Environment pollution monitoring system based on Internet of Things

By combining CNN and LSTM in the environmental monitoring system, the pollution prediction model is constructed, and the complexity of environmental monitoring data and sensor coverage problems are solved, real-time monitoring and early warning of environmental pollution is achieved, and the accuracy and response capabilities of the monitoring system are improved.

CN118981749AInactive Publication Date: 2024-11-19TIBET LANGLIHAO TECHNOLOGY CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410997264.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-11-19
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Environmental monitoring requires the installation of a large number of sensors in a wide range of geographical areas, and the monitoring data has complex spatial and temporal relationships, making it difficult to effectively monitor and analyze environmental pollution.

Method used

A pollution monitoring system based on the Internet of Things was designed, combining convolutional neural network (CNN) and long-term memory network (LSTM) to build a pollution prediction model, obtain environmental data and image data through sensors for preprocessing, and generate preprocessing. Dataset for real-time monitoring and early warning.

Benefits of technology

The accuracy and prediction capabilities of the model are improved, and changes and abnormal situations of environmental pollution can be discovered in a timely manner, and rapid response and decision-making are supported to adapt to the needs of different monitoring scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118981749A_ABST
    Figure CN118981749A_ABST
Patent Text Reader

Abstract

The invention discloses an environment pollution monitoring system based on the Internet of Things, and belongs to the technical field of environment monitoring, and the system comprises a data collection subsystem which is used for obtaining a monitoring point, deploying a sensor, and obtaining environment data and image data based on the sensor; the data processing subsystem is used for preprocessing the environment data and the image data to obtain a preprocessed data set; the model calculation subsystem is used for constructing a pollution prediction model based on a convolutional neural network and a long-short term memory network; and the monitoring subsystem is used for inputting the preprocessed data set into the pollution prediction model for calculation to obtain a prediction result, and monitoring the environmental pollution based on the prediction result. According to the invention, the model structure can be adjusted and optimized according to different environmental monitoring requirements, such as changing the size of the convolution kernel of the CNN or adjusting the time step length of the LSTM, so as to adapt to different monitoring scenes.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of environmental monitoring, and in particular relates to an environmental pollution monitoring system based on the Internet of Things. Background Art

[0002] Environmental monitoring, as the basis of environmental protection work, should be used to analyze the environmental status of the region in a targeted manner based on monitoring data, and to explore the causes of environmental problems, so as to point out the direction for environmental protection work. Applying monitoring technology to environmental protection work can grasp the changes in environmental quality in advance, and can pass on real and reliable data to environmental planning, prevention and maintenance links, laying a good foundation for the gradual recovery of the ecological environment. Introducing Internet of Things technology, intelligent control technology and big data technology into environmental monitoring work, the application of monitoring technology can monitor the actual pollution of the atmosphere, water bodies, soil, solid waste, noise, vibration, electromagnetic radiation and other aspects in the region, promote the continuous improvement of environmental protection work, and then gradually improve the environmental quality, fully understand the weaknesses of the current work, and can adjust the environmental protection strategy in a targeted manner, point out the direction for the subsequent development of environmental protection, and achieve the goal of common development with the economy.

[0003] However, current environmental monitoring requires a large number of sensors to be installed in a wide geographical area, and environmental monitoring data has complex spatial and temporal relationships. For example, the air quality at different locations in a city may be affected by multiple factors. Therefore, an IoT-based environmental pollution monitoring system is needed to solve the above problems. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes an environmental pollution monitoring system based on the Internet of Things to solve the problems existing in the above prior art.

[0005] To achieve the above object, the present invention provides an environmental pollution monitoring system based on the Internet of Things, comprising:

[0006] A data collection subsystem, used to obtain monitoring points and deploy sensors, and obtain environmental data and image data based on the sensors;

[0007] A data processing subsystem, connected to the data collection subsystem, for preprocessing the environmental data and image data to obtain a preprocessed data set;

[0008] A model calculation subsystem, connected to the data processing subsystem, for building a pollution prediction model based on a convolutional neural network and a long short-term memory network;

[0009] The monitoring subsystem is connected to the model calculation subsystem and is used to input the preprocessed data set into the pollution prediction model for calculation, obtain prediction results, and monitor environmental pollution based on the prediction results.

[0010] Preferably, the data collection subsystem includes:

[0011] A deployment module, used to select monitoring points and deploy sensors based on the geographical location of the environment to be monitored;

[0012] A data collection module is used to obtain the environmental data and the image data based on sensors.

[0013] Preferably, the data collection subsystem is also used to transmit the environmental data and the image data to a central server based on wireless communication technology, and the central server divides the data into structured data and unstructured data and then transmits the data to the data processing subsystem.

[0014] Preferably, the data processing subsystem includes:

[0015] A cleaning module, used to perform data cleaning, format conversion and data standardization processing on the environmental data in sequence to obtain processed data;

[0016] An image processing module, used for cropping the image data into small-scale image blocks, and then normalizing the pixel values ​​of the small-scale image blocks to the interval [0, 1] to generate normalized image blocks;

[0017] A fusion module is used to fuse the processed data with the normalized image block to obtain the pre-processed data set.

[0018] Preferably, the fusion module includes:

[0019] An extraction unit, used to extract features of the processed data based on an autoencoder to obtain key feature data;

[0020] An association calculation unit, used for analyzing the association between the key feature data and the normalized image block, and generating an association between the data and the image block;

[0021] A view processing unit, configured to process the key feature data based on the association between the data and the image blocks to generate a plurality of data views;

[0022] A data fusion unit is used to fuse the multiple data views and the normalized image block to obtain the preprocessed data set.

[0023] Preferably, the model calculation subsystem includes:

[0024] Model building module, used to build the convolutional neural network part, the long short-term memory network part, the feature flattening part and the time series feature construction part;

[0025] A framework construction module, used for connecting the convolutional neural network part, the long short-term memory network part, the feature flattening part and the time series feature construction part to obtain a CNN-LSTM architecture;

[0026] The training module is used to train the CNN-LSTM architecture through a sample set to obtain the pollution prediction model.

[0027] Preferably, in the model building module, the convolutional neural network part includes a convolution layer, an activation layer and a pooling layer;

[0028] The long short-term memory network part includes an LSTM layer and an LSTM output layer;

[0029] The feature flattening part is used to flatten the feature map of each time step into a one-dimensional vector;

[0030] The time series feature construction part arranges the feature vectors after flattening all time steps in chronological order to form a time series feature matrix.

[0031] Preferably, the monitoring subsystem comprises:

[0032] A calculation module, used for inputting the preprocessed data set into the pollution prediction model for calculation to obtain a prediction result;

[0033] An identification module, for identifying abnormally high or rapidly rising pollution conditions based on the prediction results and issuing an early warning;

[0034] The feedback module is used to feed back the prediction results to the model calculation subsystem to optimize the pollution prediction model.

[0035] Compared with the prior art, the present invention has the following advantages and technical effects:

[0036] The present invention combines CNN and LSTM to improve the accuracy and predictive ability of the model, because CNN is responsible for the extraction of spatial features, while LSTM processes the dynamic changes of time series. This architecture can realize the processing and analysis of real-time data, timely discover changes and abnormal conditions of environmental pollution, and help to quickly respond and make decisions. At the same time, the present invention can adjust and optimize the model structure according to different environmental monitoring needs, such as changing the convolution kernel size of CNN or adjusting the time step of LSTM to adapt to different monitoring scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0038] Figure 1 Schematic diagram of an environmental pollution monitoring system based on the Internet of Things according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0040] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0041] Embodiment 1

[0042] like Figure 1 As shown, this embodiment provides an environmental pollution monitoring system based on the Internet of Things, including:

[0043] A data collection subsystem, used to obtain monitoring points and deploy sensors, and obtain environmental data and image data based on the sensors;

[0044] A data processing subsystem, connected to the data collection subsystem, for preprocessing the environmental data and image data to obtain a preprocessed data set;

[0045] A model calculation subsystem, connected to the data processing subsystem, for building a pollution prediction model based on a convolutional neural network and a long short-term memory network;

[0046] The monitoring subsystem is connected to the model calculation subsystem and is used to input the preprocessed data set into the pollution prediction model for calculation, obtain prediction results, and monitor environmental pollution based on the prediction results.

[0047] Further optimizing the scheme, the data collection subsystem includes:

[0048] A deployment module, used to select monitoring points and deploy sensors based on the geographical location of the environment to be monitored;

[0049] A data collection module is used to obtain the environmental data and the image data based on sensors.

[0050] The data collection subsystem includes IoT sensors deployed at various monitoring points, including air quality sensors (such as PM2.5, PM10, CO2, SO2, NOx, etc.), water quality sensors (such as pH value, dissolved oxygen, ammonia nitrogen, etc.), noise sensors, etc.; other data sources can also be obtained, such as: satellite remote sensing data, meteorological data, traffic data, etc.

[0051] According to a further optimization scheme, the data collection subsystem is also used to transmit the environmental data and the image data to a central server based on wireless communication technology, and the central server divides the data into structured data and unstructured data and then transmits the data to the data processing subsystem.

[0052] The data collection subsystem is also used to transmit sensor data to the cloud or data center in real time using technologies such as NB-IoT, LoRa, Wi-Fi, 4G / 5G, etc. Cloud platforms (such as AWS, Azure, Google Cloud) are used to store large amounts of sensor data to ensure high availability and security of the data. Relational databases (such as MySQL, PostgreSQL) and non-relational databases (such as MongoDB) are used to store structured and unstructured data.

[0053] Further optimizing the scheme, the data processing subsystem includes:

[0054] A cleaning module, used to perform data cleaning, format conversion and data standardization processing on the environmental data in sequence to obtain processed data;

[0055] An image processing module, used for cropping the image data into small-scale image blocks, and then normalizing the pixel values ​​of the small-scale image blocks to the interval [0, 1] to generate normalized image blocks;

[0056] Normalization is the process of scaling image pixel values ​​from their original range (usually 0 to 255) to a new range (such as 0 to 1), which helps with numerical stability and convergence speed during model training.

[0057] Determine the cropping strategy, which can be uniform cropping (i.e. cropping patches of equal size from the image at equal intervals) or cropping based on a specific target (e.g. cropping around a target object). Cropping can be random or systematic, ensuring that the entire image area is covered.

[0058] A fusion module is used to fuse the processed data with the normalized image block to obtain the pre-processed data set.

[0059] Further optimizing the scheme, the fusion module includes:

[0060] An extraction unit, used to extract features of the processed data based on an autoencoder to obtain key feature data;

[0061] An association calculation unit, used for analyzing the association between the key feature data and the normalized image block, and generating an association between the data and the image block;

[0062] Set the key feature data to (X = {x_1, x_2, \ldots, x_n}), and the data device of the normalized block to (Y = {y_1, y_2, \ldots, y_n}). First, we need to convert the observations of each set of data into their ranks (arranged from small to large):

[0063] Calculate the rank (\text{rank}(x_i)) of (X), denoted as (R_X).

[0064] Calculate the rank (\text{rank}(y_i)) of (Y), denoted as (R_Y).

[0065] Then, the Spearman rank correlation coefficient (\rho) can be calculated by the following formula:

[0066] [\rho=1-\frac{6\sum_{i=1}^{n}(R_{X,i}-R_{Y,i})^2}{n(n^2-1)}]

[0067] in,

[0068] (R_{X,i}) is the rank of (x_i) in (X),

[0069] (R_{Y,i}) is the rank of (y_i) in (Y),

[0070] (n) is the number of sample data.

[0071] The numerator in this formula (\sum_{i=1}^{n}(R_{X,i}-R_{Y,i})^2) represents the sum of the squares of the differences in rank between the two sets of data, and the denominator (n(n^2-1)) is a factor used to normalize the coefficient so that the Spearman rank correlation coefficient ranges from -1 to 1.

[0072] A view processing unit, configured to process the key feature data based on the association between the data and the image blocks to generate a plurality of data views;

[0073] A data fusion unit is used to fuse the multiple data views and the normalized image block to obtain the preprocessed data set.

[0074] Further optimizing the scheme, the model calculation subsystem includes:

[0075] Model building module, used to build the convolutional neural network part, the long short-term memory network part, the feature flattening part and the time series feature construction part;

[0076] A framework construction module, used for connecting the convolutional neural network part, the long short-term memory network part, the feature flattening part and the time series feature construction part to obtain a CNN-LSTM architecture;

[0077] The training module is used to train the CNN-LSTM architecture through a sample set to obtain the pollution prediction model.

[0078] Specifically, the sample set contains two parts, one is the training features, and the other is the label; the training features are used as the input of the model, and the labels are used to verify the output of the pre-trained sorting model: when constructing sample pairs, a pairing method will be adopted. Assuming that there are 100 training data sets and image sets from 1 to 100, then the training features of No. 1 will form 99 groups of sample pairs with the training features of No. 2-100, and the training features of No. 2 will form 99 groups of sample pairs with the training features of No. 1 and No. 3 to 100. Therefore, in principle, for the training samples of n training blocks, n*(n-1) sample pairs can be formed, that is, a large number of samples are not required to construct the sample set.

[0079] Precision, recall, average precision (AP) and mean average precision (mAP) are used as evaluation indicators.

[0080] Precision is the proportion of true positive samples among all predicted positive samples, and is calculated as:

[0081]

[0082] Among them, TP is the number of positive samples predicted by the model as positive samples, and FP is the number of negative samples predicted by the model as positive samples.

[0083] The recall rate is the proportion of all true positive samples that are predicted to be positive samples. The formula is as follows:

[0084]

[0085] Among them, TP is the number of positive samples predicted by the model as positive samples, and FN is the number of positive samples predicted by the model as negative samples.

[0086] The horizontal axis of the PR curve is the recall rate, and the vertical axis is the precision rate, which is also called the precision-recall curve. The average precision is the surface area enclosed under the PR curve. The formula is as follows:

[0087]

[0088] In the field of target detection, data sets usually contain multiple categories. AP is often used to measure the detection effect of a model for a certain category of targets. If the detection effect of multiple categories of targets is evaluated, the AP value of each category needs to be summed and averaged to obtain the mean average precision. mAP is a more comprehensive evaluation indicator that can reflect the overall performance of the model in multi-category target detection tasks. The formula is as follows:

[0089]

[0090] The IoU threshold affects the mAP value of the model. The commonly used mAP indicators are mAP@0.5 and mAP@0.5:0.95. mAP@0.5 represents the average precision of all categories when the IoU threshold is 0.5.

[0091] Further optimizing the scheme, in the model building module, the convolutional neural network part includes a convolution layer, an activation layer and a pooling layer;

[0092] The long short-term memory network part includes an LSTM layer and an LSTM output layer;

[0093] The feature flattening part is used to flatten the feature map of each time step into a one-dimensional vector;

[0094] The time series feature construction part arranges the feature vectors after flattening all time steps in chronological order to form a time series feature matrix, which specifically includes:

[0095] 1. Input layer

[0096] The input layer receives time series data, such as a video clip or a speech signal. Assume that the input data is a time series, and each time step contains a two-dimensional image or signal segment.

[0097] 2. CNN part

[0098] The input data of each time step is processed separately by the CNN part. The CNN part usually consists of the following layers:

[0099] Convolution layer: Apply multiple convolution kernels to perform convolution operations on the input two-dimensional image to extract local features.

[0100] Activation layer: Apply an activation function (such as ReLU) to perform a nonlinear transformation on the convolution output.

[0101] Pooling layer: Perform downsampling (such as maximum pooling or average pooling) to reduce the size of the feature map and retain the main features.

[0102] Multiple convolution-activation-pooling layers: The above operations can be repeated multiple times to extract higher-level features layer by layer.

[0103] 3. Feature Flattening

[0104] After the CNN part is processed, the feature map of each time step is flattened into a one-dimensional vector. Assuming that the size of the feature map of each time step is (h\times sw\times d), the length of the vector obtained after flattening is (h\times sw\times d).

[0105] 4. Time series feature construction

[0106] Arrange the flattened feature vectors of all time steps in chronological order to form a time series feature matrix. Assuming there are (T) time steps and the length of each flattened feature vector is (F), the dimension of the time series feature matrix is ​​(T\timesF).

[0107] 5. LSTM part

[0108] The time series feature matrix is ​​input into the LSTM layer for processing. The LSTM part includes the following layers:

[0109] LSTM layer: LSTM processes time series data through memory cells and gating mechanisms, outputting a hidden state vector for each time step. Multiple layers of LSTM layers can be used to increase the expressiveness of the model.

[0110] LSTM output: The LSTM layer can output a sequence of hidden states for each time step, or only the hidden state for the last time step.

[0111] 6. Fully connected layer

[0112] The output of LSTM is connected to a fully connected layer for further feature combination and transformation. The fully connected layer can be one or more layers, depending on the complexity of the task.

[0113] 7. Output Layer

[0114] According to the requirements of the specific task, the output layer is a regression task: a fully connected layer with a linear activation function is used to output continuous values.

[0115] Further optimizing the scheme, the monitoring subsystem includes:

[0116] A calculation module, used for inputting the preprocessed data set into the pollution prediction model for calculation to obtain a prediction result;

[0117] An identification module, for identifying abnormally high or rapidly rising pollution conditions based on the prediction results and issuing an early warning;

[0118] The feedback module is used to feed back the prediction results to the model calculation subsystem to optimize the pollution prediction model.

[0119] Analysis and monitoring of model prediction results include:

[0120] Time series analysis: Analyze the temporal variation trend of pollutant concentrations.

[0121] Spatial analysis: Analyze pollution in different regions.

[0122] Anomaly detection: Identify abnormally high or rapidly rising pollution conditions and issue timely warnings.

[0123] Decision support: Based on the prediction results, formulate corresponding environmental governance measures and emergency plans.

[0124] The implementation process also includes: Assuming there is a task to predict the PM2.5 concentration of a city in the next 24 hours, the specific steps are as follows:

[0125] Data collection: Collect PM2.5 concentration and meteorological data (temperature, humidity, wind speed, etc.) over the past few months.

[0126] Data preprocessing: clean and normalize the data and generate time features (such as hours, days of the week, etc.).

[0127] Model selection and training:

[0128] The input layer receives the processed data.

[0129] Use CNN to extract spatial features of meteorological data.

[0130] Use LSTM to process the time series characteristics of PM2.5 concentration.

[0131] The extracted features are fused and combined through the fully connected layer to finally output the predicted PM2.5 concentration within the next 24 hours.

[0132] Model prediction: Input the latest meteorological data and PM2.5 historical data to predict the PM2.5 concentration in the next 24 hours.

[0133] Result analysis and monitoring: Analyze forecast results, identify potential high pollution periods, and provide decision support.

[0134] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An environmental pollution monitoring system based on the Internet of Things, characterized in that: include: A data collection subsystem, used to obtain monitoring points and deploy sensors, and obtain environmental data and image data based on the sensors; A data processing subsystem, connected to the data collection subsystem, for preprocessing the environmental data and image data to obtain a preprocessed data set; A model calculation subsystem, connected to the data processing subsystem, for building a pollution prediction model based on a convolutional neural network and a long short-term memory network; The monitoring subsystem is connected to the model calculation subsystem and is used to input the preprocessed data set into the pollution prediction model for calculation, obtain prediction results, and monitor environmental pollution based on the prediction results.

2. The environmental pollution monitoring system based on the Internet of Things according to claim 1 is characterized in that: The data collection subsystem includes: A deployment module, used to select monitoring points and deploy sensors based on the geographical location of the environment to be monitored; A data collection module is used to obtain the environmental data and the image data based on sensors.

3. The environmental pollution monitoring system based on the Internet of Things according to claim 1 is characterized in that: The data collection subsystem is also used to transmit the environmental data and the image data to a central server based on wireless communication technology, and the central server divides the data into structured data and unstructured data and transmits the data to the data processing subsystem.

4. The environmental pollution monitoring system based on the Internet of Things according to claim 1 is characterized in that: The data processing subsystem comprises: A cleaning module, used to perform data cleaning, format conversion and data standardization processing on the environmental data in sequence to obtain processed data; An image processing module, used for cropping the image data into small-scale image blocks, and then normalizing the pixel values ​​of the small-scale image blocks to the interval [0, 1] to generate normalized image blocks; A fusion module is used to fuse the processed data with the normalized image block to obtain the pre-processed data set.

5. The environmental pollution monitoring system based on the Internet of Things according to claim 4 is characterized in that: The fusion module includes: An extraction unit, used to extract features of the processed data based on an autoencoder to obtain key feature data; An association calculation unit, used for analyzing the association between the key feature data and the normalized image block, and generating an association between the data and the image block; A view processing unit, configured to process the key feature data based on the association between the data and the image blocks to generate a plurality of data views; A data fusion unit is used to fuse the multiple data views and the normalized image block to obtain the preprocessed data set.

6. The environmental pollution monitoring system based on the Internet of Things according to claim 1 is characterized in that: The model calculation subsystem includes: Model building module, used to build the convolutional neural network part, the long short-term memory network part, the feature flattening part and the time series feature construction part; A framework construction module, used for connecting the convolutional neural network part, the long short-term memory network part, the feature flattening part and the time series feature construction part to obtain a CNN-LSTM architecture; The training module is used to train the CNN-LSTM architecture through a sample set to obtain the pollution prediction model.

7. The environmental pollution monitoring system based on the Internet of Things according to claim 6 is characterized in that: In the model building module, the convolutional neural network part includes a convolution layer, an activation layer and a pooling layer; The long short-term memory network part includes an LSTM layer and an LSTM output layer; The feature flattening part is used to flatten the feature map of each time step into a one-dimensional vector; The time series feature construction part arranges the feature vectors after flattening all time steps in chronological order to form a time series feature matrix.

8. The environmental pollution monitoring system based on the Internet of Things according to claim 1 is characterized in that: The monitoring subsystem includes: A calculation module, used for inputting the preprocessed data set into the pollution prediction model for calculation to obtain a prediction result; An identification module, for identifying abnormally high or rapidly rising pollution conditions based on the prediction results and issuing an early warning; The feedback module is used to feed back the prediction results to the model calculation subsystem to optimize the pollution prediction model.

Citation Information

Cited By

  • Roadway surrounding rock anisotropic aging deformation monitoring and predicting model under mining disturbance

    CN119860991A