Tea tree disease and pest monitoring method based on fusion of machine learning and gas sensor array

Abstract

The invention relates to the technical field of intelligent monitoring, in particular to a tea tree disease and insect pest monitoring method based on machine learning fused with a gas sensor array. Volatile substances such as leaf alcohol, phenethyl alcohol, capraldehyde, linalool, farnesene, ocimene, geraniol, n-caprylic alcohol, benzaldehyde and leaf acetate are monitored respectively. By adjusting the working temperature of the sensor, multiple groups of sensor response data are collected, and the data are preprocessed. By comparing the combination of different dimensionality reduction methods and classification models such as a support vector machine and a convolutional neural network, optimizing the model performance by using cross validation and hyper-parameter tuning, selecting an optimal classification model, and in a concentration prediction stage, comparing various regression models respectively, and optimizing the model performance by combining cross validation and hyper-parameter tuning. According to the method, the tea plant diseases and insect pests can be efficiently and accurately monitored, and the method has a wide application prospect.

Description

Technical Field

[0001] This invention relates to the fields of sensor data acquisition, gas identification and classification, and concentration prediction, specifically to a method for monitoring tea tree diseases and pests using machine learning and a gas sensor array. Background Technology

[0002] In modern agricultural production, gas detection technology has demonstrated significant application value in agricultural monitoring, environmental monitoring, food processing, and medical diagnostics. Particularly in tea cultivation, the monitoring and control of pests and diseases are crucial for ensuring tea quality and yield. However, due to the complex, variable, multi-dimensional, non-linear, and real-time dynamic characteristics of gas signals in the agricultural environment, traditional data analysis methods struggle to effectively process this data, resulting in low efficiency, significant lag, and large errors in existing pest and disease monitoring technologies. Currently, pest and disease monitoring mainly relies on manual observation and traditional detection methods, which are not only time-consuming and labor-intensive but also fail to meet the demands of modern agriculture for efficient and accurate monitoring.

[0003] With the rapid development of machine learning and artificial intelligence algorithms, related technologies have demonstrated significant advantages in the field of gas detection. For example, machine learning models such as Support Vector Machines (SVM), Random Forests, and Neural Networks can extract key information from large amounts of gas sensor data, significantly improving the accuracy of pest and disease classification and prediction. Gas monitoring systems typically consist of two parts: a gas sensor array and artificial intelligence algorithms. When gas molecules pass through the sensor array, the sensors react and generate corresponding electrical signals. After data preprocessing and feature transformation, these signals are used for feature extraction and analysis using deep learning models, thereby enabling the identification of the components of the gas mixture and accurate prediction of its concentration.

[0004] Despite this, existing technologies still have many shortcomings in the field of tea pest and disease monitoring. First, the agricultural environment contains a wide variety of volatile organic compounds (VOCs) with varying concentrations, making it difficult for a single sensor or a sensor array operating under fixed conditions to comprehensively capture these complex signals. Second, existing methods lack efficient dimensionality reduction and optimization strategies for data processing and feature extraction, resulting in low model training efficiency and limited classification and prediction accuracy. Furthermore, for predicting the concentration of different gases, existing technologies often employ a uniform model, failing to fully consider the impact of different gas characteristics on model selection, thus limiting further improvements in prediction performance.

[0005] Therefore, developing a tea plant pest and disease monitoring method based on machine learning and a fusion gas sensor array, which modulates sensor operating conditions through temperature modulation and combines multi-sensor data acquisition, data fusion, feature optimization, and classification prediction techniques, is of significant research importance and application value for achieving efficient identification and accurate concentration prediction of VOCs related to tea pests and diseases. This technology not only overcomes the limitations of existing monitoring methods but also provides scientific support for tea pest and disease control and quality improvement, while simultaneously promoting technological progress and development in related industries. Summary of the Invention

[0006] The purpose of this invention is to provide a method for monitoring tea tree diseases and pests by integrating a gas sensor array with machine learning. By changing the operating temperature of the sensors through temperature modulation, and combining sensor data acquisition with gas identification and concentration prediction technologies, multi-sensor data fusion is performed to identify the components and predict the concentration of VOCs related to tea diseases and pests.

[0007] The technical problem solved by this invention is: The objective of this invention can be achieved through the following technical solutions: A method for monitoring tea tree diseases and pests using machine learning and a gas sensor array includes the following steps: S1. Data Acquisition: The operating temperature of the sensor array is dynamically adjusted using temperature modulation technology, and sensor response data at different temperatures are acquired; multiple volatile substances are dropped into multiple conical flasks, heated and evaporated, and then injected into the test system using a syringe according to the concentration gradient, and the sensor response values ​​at different concentrations are recorded; S2. Data Preprocessing: The collected data is preprocessed, including noise reduction, normalization and data transformation, and dimensionality reduction techniques are used to optimize the data representation. S3. Classification and Prediction: The classification model is used to identify the types of gases and obtain the gas classification results. The gas concentration is then predicted through a regression model, ultimately generating high-precision monitoring results. Based on the performance evaluation results, the feature selection strategy can be further adjusted to improve the model's performance.

[0008] As a further aspect of the present invention: in S1, the various volatile substances include leaf alcohol, phenethyl alcohol, decanal, linalool, phenethyl alcohol, farnesene, ocimene, geraniol, n-octanol, benzaldehyde, and leaf alcohol acetate.

[0009] As a further aspect of the present invention: In S1, the gas sensor array is designed to include multiple planar electrode sensors, each of which has an independent sensitivity adjustment module, capable of generating specific responses for different gas molecules.

[0010] As a further aspect of the present invention: In S2, the data conversion process is as follows: The baseline voltage data is converted into response values ​​and combined for subsequent data training. The conversion formula for the baseline voltage is as follows: Where 5 represents the test system voltage, V test For testing voltage, R test R is the sensor resistance, R is the matching resistance, and S is the response value. air R is the resistance in air. gas S is the resistance in the target gas. j For the response value, S j For the response value, j represents different sensors, Ratio_S j This is the data after the response value changes.

[0011] As a further aspect of the present invention: In S3, the specific process of identifying gas types using a classification model is as follows: By calculating accuracy, R 2 The performance of the classification model is evaluated using metrics such as root mean square error (RMSE) and F1 score, and the best classification model is selected. The data to be predicted is then input into the trained classification model to obtain the classification result.

[0012] As a further aspect of the present invention: In S3, the specific process of predicting gas concentration using a regression model is as follows: By calculating accuracy, R 2 The performance of the classification model is evaluated using metrics such as root mean square error (RMSE) and F1 score, and the best regression model is selected. The data to be predicted is then input into the trained regression model to obtain the concentration prediction results.

[0013] As a further aspect of the present invention: the classification model includes support vector machine, decision tree, and neural network; the regression model includes vector machine, decision tree, and neural network.

[0014] As a further aspect of the present invention: using the coefficient of determination R 2 As a model evaluation metric, Ri 2 The expression is as follows: in, These are actual observed values. This is the average of the actual observed values. This represents the model's predicted output value, where n is the number of test points.

[0015] As a further aspect of this invention: the root mean square error (RMSE) is used as the model evaluation metric. The expression for RMSE is as follows: .

[0016] As a further aspect of the present invention, the expression for the F1 value is as follows: Precision is the accuracy rate, and Recall is the recall rate.

[0017] The beneficial effects of this invention are: This invention, under laboratory conditions, utilizes self-made sensors to monitor volatile organic compounds (VOCs) such as leaf alcohol, phenylethanol, decanal, linalool, farnesene, ocimene, geraniol, n-octanol, benzaldehyde, and leaf alcohol acetate. Multiple sets of sensor response data are collected by adjusting the sensor's operating temperature and preprocessing the data. In the gas classification stage, different combinations of dimensionality reduction methods and classification models such as Support Vector Machines (SVM) and Convolutional Neural Networks (CNN) are compared. Cross-validation and hyperparameter tuning are used to optimize model performance, selecting the optimal classification model to improve the accuracy and robustness of classification. In the concentration prediction stage, multiple regression models are compared, again using cross-validation and hyperparameter tuning to optimize model performance and achieve high-precision gas concentration prediction. The optimized gas concentration prediction model provides the best prediction results for different gases. This method can efficiently and accurately monitor tea pests and diseases and has broad application prospects. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a tea tree disease and pest monitoring method based on machine learning and gas sensor array provided in Embodiment 1 of the present invention; Figure 2 This is a flowchart of classification prediction provided in Embodiment 1 of the present invention; Figure 3 This is an effect analysis diagram of the dimensionality reduction scheme provided in Embodiment 1 of the present invention; Figure 4 This is an analysis diagram of the model classification effect provided in Embodiment 1 of the present invention; Figure 5 This is an analysis chart of the concentration prediction effect provided in Embodiment 1 of the present invention; Figure 6 This is a schematic diagram of the structure of a tea tree disease and pest monitoring system based on a machine learning fusion gas sensor array, provided in Embodiment 2 of the present invention. Figure 7 This is a graph showing the response values ​​of four sensors for leaf alcohol at four operating temperatures provided in Embodiment 1 of the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0021] Example 1 like Figure 1-2 As shown in the figure, the tea tree disease and pest monitoring method based on machine learning fusion gas sensor array provided by this invention specifically includes the following steps: S1. Data Acquisition: The operating temperature of the sensor array is dynamically adjusted using temperature modulation technology, and sensor response data at different temperatures are acquired; multiple volatile substances are dropped into multiple conical flasks, heated and evaporated, and then injected into the test system using a syringe according to the concentration gradient, and the sensor response values ​​at different concentrations are recorded; In S1, temperature modulation is achieved by using four sensors prepared in the laboratory (the four sensors include pure ZnO, zinc oxide sensors with different ruthenium loading contents, i.e., ZnO sensors, 1-Ru@ZnO sensors, 6-Ru@ZnO sensors, and 10-Ru@ZnO sensors) to set four different operating temperatures (175℃, 225℃, 275℃, and 325℃) to test various gases. The various gases include leaf alcohol, phenethyl alcohol, decanal, linalool, phenethyl alcohol, farnesene, ocimene, geraniol, n-octanol, benzaldehyde, and leaf alcohol acetate; Response data reference Figure 7 As shown, it is a graph of the response values ​​of four sensors for leaf alcohol at four operating temperatures; For example, solutions of leaf alcohol, phenethyl alcohol, decanal, linalool, phenethyl alcohol, farnesene, ocimene, geraniol, n-octanol, benzaldehyde, and leaf alcohol acetate were dropped into 10 conical flasks respectively. After the solutions were placed in an oven to evaporate into gas, they were taken out for use. Then, the gas was injected into the laboratory testing system in sequence with equal concentration gradients using a syringe. After the system stabilized for 20 minutes, a gas concentration of 2.5 ppm was added to trigger the sensor response. Then, the lid was opened to allow the testing system to recover. After stabilizing for 20 minutes, multiple sets of cycles were performed. The gas sensor array design incorporates multiple planar electrode sensors, each with an independent sensitivity adjustment module, enabling it to generate specific responses for different gas molecules. In the relevant test process of gas sampling and sensing response, a self-made planar electrode sensor was used, with a total of 9 gas sensors, which monitored geraniol, phenethyl alcohol, decanal, linalool, phenethyl alcohol, farnesene, ocimene, geraniol, n-octanol, benzaldehyde and geraniol acetate, respectively. In the test results, the corresponding gases were labeled as gas_0, gas_1, gas_2, gas_3, gas_4, gas_5, gas_6, gas_7, gas_8, gas_9 and gas_10, respectively. S2. Data Preprocessing: The collected data is preprocessed, including noise reduction, normalization and data transformation, and dimensionality reduction techniques are used to optimize the data representation. In S2, the data transformation process is as follows: The baseline voltage data is converted into response values ​​and combined for subsequent data training. The conversion formula for the baseline voltage is as follows: Where 5 represents the test system voltage, V test For testing voltage, R test R is the sensor resistance, R is the matching resistance, and S is the response value. air R is the resistance in air. gas S is the resistance in the target gas. j For the response value, S j For the response value, j represents different sensors, Ratio_S j The data reflects the change in response value; The process by which dimensionality reduction techniques optimize the representation of data is as follows: Useful features can be extracted from preprocessed data using various feature extraction techniques, such as principal component analysis (PCA) and linear discriminant analysis (LDA). The effectiveness of each method can be compared and evaluated, and the optimal dimensionality reduction scheme can be selected. During data preprocessing, all features were selected, totaling 16 features (the 16 features include the Ratio_S of 4 sensors at 4 temperatures). j To reduce redundant features and enhance classification and prediction performance, dimensionality reduction methods such as PCA (Principal Component Analysis), t-SNE (t-Distributed Random Neighborhood Embedding), UMAP (Uniform Manifold Approximation and Projection), and LDA (Linear Discriminant Analysis) are applied to process the data. This allows for the extraction and visualization of important information from high-dimensional data, facilitating subsequent model training and evaluation. By analyzing and comparing different dimensionality reduction methods and related combinations (such as PCA, t-SNE, UMAP, PCA+t-SNE, etc.), the optimal data representation method is determined, providing clear, low-dimensional feature representations for subsequent model training. like Figure 3 As shown, the dimensionality reduction scheme combining PCA and t-SNE achieved the best results.

[0022] S3. Classification and Prediction: The classification model is used to identify gas types and obtain gas classification results. Then, the gas concentration is predicted through a regression model, ultimately generating high-precision monitoring results. Based on the performance evaluation results, the feature selection strategy can be further adjusted to improve the model's performance. In S3, the specific process of identifying gas types using a classification model is as follows: By calculating accuracy, R 2 The performance of the classification model is evaluated using metrics such as root mean square error (RMSE) and F1 score, and the best classification model is selected. The data to be predicted is input into the trained classification model to obtain the classification result. The classification models include support vector machines (SVM), decision trees, neural networks, etc. The specific process of predicting gas concentration using a regression model is as follows: By calculating accuracy, R 2 The performance of the classification model is evaluated using metrics such as root mean square error (RMSE) and F1 score, and the best regression model is selected. The data to be predicted is then input into the trained regression model to obtain the concentration prediction results. Among them, regression models also include vector machines (SVM), decision trees, neural networks, etc. More specifically, the process of evaluating model performance by calculating metrics such as accuracy, R², root mean square error (RMSE), and F1 score is as follows: The performance metrics for optimizing and tuning related algorithm models mainly use the coefficient of determination R. 2 As a model evaluation metric, Ri 2 The expression is as follows: in, These are actual observed values. This is the average of the actual observed values. The model predicts the output value, where n is the number of test points; The root mean square error (RMSE) is mainly used as the performance metric for optimizing and tuning related algorithm models. The expression for RMSE is as follows: The F1 score is mainly used as the performance metric for optimizing and tuning related algorithm models. The expression for the F1 score is as follows: Precision is the accuracy rate, and Recall is the recall rate.

[0023] Based on the performance evaluation results, the process of further adjusting the feature selection strategy is as follows: Regarding classification accuracy, such as Figure 4 As shown, various models such as Support Vector Machine (SVM), Random Forest, and Multilayer Perceptron (MLP) all perform well and can effectively improve classification accuracy. In gas concentration prediction, such as Figure 5 As shown, different gases exhibit different optimal prediction models. Specifically, for each gas, multiple regression models were applied for prediction optimization, resulting in the following optimal configuration: for gas_0, the XGBoost model was used; for gas_1, the Support Vector Regression (SVR) model was used; for gas_2, a linear regression model was used; for gas_3, an SVR model was used; for gases_4, gas_5, and gas_6, MLP models were used; for gas_7, a decision tree model was used; and for gases_8 and gas_9, linear regression models were used respectively. These different gas concentration prediction models were optimized to achieve the best prediction results for each gas. Through the selection, training, and optimization of different models and algorithms, this system achieves accurate concentration prediction for different gases, demonstrating high application value and potential for widespread adoption.

[0024] Example 2 like Figure 6 As shown in the figure, the tea tree disease and pest monitoring system based on machine learning and gas sensor array provided in this embodiment of the invention specifically includes the following modules: Data acquisition module: The operating temperature of the sensor array is dynamically adjusted through temperature modulation technology, and sensor response data at different temperatures are collected; multiple volatile substances are dropped into multiple conical flasks, heated and evaporated, and then injected into the test system according to the concentration gradient using a syringe, and the sensor response values ​​at different concentrations are recorded; Data preprocessing module: preprocesses the collected data, including noise reduction, normalization and data transformation, and uses dimensionality reduction techniques to optimize the data representation; Classification and prediction module: It uses a classification model to identify gas types, obtain gas classification results, and uses a regression model to predict gas concentrations, ultimately generating high-precision monitoring results. Based on the performance evaluation results, the feature selection strategy can be further adjusted to improve the model's performance.

[0025] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for monitoring tea tree diseases and pests using machine learning fusion gas sensor arrays, characterized in that, Includes the following steps: S1. Data Acquisition: The operating temperature of the sensor array is dynamically adjusted through temperature modulation technology, and sensor response data at different temperatures are acquired. Multiple volatile substances were dropped into several conical flasks, heated and evaporated, and then injected into the test system using a syringe according to the concentration gradient. The response values ​​of the sensor at different concentrations were recorded. S2. Data Preprocessing: The collected data is preprocessed, including noise reduction, normalization and data transformation, and dimensionality reduction techniques are used to optimize the data representation. S3. Classification and Prediction: The classification model is used to identify the types of gases and obtain the gas classification results. The gas concentration is then predicted through a regression model, ultimately generating high-precision monitoring results. Based on the performance evaluation results, the feature selection strategy can be further adjusted to improve the model's performance.

2. The method for monitoring tea tree diseases and pests using a machine learning-integrated gas sensor array according to claim 1, characterized in that, In S1, various volatile substances include leaf alcohol, phenethyl alcohol, decanal, linalool, phenethyl alcohol, farnesene, ocimene, geraniol, n-octanol, benzaldehyde, and leaf alcohol acetate.

3. The method for monitoring tea tree diseases and pests using a machine learning-integrated gas sensor array according to claim 1, characterized in that, In S1, the gas sensor array is designed to include multiple planar electrode sensors, each with an independent sensitivity adjustment module, which can generate specific responses for different gas molecules.

4. The method for monitoring tea tree diseases and pests using a machine learning-integrated gas sensor array according to claim 1, characterized in that, In S2, the data transformation process is as follows: The baseline voltage data is converted into response values ​​and combined for subsequent data training. The conversion formula for the baseline voltage is as follows: Where 5 represents the test system voltage, V test For testing voltage, R test R is the sensor resistance, R is the matching resistance, and S is the response value. air R is the resistance in air. gas S is the resistance in the target gas. j For the response value, S j For the response value, j represents different sensors, Ratio_S j This is the data after the response value changes.

5. The method for monitoring tea tree diseases and pests using a machine learning-integrated gas sensor array according to claim 1, characterized in that, In S3, the specific process of identifying gas types using a classification model is as follows: By calculating accuracy, R 2 The performance of the classification model is evaluated using metrics such as root mean square error (RMSE) and F1 score, and the best classification model is selected. The data to be predicted is then input into the trained classification model to obtain the classification result.

6. The method for monitoring tea tree diseases and pests using a machine learning-integrated gas sensor array according to claim 1, characterized in that, In S3, the specific process of predicting gas concentration using a regression model is as follows: By calculating accuracy, R 2 The performance of the classification model is evaluated using metrics such as root mean square error (RMSE) and F1 score, and the best regression model is selected. The data to be predicted is then input into the trained regression model to obtain the concentration prediction results.

7. The method for monitoring tea tree diseases and pests using a machine learning fusion gas sensor array according to claim 6, characterized in that, Classification models include support vector machines, decision trees, and neural networks; regression models include vector machines, decision trees, and neural networks.

8. The method for monitoring tea tree diseases and pests using a machine learning fusion gas sensor array according to claim 6, characterized in that, Using the coefficient of determination R 2 As a model evaluation metric, Ri 2 The expression is as follows: in, These are actual observed values. This is the average of the actual observed values. This represents the model's predicted output value, where n is the number of test points.

9. A method for monitoring tea tree diseases and pests using a machine learning-integrated gas sensor array according to claim 8, characterized in that, The root mean square error (RMSE) is used as the model evaluation metric. The expression for RMSE is as follows: 。 10. A method for monitoring tea tree diseases and pests using a machine learning-integrated gas sensor array according to claim 9, characterized in that, The expression for the F1 value is as follows: Precision is the accuracy rate, and Recall is the recall rate.

Images