easier to understand and process. Arranging the feature vectors into a matrix form can facilitate matrix operations, such as matrix multiplication, matrix transposition, etc. This helps to process data more efficiently in mathematics and algorithms, and supports various linear algebra operations and calculations. The feature matrix form can be easily input into a deep learning model or other machine learning algorithm for training and prediction. The matrix structure provides a general data representation form that is suitable for various data processing and analysis tasks. By arranging the feature vectors into a matrix form, the feature associations and patterns between different time points can be more clearly displayed. This helps the model to better learn the relationship between the data and improve the accuracy of data analysis and prediction. The feature matrix form is also convenient for visual analysis of data, and can intuitively display the feature distribution and change trend of data through heat maps and other methods, helping users to understand the data more intuitively. Therefore, arranging the first pollutant feature vectors at each predetermined time point in two dimensions into a first pollutant feature matrix can provide advantages such as structured representation, support for matrix operations, and convenience for data processing and feature association analysis, which helps to process and analyze data more efficiently.
[0039] Accordingly, in a specific example of the present application, the pollutant extraction unit 122 is used to extract the second pollutant feature matrix from the sequence of the pollutant detection data at the plurality of predetermined time points. It should be understood that by extracting the second pollutant feature matrix, characteristic information related to the second pollutant can be extracted from the pollutant detection data at the plurality of predetermined time points. This helps to capture the changes and trends of the second pollutant in the time series. The second pollutant feature matrix can help establish the correlation between different pollutants. By analyzing the second pollutant feature matrix, the correlation between different pollutants can be understood, so as to better understand the mutual influence and effect between pollutants. The second pollutant feature matrix provides a multidimensional data representation, which can simultaneously consider the relationship between different time points and different features. This helps to comprehensively consider various information and improve the comprehensiveness and accuracy of data analysis. Based on the second pollutant feature matrix, a model can be trained to predict the change trend of the second pollutant at a future time point. By establishing a model, the concentration of the second pollutant or other related indicators can be better predicted to provide support for environmental monitoring and management. The second pollutant feature matrix can also be used for data visualization, showing the relationship between different time points and different features through heat maps and the like. This helps to intuitively understand the distribution and changes of data and provide visual support for decision-making. Therefore, extracting the second pollutant feature matrix from the sequence of pollutant detection data at multiple predetermined time points can help extract feature information, establish data associations, integrate multidimensional data, support model training and data visualization, and provide more possibilities for environmental data analysis and application.
[0040] further, Figure 4 The figure shows a block diagram of various pollutant extraction units in the pollution source automatic monitoring system based on the Internet of Things according to an embodiment of the present application. Figure 4As shown, in the pollutant data processing module 120 of the above-mentioned pollution source automatic monitoring system 100 based on the Internet of Things, the pollutant extraction units 122 include: a time series encoder subunit 1221, which is used to pass the sequence of each of the pollutant detection data at the plurality of predetermined time points through a time series encoder including a one-dimensional convolutional layer and a fully connected layer to obtain a second pollutant feature vector corresponding to each of the pollutant detection data; a Gaussian density map construction subunit 1222, which is used to construct a Gaussian density map of the second pollutant feature vector of each of the pollutant detection data, wherein the mean vector of the Gaussian density map is the second pollutant feature vector. Eigenvector, the value of each position in the covariance matrix of the Gaussian density map is the variance between the eigenvalues of each position in the second pollutant feature vector; constructing a Gaussian mixture model subunit 1223, for constructing a Gaussian mixture model of the Gaussian density map of each of the pollutant detection data, the mean vector of the Gaussian mixture model is the position-weighted sum of the mean vectors of each of the Gaussian density maps, and the covariance matrix of the Gaussian mixture model is the position-weighted sum of the covariance matrices of each of the Gaussian density maps; a Gaussian discretization subunit 1224, for performing Gaussian discretization on the Gaussian mixture model to obtain a second pollutant feature matrix.
[0041] Specifically, the time series encoder subunit 1221 is used to pass the sequence of each of the pollutant detection data at the plurality of predetermined time points through a time series encoder comprising a one-dimensional convolution layer and a fully connected layer to obtain a second pollutant feature vector corresponding to each of the pollutant detection data. It should be understood that the one-dimensional convolution layer and the fully connected layer can effectively extract features from the time series data. In this way, the changes and patterns of the pollutant detection data in the time dimension can be captured, so as to better characterize the data characteristics. The time series encoder can fuse the information of the pollutant detection data at each time point and convert the time series information into a feature vector. This helps to comprehensively consider the data at different time points and improve the data characterization ability. Through the convolution layer and the fully connected layer, the model can automatically learn the important features in the data without manually designing the feature extractor. This can better adapt to different data patterns and changes. Converting time series data into feature vectors can reduce the dimension of the data, thereby simplifying the data representation and processing process. This helps to improve computational efficiency and reduce computational costs. The time series encoder using convolution layers and fully connected layers can make the model more adaptable and generalizable, and can better adapt to different pollutant data characteristics and change patterns. Therefore, converting pollutant detection data into feature vectors through a temporal encoder of one-dimensional convolutional layers and fully connected layers can effectively extract features, fuse information, automatically learn data features, reduce dimensions, and enhance model adaptability, thereby providing a better foundation for subsequent data processing and analysis.
[0042] Correspondingly, the time series encoder subunit includes: an input vector construction secondary subunit, which is used to arrange the sequence of the pollutant detection data at the multiple predetermined time points into a pollutant input vector according to the time dimension; a normalization mapping secondary subunit, which is used to perform maximum value-based normalization on the pollutant input vector to obtain a normalized pollutant input vector; a fully connected encoding secondary subunit, which is used to use the fully connected layer of the time series encoder to perform fully connected encoding on the normalized pollutant input vector to extract high-dimensional implicit features of the eigenvalues at each position in the normalized pollutant input vector; and a one-dimensional convolution encoding secondary subunit, which is used to use the one-dimensional convolution layer of the time series encoder to perform one-dimensional encoding on the normalized pollutant input vector to extract high-dimensional implicit correlation features of the correlation between the eigenvalues at each position in the normalized pollutant input vector.
[0043] Specifically, the Gaussian density map construction subunit 1222 is used to construct a Gaussian density map of the second pollutant feature vector of each of the pollutant detection data, the mean vector of the Gaussian density map is the second pollutant feature vector, and the values of each position in the covariance matrix of the Gaussian density map are the variances between the eigenvalues of each position in the second pollutant feature vector. It should be understood that by constructing a Gaussian density map, the distribution of the second pollutant feature vector can be better modeled. Gaussian distribution is a commonly used probability distribution model that can describe the distribution of data. The values in the covariance matrix reflect the correlation and variance between different features. This can help us understand the correlation between features and the uncertainty of data, and help to understand the data more comprehensively. Constructing a Gaussian density map can be used to generate new data samples that conform to a specific distribution. This is very useful in tasks such as data enhancement and data expansion. The mean vector and covariance matrix of the Gaussian density map provide an effective representation of the data distribution. This helps to extract important features and patterns in the data. Based on the Gaussian density map, we can perform probabilistic inference, such as calculating probability density, anomaly detection, etc., to better analyze the data. Therefore, constructing a Gaussian density map of the second pollutant feature vector of each pollutant detection data can help us better understand data distribution, express data uncertainty, generate data samples, extract features, perform probabilistic inference, etc., thereby providing more useful information for data analysis and application.
[0044] Specifically, the Gaussian mixture model construction subunit 1223 is used to construct a Gaussian mixture model of the Gaussian density map of each of the pollutant detection data, the mean vector of the Gaussian mixture model is the weighted sum of the mean vectors of each of the Gaussian density maps by position, and the covariance matrix of the Gaussian mixture model is the weighted sum of the covariance matrices of each of the Gaussian density maps by position. It should be understood that the pollutant detection data may contain a variety of different distribution patterns and features, which may not be completely covered by a single Gaussian distribution. By using a Gaussian mixture model, the data can be modeled as a superposition of multiple Gaussian distributions, thereby better capturing the complexity and diversity of the data. The Gaussian mixture model has greater flexibility and can adapt to different data distribution situations. Each Gaussian distribution can capture different patterns in the data, and by weighting and these distributions, the overall distribution of the data can be better fitted. By weighting the mean and covariance matrices of each Gaussian density map by position, the model can be better parameterized so that it can adapt to the correlation and difference between different features. This parameterized modeling method helps to more accurately describe the distribution characteristics of the data. Therefore, building a Gaussian mixture model can better model and describe complex data, while providing more flexible and accurate data distribution fitting. This method can better capture the diversity and complexity of data, and help improve the performance and generalization ability of the model.
[0045] Specifically, the Gaussian discretization subunit 1224 is used to perform Gaussian discretization on the Gaussian mixture model to obtain a second pollutant feature matrix. It should be understood that by discretizing the Gaussian mixture model, a continuous feature space can be converted into a discrete feature space, thereby extracting a more representative feature matrix. This helps to reduce the dimension and complexity of the data while retaining the key information of the data. Discretization can help simplify the model, making the data easier to process and analyze. In some cases, discretization can improve the computational efficiency and accuracy of the model. Through discretization, the data can be compressed into a more compact representation, thereby saving storage space and computing resources. The discretized feature matrix may be easier to interpret and understand, which helps to deeply analyze the patterns and laws behind the data. For some machine learning algorithms and models, discretization can make data easier to process and train. Some algorithms may be more suitable for processing discrete features rather than continuous features. Therefore, Gaussian discretization of the Gaussian mixture model to obtain a second pollutant feature matrix can help simplify the model, extract key features, compress data, improve model interpretability, etc., so as to better understand and process pollutant detection data.
[0046] In an embodiment of the present application, the pollutant data fusion module 130 is used to construct an air quality classification feature matrix between the first pollutant feature matrix and the second pollutant feature matrix, and to activate the expected activation of the feature representation of the air quality classification feature matrix to obtain an optimized air quality classification feature matrix. It should be understood that by combining the features of different pollutants, the air quality classification feature matrix can provide a more comprehensive air quality assessment. Comprehensive consideration of the impact of multiple pollutants can more accurately reflect the comprehensive situation of air quality. Merging the features of different pollutants into one feature matrix can reduce the dimension of the data, making the data easier to process and analyze. Some machine learning models may need to consider the relationship between multiple features, so merging the features of different pollutants into one feature matrix can better train the model. Air quality usually depends not only on a single pollutant, but is affected by multiple factors. Therefore, by integrating the features of multiple pollutants, the level and trend of air quality can be more comprehensively evaluated. Based on the comprehensive air quality classification feature matrix, the government and relevant departments can better formulate air quality management strategies and decisions to improve air quality and protect public health.
[0047] Accordingly, in one embodiment of the present application, the pollutant data fusion module 130 includes: a fusion feature matrix unit, used to fuse the first pollutant feature matrix and the second pollutant feature matrix to obtain a pollutant fusion correlation feature matrix; a mutually transposed convolution kernel unit, used to pass the pollutant fusion correlation feature matrix through an adjacent layer using a second convolutional neural network model with mutually transposed convolution kernels to obtain an air quality classification feature matrix; an optimization unit, used to activate the expected activation of the feature representation of the air quality classification feature matrix to obtain the optimized air quality classification feature matrix.
[0048] Specifically, in a specific example of the present application, the fusion feature matrix unit is used to fuse the first pollutant feature matrix and the second pollutant feature matrix to obtain a pollutant fusion correlation feature matrix. It should be understood that different pollutants may be interrelated or mutually influential. By fusing the features of different pollutants together, a more comprehensive and integrated feature representation can be obtained, which helps to more accurately describe the air quality situation. Combining the features of multiple pollutants can improve the accuracy of the air quality prediction model. Comprehensive consideration of multiple factors can better capture the complexity of air quality. Fusion of the features of multiple pollutants into a feature matrix can reduce the dimension of the data and simplify the complexity and computational cost of the model. Some machine learning models may require multiple input features for better training and prediction. By fusing the features of different pollutants, model performance and generalization ability can be improved. Comprehensive consideration of the features of multiple pollutants can help decision makers more comprehensively assess the air quality situation, thereby formulating more effective air quality management measures. Therefore, fusing the feature matrices of different pollutants to obtain a pollutant fusion correlation feature matrix can provide a more comprehensive and accurate description of air quality, which helps to improve the effect of air quality assessment and prediction.
[0049] Specifically, in a specific example of the present application, the mutually transposed convolution kernel unit is used to pass the pollutant fusion associated feature matrix through the adjacent layer using the second convolution neural network model of the mutually transposed convolution kernel to obtain the air quality classification feature matrix. It should be understood that the convolution neural network (CNN) performs well in image and sequence data processing, and the spatial features in the data can be extracted through the convolution operation. By using the convolution kernel to process the pollutant fusion associated feature matrix, the key features can be better extracted, which is helpful for the mapping and identification of air quality features. The convolution operation can effectively retain spatial information when processing spatial data such as images. By using the convolution operation, the spatial correlation in the pollutant fusion associated feature matrix can be retained, which helps to more accurately capture the patterns and features in the air quality data. Using mutually transposed convolution kernels can achieve upsampling of feature maps, thereby increasing the size and details of the feature maps. This helps to perform deeper feature representation in the neural network and improve the expressiveness and performance of the model. Convolutional neural networks perform well in classification tasks. By using the convolution kernel to process the pollutant fusion associated feature matrix, air quality classification tasks can be better performed and different air quality levels or categories can be identified. The structure and parameter sharing characteristics of convolutional neural networks make them more efficient in processing large-scale data. By using a convolutional neural network model, the accuracy of air quality classification tasks can be improved while ensuring model efficiency. Therefore, by using mutually transposed convolution kernels to process the pollutant fusion correlation feature matrix in the second convolutional neural network model, features can be better extracted, spatial information can be retained, and a more accurate feature representation can be provided for air quality classification tasks, thereby improving the accuracy and efficiency of air quality classification.
[0050] Furthermore, the mutually transposed convolution kernel unit is used for: the layers of the second convolution neural network model in the adjacent layers using mutually transposed convolution kernels respectively perform the following on the input data in the forward pass of the layer: performing convolution processing, pooling processing and nonlinear processing based on the second two-dimensional convolution kernel on the input data to obtain a first activation feature map; performing convolution processing, pooling processing and nonlinear activation processing based on the second two-dimensional convolution kernel on the first activation feature map to obtain a second activation feature map, wherein the second two-dimensional convolution kernel and the second two-dimensional convolution kernel are transposed of each other; wherein the input of the first layer of the second convolution neural network model in the adjacent layers using mutually transposed convolution kernels is the pollutant fusion correlation feature matrix, and the output of the last layer of the first convolution neural network model in the adjacent layers using mutually transposed convolution kernels is the air quality classification feature matrix.
[0051] Specifically, in a specific example of the present application, the optimization unit is used to activate the expected activation of the feature representation of the air quality classification feature matrix to obtain the optimized air quality classification feature matrix. It should be understood that, in particular, in the technical solution of the present application, it is considered that in the process of collecting pollutant detection data, abnormal fluctuations in data may be caused by factors such as equipment accuracy, environmental interference, and operational errors. When constructing models such as embedding layers, convolutional layers, and fully connected layers, there may be problems of unreasonable design or inaccurate parameter settings, resulting in the model being unable to accurately capture and reflect changes in the real environment. Emergencies such as fires and traffic accidents may occur in the area to be monitored. These events will cause a sharp change in the concentration of pollutants in a short period of time, which is reflected in the air quality classification feature matrix to form a local abnormal distribution. Local abnormal distribution may interfere with the training and prediction of the model, making it impossible for the classifier to accurately identify and judge whether the air quality is normal or not, thereby affecting the final classification result. If the local abnormal distribution is not properly handled, it may lead to wrong decisions and countermeasures, such as taking unnecessary emergency measures when the air quality is actually normal, or failing to take timely measures when the air quality is actually very poor. In order to solve the problem of local abnormal distribution in the air quality classification feature matrix, in the technical solution of the present application, the expected activation of the feature representation of the air quality classification feature matrix is activated.
[0052] Correspondingly, the optimization unit includes: a calculation adjustment eigenvalue subunit, used to calculate the exponential function value of the eigenvalue at the predetermined position of the air quality classification feature matrix raised to the power and then subtract one to obtain the exponential adjustment eigenvalue; a normalized exponential adjustment eigenvalue ratio subunit, used to calculate the exponential adjustment eigenvalue divided by the eigenvalue at the predetermined position of the air quality classification feature matrix, and calculate the absolute value of the result to obtain the normalized exponential adjustment eigenvalue ratio; a Softmax activation subunit, used to activate the normalized exponential adjustment eigenvalue ratio through a softmax function and then multiply it by the eigenvalue at the predetermined position of the air quality classification feature matrix to obtain the eigenvalue at the predetermined position of the optimized air quality classification feature matrix.
[0053] Here, in view of the local abnormal distribution in the air quality classification feature matrix, in the technical solution of the present application, the expected activation of the feature representation of the air quality classification feature matrix is activated, and each element in the air quality classification feature matrix is analyzed to determine its contribution to the final classification result. In this way, those features that contribute more to the classification result can be strengthened, while those features that contribute less or are unstable can be suppressed. In this way, not only the overall consistency of the air quality classification feature matrix can be improved, but also the sensitivity and recognition ability of the model for important features in the data can be enhanced.
[0054] In an embodiment of the present application, the pollutant data analysis module 140 is used to pass the optimized air quality classification feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the air quality of the monitored area is normal. It should be understood that the classifier maps the feature matrix to a specific category or label, so that we can classify the air quality, for example, into good, general, polluted and other levels. These categories can help decision makers and the public better understand and respond to different air quality conditions. By inputting the feature matrix into the classifier, real-time monitoring and evaluation of the air quality in the monitored area can be achieved. The classification results can be fed back to the relevant departments and the public in a timely manner to help them take necessary measures to improve the air quality. The classification results can be used in an early warning system to help predict the changing trend of air quality in the future. Once the air quality is abnormal, measures can be taken in time to reduce the impact on human health and the environment. The classification results are used to indicate the normal and abnormal state of air quality, and the air quality status can be intuitively displayed in a visual way to improve the public's awareness and attention to air quality issues. The classification results help regulatory authorities to supervise and manage air quality. According to the classification results, relevant departments can formulate corresponding policies and measures to improve and maintain air quality.
[0055] In summary, the automatic monitoring system and method of pollution sources based on the Internet of Things according to the embodiment of the present application obtains pollutant detection data (including PM2.5, PM10, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone) at multiple predetermined time points, extracts feature matrices and constructs air quality classification feature matrices, and finally uses a classifier to determine whether the air quality is normal. This method realizes real-time and continuous monitoring of air quality, improves monitoring efficiency and accuracy, and contributes to the automation and intelligent development of environmental protection and urban air quality management.
[0056] As described above, the pollution source automatic monitoring system 100 based on the Internet of Things according to the embodiment of the present application can be implemented in various terminal devices, such as a server of the pollution source automatic monitoring system based on the Internet of Things. In one example, the pollution source automatic monitoring system 100 based on the Internet of Things can be integrated into the terminal device as a software module and / or a hardware module. For example, the pollution source automatic monitoring system 100 based on the Internet of Things can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the pollution source automatic monitoring system 100 based on the Internet of Things can also be one of the many hardware modules of the terminal device.
[0057] Alternatively, in another example, the IoT-based pollution source automatic monitoring system 100 and the terminal device may also be separate devices, and the IoT-based pollution source automatic monitoring system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.
[0058] Figure 5 FIG. 1 is a flow chart of a method for automatically monitoring pollution sources based on the Internet of Things according to an embodiment of the present application. Figure 5 As shown, according to the automatic monitoring method of pollution sources based on the Internet of Things according to the embodiment of the present application, the steps include: S110, obtaining a plurality of pollutant detection data of the monitored area at a plurality of predetermined time points, the pollutant detection data including PM2.5, PM10, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone; S120, extracting a first pollutant feature matrix and a second pollutant feature matrix from the plurality of pollutant detection data at the plurality of predetermined time points respectively; S130, constructing an air quality classification feature matrix between the first pollutant feature matrix and the second pollutant feature matrix, and performing expected activation of feature representation on the air quality classification feature matrix to obtain an optimized air quality classification feature matrix; S140, passing the optimized air quality classification feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the air quality of the monitored area is normal.
[0059] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned method for automatically monitoring pollution sources based on the Internet of Things have been referred to above. Figures 1 to 4 The description of the automatic monitoring system of pollution sources based on the Internet of Things has been introduced in detail, and therefore, its repeated description will be omitted.
[0060] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0061] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0062] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0063] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0064] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
[0065] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.
[0066] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
Claims
1. An automatic monitoring system for pollution sources based on the Internet of Things, characterized in that: include: A pollutant data acquisition module is used to obtain multiple pollutant detection data of the monitored area at multiple predetermined time points, wherein the pollutant detection data includes PM2.5, PM10, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone; A pollutant data processing module, used to extract a first pollutant characteristic matrix and a second pollutant characteristic matrix from the plurality of pollutant detection data at the plurality of predetermined time points; A pollutant data fusion module, used for constructing an air quality classification feature matrix between the first pollutant feature matrix and the second pollutant feature matrix, and performing expected activation of feature representation on the air quality classification feature matrix to obtain an optimized air quality classification feature matrix; A pollutant data analysis module, used for passing the optimized air quality classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the air quality of the area to be monitored is normal; Wherein, the pollutant data fusion module includes: A fusion feature matrix unit, used for fusing the first pollutant feature matrix and the second pollutant feature matrix to obtain a pollutant fusion correlation feature matrix; A mutually transposed convolution kernel unit, used for passing the pollutant fusion correlation feature matrix through an adjacent layer using a second convolution neural network model with mutually transposed convolution kernels to obtain an air quality classification feature matrix; An optimization unit, configured to perform expected activation activation of feature representation on the air quality classification feature matrix to obtain the optimized air quality classification feature matrix; Wherein, the optimization unit comprises: A calculation adjustment eigenvalue subunit, used for calculating an exponential function value raised to a power of an eigenvalue at a predetermined position of the air quality classification feature matrix and then subtracting one to obtain an exponential adjustment eigenvalue; a normalized index-adjusted eigenvalue ratio subunit, configured to calculate the index-adjusted eigenvalue divided by the eigenvalue at a predetermined position of the air quality classification feature matrix, and calculate an absolute value of the result to obtain a normalized index-adjusted eigenvalue ratio; The softmax activation subunit is used to activate the normalized exponential adjustment eigenvalue ratio through a softmax function and then multiply it by the eigenvalue of the predetermined position of the air quality classification feature matrix to obtain the eigenvalue of the predetermined position of the optimized air quality classification feature matrix.
2. The pollution source automatic monitoring system based on the Internet of Things according to claim 1 is characterized in that: The pollutant data processing module comprises: A pollutant extraction unit at each time point, used to extract the first pollutant feature matrix from the multiple pollutant detection data at each of the predetermined time points; The pollutant extraction unit is used to extract the second pollutant feature matrix from the sequence of the pollutant detection data at the plurality of predetermined time points.
3. The pollution source automatic monitoring system based on the Internet of Things according to claim 2 is characterized in that: The pollutant extraction unit at each time point includes: A vector extraction subunit, configured to obtain a plurality of feature vectors by passing the plurality of pollutant detection data at each of the predetermined time points through a context encoder including an embedding layer, and cascade the plurality of feature vectors to obtain a first pollutant feature vector corresponding to each of the predetermined time points; The vector two-dimensional arrangement subunit is used to arrange the first pollutant characteristic vectors at each of the predetermined time points in two dimensions into a first pollutant characteristic matrix.
4. The pollution source automatic monitoring system based on the Internet of Things according to claim 3 is characterized in that: The extraction vector subunit comprises: An embedding vectorization secondary subunit, configured to use the embedding layer of the context encoder to convert the plurality of pollutant detection data at each of the predetermined time points into an embedding vector to obtain a sequence of embedding vectors; A context encoding secondary subunit, configured to perform global context semantic encoding on the sequence of the embedded vectors using a transformer-based Bert model of the context encoder to obtain the plurality of feature vectors; The cascade secondary subunit is used to cascade the multiple feature vectors to obtain the first pollutant feature vector corresponding to each of the predetermined time points.
5. The pollution source automatic monitoring system based on the Internet of Things according to claim 4 is characterized in that: The pollutant extraction units include: A time series encoder subunit, configured to pass the sequence of each of the pollutant detection data at the plurality of predetermined time points through a time series encoder comprising a one-dimensional convolutional layer and a fully connected layer to obtain a second pollutant feature vector corresponding to each of the pollutant detection data; Constructing a Gaussian density map subunit, used to construct a Gaussian density map of the second pollutant feature vector of each pollutant detection data, wherein the mean vector of the Gaussian density map is the second pollutant feature vector, and the value of each position in the covariance matrix of the Gaussian density map is the variance between the eigenvalues of each position in the second pollutant feature vector; Constructing a Gaussian mixture model subunit, which is used to construct a Gaussian mixture model of the Gaussian density map of each of the pollutant detection data, wherein the mean vector of the Gaussian mixture model is the position-weighted sum of the mean vectors of each of the Gaussian density maps, and the covariance matrix of the Gaussian mixture model is the position-weighted sum of the covariance matrices of each of the Gaussian density maps; The Gaussian discretization subunit is used to perform Gaussian discretization on the Gaussian mixture model to obtain a second pollutant characteristic matrix.
6. The pollution source automatic monitoring system based on the Internet of Things according to claim 5 is characterized in that: The timing encoder subunit comprises: An input vector construction secondary subunit is used to arrange the sequences of the various pollutant detection data at the multiple predetermined time points into a pollutant input vector according to the time dimension; A normalization mapping secondary subunit, used for performing a maximum value-based normalization process on the pollutant input vector to obtain a normalized pollutant input vector; A fully connected encoding secondary subunit, used to use the fully connected layer of the temporal encoder to perform fully connected encoding on the normalized pollutant input vector to extract high-dimensional implicit features of the feature values of each position in the normalized pollutant input vector; The one-dimensional convolutional encoding secondary subunit is used to use the one-dimensional convolutional layer of the temporal encoder to perform one-dimensional encoding on the normalized pollutant input vector to extract high-dimensional implicit correlation features of the correlation between the feature values of each position in the normalized pollutant input vector.
7. The pollution source automatic monitoring system based on the Internet of Things according to claim 6 is characterized in that: The mutually transposed convolution kernel unit is used for: each layer of the second convolution neural network model in which adjacent layers use mutually transposed convolution kernels respectively performs the following on the input data in the forward transfer of the layer: Performing convolution processing, pooling processing and nonlinear processing based on a second two-dimensional convolution kernel on the input data to obtain a first activation feature map; Performing convolution processing, pooling processing, and nonlinear activation processing based on a second two-dimensional convolution kernel on the first activation feature map to obtain a second activation feature map, wherein the second two-dimensional convolution kernel and the second two-dimensional convolution kernel are transposed of each other; Among them, the input of the first layer of the second convolutional neural network model in which the adjacent layers use mutually transposed convolution kernels is the pollutant fusion correlation feature matrix, and the output of the last layer of the first convolutional neural network model in which the adjacent layers use mutually transposed convolution kernels is the air quality classification feature matrix.
8. A pollution source automatic monitoring method based on the Internet of Things, characterized in that: include: Acquire multiple pollutant detection data of the monitored area at multiple predetermined time points, the pollutant detection data including PM2.5, PM10, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone; respectively extracting a first pollutant characteristic matrix and a second pollutant characteristic matrix from the plurality of pollutant detection data at the plurality of predetermined time points; Constructing an air quality classification feature matrix between the first pollutant feature matrix and the second pollutant feature matrix, and performing expected activation of feature representation on the air quality classification feature matrix to obtain an optimized air quality classification feature matrix; Passing the optimized air quality classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the air quality of the area to be monitored is normal; The step of constructing an air quality classification feature matrix between the first pollutant feature matrix and the second pollutant feature matrix, and performing expected activation of feature representation on the air quality classification feature matrix to obtain an optimized air quality classification feature matrix includes: Fusing the first pollutant characteristic matrix and the second pollutant characteristic matrix to obtain a pollutant fusion correlation characteristic matrix; The pollutant fusion correlation feature matrix is passed through a second convolutional neural network model using mutually transposed convolution kernels in adjacent layers to obtain an air quality classification feature matrix; Performing expected activation activation of feature representation on the air quality classification feature matrix to obtain the optimized air quality classification feature matrix; The step of performing expected activation of feature representation on the air quality classification feature matrix to obtain the optimized air quality classification feature matrix includes: Calculate an exponential function value raised to a power of the eigenvalue at a predetermined position of the air quality classification feature matrix and then subtract one to obtain an exponentially adjusted eigenvalue; Calculating the index-adjusted eigenvalue divided by the eigenvalue of a predetermined position of the air quality classification feature matrix, and calculating an absolute value of the result to obtain a normalized index-adjusted eigenvalue ratio; The normalized index adjusted eigenvalue ratio is activated by a softmax function and then multiplied by the eigenvalue of the predetermined position of the air quality classification feature matrix to obtain the eigenvalue of the predetermined position of the optimized air quality classification feature matrix.
