Method for distinguishing tea leaves treated by different ultraviolet lights by using electronic nose

Abstract

The invention discloses a method for distinguishing tea leaves under different ultraviolet light treatment by using an electronic nose, which comprises the following steps of: collecting gas volatilized by tea leaves in an ultraviolet light treatment group and an untreated group by using a sensor array in an electronic nose system, and transmitting the collected data to an upper computer in real time; the real-time response curve of each sensor is analyzed through the upper computer to make a timely judgment on the experimental effect, and all data is subjected to multi-classification division through a machine learning artificial intelligence algorithm, so that the volatile organic compounds among the tea leaves treated by different ultraviolet light can be quickly, efficiently and accurately distinguished. According to the electronic nose system, the requirement for the experiment environment is widened, the time spent on personnel training and the experiment process is shortened, the electronic nose system conducts automatic collection and automatic data uploading and analysis, the hardware cost of gas analysis is greatly reduced, and the analysis accuracy and efficiency are improved.

Description

Technical Field

[0001] The invention relates to the technical field of plant volatile detection, and in particular to a method for distinguishing tea leaves treated with different ultraviolet lights by using an electronic nose. Background Art

[0002] Insect feeding is one of the main biological factors threatening plant survival. Plants have developed complex defense systems to fight herbivorous insects during the process of co-evolution. After sensing the damage caused by insect feeding or certain elicitors in insect oral secretions, plants activate signaling networks such as Ca2+, mitogen-activated protein kinase (MAPK) and plant hormone signaling cascades to regulate the biosynthesis of defense-related metabolites. Light is a major factor controlling plant growth and development and stress response. In addition to being an energy source, it is also an important regulator of the interaction between plants and pests. Supplementing ultraviolet light as a biological stress can change the internal metabolism and hormone content of plants. In the prior art, there is extensive research on the effects of ultraviolet light of different bands on hormone expression, insect resistance and antioxidant properties in plants, but research on volatile organic compounds (VOCs) outside the body of plants is very limited.

[0003] Tea trees, as an important economic crop, not only provide humans with a popular drink, but also have important ecological and social values. In addition to causing some changes inside the plants, ultraviolet radiation to tea trees will theoretically affect the types and contents of volatile organic compounds released by the plants. If there is a method to quickly distinguish between untreated tea trees and tea trees treated with different doses of ultraviolet light, we can have a deeper understanding of the effects of ultraviolet light on plant insect resistance, antioxidants, and hormone metabolism levels from the VOC level. Therefore, it is of great significance to develop a convenient, fast, and accurate detection technology.

[0004] At present, the main method for detecting gas is gas chromatography-mass spectrometry, which combines gas chromatography and mass spectrometry to separate and identify compounds in complex mixtures. Gas chromatography is used to separate the components in the sample, while mass spectrometry is used to identify and quantify the molecular structure of each component. Although this method has the advantage of high precision, the instruments and equipment are expensive, the experimental steps are cumbersome, time-consuming, strict laboratory requirements, experimental personnel need to be strictly trained, and can only be used in the laboratory. Therefore, it is necessary to develop a simple, accurate and portable detection method. Summary of the invention

[0005] The problem to be solved by the present invention is to provide a method for distinguishing tea leaves treated with different ultraviolet light by using an electronic nose, quickly distinguish the difference in volatile organic compounds between tea leaves treated with different ultraviolet light, reduce the hardware cost of gas analysis, and improve the accuracy and efficiency of analysis.

[0006] The present invention adopts the following technical scheme: a method for distinguishing tea leaves treated with different ultraviolet light by using an electronic nose, comprising the following steps:

[0007] S1. treating tea leaves with ultraviolet light, dividing tea tree samples into groups, and continuously supplementing ultraviolet light;

[0008] S2, collecting the volatilized gas from tea leaves through the sensor array in the electronic nose system, and obtaining the original response data set X of the accumulated organic volatile gas released by the tea tree samples;

[0009] S3, preprocessing the original response data set X to obtain a sample set X1, and labeling the collected sensor response files in sequence to obtain a label set Y1;

[0010] S4, extracting features from the sample set X1 by using the window selection and sensor selection methods to obtain a sample set X2 after feature extraction;

[0011] S5, perform feature dimension reduction on the sample set X2 after feature extraction to generate a new feature sample set X 2-pca ;

[0012] S6. Create a support vector machine classifier with a linear kernel and obtain the optimal parameters by setting a grid search.

[0013] S7, feature sample set X 2-pca A ten-fold cross validation was performed, and the average accuracy was calculated and outputted, and used to evaluate the classification of volatile organic compounds in tea leaves treated with different doses of UV light.

[0014] Specifically, in step S1, the tea tree samples are divided into four groups, including a control group and three experimental groups with different ultraviolet light doses, the experimental groups are: 0.6 kJ m-2day-1 dose group, 1.2 kJ m-2day-1 dose group, 1.8 kJm-2day-1 dose group, and the tea tree samples are subjected to supplementary ultraviolet light irradiation treatment at the same time every day for seven consecutive days.

[0015] Specifically, in step S2, the electronic nose system includes: a gas collection device, a gas detection device, a circuit, and a host computer;

[0016] The gas collection device comprises: an air intake hose and a rubber plug, an air pump, and a flow meter. The air intake hose is fixed to the gas collection bottle through the rubber plug. The organic volatiles enriched in the gas collection bottle are sucked into the electronic nose through the hose by the air pump. The gas flow is detected by the flow meter and fed back to control the air pump.

[0017] The gas detection device is composed of a sensor array, wherein at least one temperature sensor and a humidity sensor and 12 gas sensors are arranged inside the sensor array, which are used to detect volatile organic compounds of tea tree samples with different treatments, and generate original response signals to obtain an original response data set X, which is transmitted to a host computer through the circuit. The host computer receives the original response signal and generates a corresponding data file, and distinguishes the differences between tea trees with different treatments through calculation.

[0018] Specifically, in step S3, the original response data set X is processed into a matrix with the same number of rows and columns, as shown below:

[0019]

[0020] Among them, X1 (n) is the nth sample data of sample set X1, x1, x2, x3, …, x 12 For the output of 12 gas sensors, m rows are intercepted as initial data according to the time series. The number of rows of the initial data is determined by the optimal range after window selection, and the number of columns is determined by the optimal sensor combination after sensor selection. Then, the corresponding samples are labeled and recorded in the label set Y1. The label set is used as the actual label to verify the prediction results.

[0021] Specifically, in step S4, the window selection and sensor selection method is used to extract features from the sample set X1, and the method is as follows:

[0022] Two parameters are set to represent the window selection range and the sensor selection number respectively. The optimal selection of the two parameters is grid searched through the control variable method. The window range is controlled to remain unchanged at the initial value. 4 to 12 sensors are randomly selected from the 12 sensors, and the response accuracy is calculated in turn to determine the best sensor selection. The selected sensors are fixed and the window range is changed in turn. The optimal window range parameters are determined according to the accuracy. Each data sample is flattened and superimposed in a matrix to obtain the sample set X2 that needs to be reduced in dimensionality, where the number of elements in each row of X2 is the number of rows and columns of the initial data multiplied by the number of columns.

[0023] Specifically, in step S5, feature dimension reduction is performed on the sample set X2 after feature extraction, including the following steps:

[0024] S5.1, standardize the sample set X2 in step S4, and obtain

[0025] S5.2, use superscript T to indicate transpose, calculate The covariance matrix C of

[0026] S5.3. Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues ​​and eigenvectors, and arrange them in order of eigenvalue size to obtain λ1≥λ2≥…≥λ m , the corresponding eigenvectors are γ1, γ2, …, γ3, which are the 1st, 2nd, …, mth principal components respectively;

[0027] S5.4. Calculate the contribution rate E of each principal component j , determine the number of principal components according to the variance contribution rate, select the eigenvectors corresponding to the first k largest eigenvalues ​​so that their cumulative contribution rate is greater than 95%, calculate the score Z of the standardized measurement value in the first d principal component directions, and take the first two principal components with cumulative variance contribution rates greater than 95% as the new feature sample set to form the matrix V k ;

[0028] S5.5. Project the data matrix to the selected principal component space to obtain a new feature sample set X 2-pca .

[0029] Specifically, in step S6, the support vector machine, the decision function is expressed as follows:

[0030] f(x)=sign(ω T x+b)

[0031] Among them, x is the feature vector of the input sample, ω is the normal vector, which determines the direction of the hyperplane; b is the bias term, which determines the distance between the hyperplane and the origin; sign represents the sign of the function, which is 1 when it is greater than or equal to 0 and -1 when it is less than 0;

[0032] If f(x)>0, it is classified as positive, if f(x)<0, it is classified as negative;

[0033] Specifically, in step S7, the new feature sample set X 2-pca Perform 10-fold cross validation, divide the data into 10 folds, use each fold as a validation set, and the rest as a training set, and perform 10 training and validation cycles, including the following sub-steps:

[0034] S7.1. New feature sample set X 2-pca Randomly divide into 10 equal subsets as folds, represented as: X 2-pca (1) ,X 2-pca (2) ,X 2-pca (3) ,…,X 2-pca (10) ;

[0035] S7.2, using the first fold as a validation set and the remaining 9 folds as training sets, using the support vector machine classifier in step S6 to train the sample set model, and evaluating the performance of the sample set model on the first fold;

[0036] S7.3, repeat the steps in S7.2 10 times, and select the first to tenth folds as the validation set;

[0037] S7.4. Summarize the results of 10 verifications, calculate the average and standard deviation, evaluate the overall performance of the sample set model, and finally output the average accuracy. The average accuracy is used to evaluate the classification of volatile organic compounds in tea leaves treated with different doses of ultraviolet light. The higher the average accuracy, the higher the prediction accuracy and the more obvious the classification effect.

[0038] The technical solution of the present invention also provides: an electronic device, comprising:

[0039] one or more processors;

[0040] a storage device having one or more programs stored thereon;

[0041] When the one or more programs are executed by the one or more processors, the one or more processors implement any of the above-mentioned methods of using an electronic nose to distinguish tea leaves treated with different ultraviolet lights.

[0042] The technical solution of the present invention also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in any of the above methods of using an electronic nose to distinguish tea leaves treated with different ultraviolet lights are implemented.

[0043] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0044] 1. The utilization method of the present invention utilizes the sensor array in the electronic nose system to collect the volatile gases of the tea leaves in the ultraviolet light treatment group and the untreated group, transmits the collected data to the host computer in real time, makes timely judgments on the experimental effects through the real-time response curves of each sensor, and divides all the data into multiple categories through machine learning artificial intelligence algorithms, so as to quickly, efficiently and accurately distinguish the differences in volatile organic compounds between tea leaves treated with different ultraviolet lights.

[0045] 2. The present invention broadens the requirements for the experimental environment through the electronic nose system, reduces the time spent on personnel training and the experimental process, and the electronic nose system has the functions of automatic collection, automatic data uploading and analysis, which greatly reduces the hardware cost of gas analysis and improves the accuracy and efficiency of the analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a schematic flow chart of a method for distinguishing tea leaves treated with different ultraviolet light using an electronic nose according to the present invention;

[0047] Figure 2 This is a structural block diagram of the electronic nose system of the present invention;

[0048] Figure 3 It is a PCA dimension reduction visualization diagram of tea samples in the UV treatment group and the control group based on the detection method of the present invention in the embodiment;

[0049] Figure 4 It is a classification confusion matrix diagram of tea samples in the ultraviolet light treatment group and the control group based on the detection method of the present invention in the embodiment. DETAILED DESCRIPTION

[0050] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0051] Meanwhile, the step numbers in the embodiments of the present invention are only provided for the convenience of explanation and description, and no limitation is imposed on the order of the steps. The execution order of the steps in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0052] In one embodiment of the present invention, the electronic nose system was used to collect samples of volatile gases. A total of 415 tea tree samples were collected, with a concentration of 0.6 kJ m -2 day -1 103 tea tree samples in the dose group, 1.2 kJ m -2 day -1 104 tea tree samples in the dose group, 1.8 kJ m -2 day -1 There were 104 tea tree samples in the dosage group and 104 tea tree samples without UV treatment. All tea tree samples came from the School of Instrument Science and Optoelectronic Engineering of Hefei University of Technology.

[0053] Based on the collected tea tree samples, the method of the present invention is used to distinguish tea leaves under different UV treatments through an electronic nose. It is implemented based on the DESKTOP-DTRJMNR computer and the running environment is: Windows 10, Intel i5 processor, 8G running memory, Pycharm 2021, Python 3.9, scikit-learn 1.2.1.

[0054] like Figure 1 As shown, the method of using an electronic nose to distinguish tea leaves treated with different ultraviolet light specifically comprises the following steps:

[0055] S1. Tea leaves were treated with ultraviolet light and divided into a control group and three experimental groups with different ultraviolet light doses. The supplementary ultraviolet light irradiation lasted for 7 days.

[0056] S2. The gas volatilized by the tea leaves is collected through the sensor array in the electronic nose system to obtain an original response data set X of the organic volatile gas released and accumulated by the tea tree sample.

[0057] In this embodiment, the electronic nose system is as follows Figure 2 As shown, it includes: a gas collection device, a gas detection device, a circuit, and a host computer;

[0058] The gas collection device includes: an air intake hose and a rubber plug, an air pump, and a flow meter; the hose is fixed to the gas collection bottle through the rubber plug, and the enriched organic volatiles in the gas bottle are sucked into the electronic nose by the air pump through the hose. At the same time, the flow meter detects the gas flow and feedback controls the air pump. The sensor is first preheated for 40 minutes before starting work. The operating voltage of the sensor is set to 5V and the sampling frequency is set to 1Hz.

[0059] The gas detection device is mainly composed of a sensor array, which contains one temperature sensor, one humidity sensor and 12 gas sensors. It is used to detect volatile organic compounds in tea tree samples with different treatments and generate original response signals.

[0060] The circuit plays the role of signal transmission, transmitting the original response signal of the sensor array to the host computer. The host computer receives the original response signal and generates a corresponding data file, and distinguishes the differences between tea trees with different treatments through algorithms.

[0061] S3. Preprocess the original response data set X to obtain a sample set X1, and label the collected sensor response files in sequence to obtain a label set Y1.

[0062] First, process all the original data into matrices with the same number of rows and columns, as shown below:

[0063]

[0064] Among them, X1 (n) is the nth sample data of sample set X1 (n=1,2,3,…,n), where x1,x2,x3,…,x 12Represents the output of 12 gas sensors. According to the time series, m rows of data are intercepted as the initial data. The number of rows in the initial data is determined by the optimal range after window selection, and the number of columns is determined by the optimal sensor combination after sensor selection. This matrix indicates that rows 1 to m are selected as the window range, all sensors are selected as the best combination, and then the corresponding samples are labeled and recorded in the label set Y1. The label set is used as the actual label to verify the prediction results.

[0065] S4, using the window selection and sensor selection methods to perform feature extraction on the sample set X1, to obtain a sample set X2 after feature extraction;

[0066] Two parameters are set to represent the window selection range and the sensor selection number, respectively, to facilitate the subsequent data to select the window range and sensor to achieve the best accuracy; the best choice of the two parameters is grid searched by controlling the variables. First, the window range is controlled to remain unchanged at the initial test value, and the accuracy of the response of 4 to 12 sensors randomly selected from 12 sensors is calculated in turn to determine which sensors are the best; then the selected sensors are fixed and the window range is changed in turn, and the parameters of the best window range are determined according to the accuracy; each sample processed in this step is flattened and then superimposed in a matrix, and finally the matrix X2 that needs to be reduced in dimension is obtained, as shown below:

[0067]

[0068] The number of elements in each row is equal to the number of rows and columns of the initial data, and n is the number of total samples;

[0069] S5, perform feature dimension reduction on the sample set X2 after feature extraction to generate a new feature sample set X 2-pca , specifically including the following steps:

[0070] S5.1. Standardize the data X2 in step S4:

[0071]

[0072] Where μ is the mean vector of each column in X, and σ is the standard deviation vector of each column in X;

[0073] S5.2. Calculate the covariance matrix, where n is the number of samples:

[0074]

[0075] S5.3. Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues ​​and eigenvectors, where Λ is the eigenvalue diagonal matrix and V is the eigenvector matrix:

[0076] C=VΛVT

[0077] Calculate the eigenvalues ​​of the covariance matrix C and arrange them in order of size to obtain λ1≥λ2≥…≥λ m , the corresponding eigenvectors are γ1, γ2, …γ3, and the required eigenvectors are called the 1st, 2nd, …mth principal components in order;

[0078] S5.4. Calculate the contribution rate E of each principal component j , as shown below:

[0079]

[0080] Among them, λ j is the jth eigenvalue, is the cumulative sum of all eigenvalues;

[0081] The number of principal components is determined according to the variance contribution rate, and the eigenvectors corresponding to the first k largest eigenvalues ​​are selected so that their cumulative contribution rate is greater than 95%. The score Z of the standardized measurement value in the direction of the first d principal components is calculated as shown in the following formula:

[0082]

[0083] Take the first two principal components whose cumulative variance contribution rate is greater than 95% as the new feature sample set to form the matrix V k ;

[0084] S5.5. Project the data matrix into the selected principal component space:

[0085]

[0086] S6. Create a support vector machine (SVM) classifier with a linear kernel and set up a grid search to obtain the optimal parameters;

[0087] The decision function of the support vector machine is shown as follows:

[0088] f(x)=sign(ω T x+b)

[0089] Among them, x is the feature vector of the input sample; ω is the normal vector, which determines the direction of the hyperplane; b is the bias term, which determines the distance between the hyperplane and the origin; sign represents the sign of the function, which is 1 when it is greater than or equal to 0 and -1 when it is less than 0. If f(x)>0, it is classified as a positive class, and if f(x)<0, it is classified as a negative class.

[0090] S7, the new feature sample set X 2-pcaPerform ten-fold cross validation, divide the data into 10 folds, use each fold as a validation set, and the rest as a training set. Perform 10 training and validation cycles, calculate and output the average accuracy, and include the following steps:

[0091] S7.1. Randomly divide the data set into 10 roughly equal subsets, namely: X 2-pca (1) ,X 2-pca (2) ,X 2-pca (3) ,…,X 2-pca (10) The specific division method is shown in Table 1 below:

[0092] Table 1 Ten-fold cross validation division method

[0093]

[0094]

[0095] S7.2, using the first fold as a validation set and the remaining 9 folds as training sets, train the model using the SVM classifier in step S6 and evaluate the performance of the model on the first fold;

[0096] S7.3, repeat the steps in S7.2 10 times, and select the first to tenth folds as the validation set;

[0097] S7.4. Summarize the results of the 10 verifications, calculate the mean and standard deviation of these verification results to evaluate the overall performance of the model, and finally output the average accuracy rate. This is used to evaluate the classification of volatile organic compounds in tea leaves treated with different doses of UV light. A high accuracy rate indicates a high prediction accuracy and a significant classification effect.

[0098] In order to evaluate the measurement data of the electronic nose, PCA dimensionality reduction was performed on the eigenvalues ​​for visualization, such as Figure 3 As shown, tea plants treated with different UV light were clearly separated from the control sample.

[0099] The SVM classification results are shown in the confusion matrix, as shown in Figure 4 As shown, tea trees treated with different UV light and the control samples all had extremely high prediction accuracy.

[0100] The results show that the method of this embodiment can quickly, efficiently and accurately distinguish the volatile organic compounds between teas treated with different ultraviolet lights. At the same time, the electronic nose system broadens the requirements for the experimental environment, reduces personnel training and the time spent on the experimental process, and the electronic nose system has the functions of automatic collection, automatic data uploading and analysis, which greatly reduces the hardware cost of gas analysis and improves the accuracy and efficiency of the analysis.

[0101] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for distinguishing tea leaves treated with different ultraviolet light using an electronic nose, characterized in that: The following steps are involved: S1. treating tea leaves with ultraviolet light, dividing tea tree samples into groups, and continuously supplementing ultraviolet light; S2, collecting the volatilized gas from tea leaves through the sensor array in the electronic nose system, and obtaining the original response data set X of the accumulated organic volatile gas released by the tea tree samples; S3, preprocessing the original response data set X to obtain a sample set X1, and labeling the collected sensor response files in sequence to obtain a label set Y1; S4, extracting features from the sample set X1 by using the window selection and sensor selection methods to obtain a sample set X2 after feature extraction; S5, perform feature dimension reduction on the sample set X2 after feature extraction to generate a new feature sample set X 2-pca ; S6. Create a support vector machine classifier with a linear kernel and obtain the optimal parameters by setting a grid search. S7, feature sample set X 2-pca A ten-fold cross validation was performed, and the average accuracy was calculated and outputted, and used to evaluate the classification of volatile organic compounds in tea leaves treated with different doses of UV light.

2. The method for distinguishing tea leaves treated with different ultraviolet light using an electronic nose according to claim 1, characterized in that: In step S1, the tea tree samples are divided into four groups, including a control group and three experimental groups with different ultraviolet light doses, the experimental groups are: 0.6 kJ m-2day-1 dose group, 1.2 kJ m-2day-1 dose group, 1.8 kJ m-2day-1 dose group, the tea tree samples are subjected to supplementary ultraviolet light irradiation treatment at the same time every day for seven consecutive days.

3. The method for distinguishing tea leaves treated with different ultraviolet light using an electronic nose according to claim 1, characterized in that: In step S2, the electronic nose system includes: a gas collection device, a gas detection device, a circuit, and a host computer; The gas collection device comprises: an air intake hose and a rubber plug, an air pump, and a flow meter. The air intake hose is fixed to the gas collection bottle through the rubber plug. The organic volatiles enriched in the gas collection bottle are sucked into the electronic nose through the hose by the air pump. The gas flow is detected by the flow meter and fed back to control the air pump. The gas detection device is composed of a sensor array, wherein at least one temperature sensor and a humidity sensor and 12 gas sensors are arranged inside the sensor array, which are used to detect volatile organic compounds of tea tree samples with different treatments, and generate original response signals to obtain an original response data set X, which is transmitted to a host computer through the circuit. The host computer receives the original response signal and generates a corresponding data file, and distinguishes the differences between tea trees with different treatments through calculation.

4. The method for distinguishing tea leaves treated with different ultraviolet light using an electronic nose according to claim 3, characterized in that: In step S3, the original response data set X is processed into a matrix with the same number of rows and columns, as shown below: Among them, X1 (n) is the nth sample data of sample set X1, x1, x2, x3, …, x 12 For the output of 12 gas sensors, m rows are intercepted as initial data according to the time series. The number of rows of the initial data is determined by the optimal range after window selection, and the number of columns is determined by the optimal sensor combination after sensor selection. Then, the corresponding samples are labeled and recorded in the label set Y1. The label set Y1 is used as the actual label to verify the prediction results.

5. The method for distinguishing tea leaves treated with different ultraviolet light using an electronic nose according to claim 4, characterized in that: In step S4, the window selection and sensor selection method is used to extract features from the sample set X1, as follows: Two parameters are set to represent the window selection range and the sensor selection number respectively. The optimal selection of the two parameters is grid searched by the control variable method. The control window range is kept unchanged at the initial value. 4 to 12 sensors are randomly selected from 12 sensors, and the response accuracy is calculated in turn to determine the best sensor selection; the selected sensors are fixed and the window range is changed in turn, and the optimal window range parameters are determined according to the accuracy; each data sample is flattened and superimposed in a matrix to obtain the sample set X2 that needs to be reduced in dimension, which is expressed as follows: The number of elements in each row is the number of rows and columns of the initial data multiplied by the number of columns, and n is the total number of samples.

6. The method for distinguishing tea leaves treated with different ultraviolet light using an electronic nose according to claim 5, characterized in that: In step S5, feature dimension reduction is performed on the sample set X2 after feature extraction, including the following steps: S5.

1. Standardize the sample set X2 in step S4: Where μ is the mean vector of each column in X2, and σ is the standard deviation vector of each column in X2; S5.

2. Use superscript T to indicate transpose and calculate the covariance matrix C: S5.

3. Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues ​​and eigenvectors: C=VΛV T Where Λ is the eigenvalue diagonal matrix and V is the eigenvector matrix; Calculate the eigenvalues ​​of the covariance matrix C and arrange them in order of size to obtain λ1≥λ2≥…≥λ m , the corresponding eigenvectors are γ1, γ2, …, γ3, which are the 1st, 2nd, …, mth principal components respectively; S5.

4. Calculate the contribution rate E of each principal component j : Among them, λ j is the jth eigenvalue, is the cumulative sum of all eigenvalues; Determine the number of principal components according to the variance contribution rate, select the eigenvectors corresponding to the first k largest eigenvalues ​​so that their cumulative contribution rate is greater than 95%, and calculate the score Z of the standardized measurement value in the direction of the first d principal components: Take the first two principal components with cumulative variance contribution rates greater than 95% as the new feature sample set to form the matrix V k ; S5.

5. Project the data matrix into the selected principal component space: Get a new feature sample set X 2-pca .

7. The method for distinguishing tea leaves treated with different ultraviolet light using an electronic nose according to claim 6, characterized in that: In step S6, the support vector machine, the decision function is expressed as follows: f(x)=sign(ω T x+b) Among them, x is the feature vector of the input sample, ω is the normal vector, which determines the direction of the hyperplane; b is the bias term, which determines the distance between the hyperplane and the origin; sign represents the sign of the function, which is 1 when it is greater than or equal to 0 and -1 when it is less than 0; If f(x)>0, it is classified as positive, and if f(x)<0, it is classified as negative.

8. The method for distinguishing tea leaves treated with different ultraviolet light using an electronic nose according to claim 6, characterized in that: In step S7, the new feature sample set X 2-pca Perform 10-fold cross validation, divide the data into 10 folds, use each fold as a validation set, and the rest as a training set, and perform 10 training and validation cycles, including the following sub-steps: S7.

1. New feature sample set X 2-pca Randomly divide into 10 equal subsets as folds, represented as: X 2-pca (1) ,X 2-pca (2) ,X 2-pca (3) ,…,X 2-pca (10) ; S7.2, using the first fold as a validation set and the remaining 9 folds as training sets, using the support vector machine classifier in step S6 to train the sample set model, and evaluating the performance of the sample set model on the first fold; S7.3, repeat the steps in S7.2 10 times, and select the first to tenth folds as the validation set; S7.

4. Summarize the results of 10 verifications, calculate the average and standard deviation, evaluate the overall performance of the sample set model, and finally output the average accuracy; the average accuracy is used to evaluate the classification of volatile organic compounds in tea leaves treated with different doses of ultraviolet light. The higher the average accuracy, the higher the prediction accuracy and the more obvious the classification effect.

9. An electronic device, characterized in that: include: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for distinguishing tea leaves treated with different ultraviolet light using an electronic nose as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, the steps in the method for distinguishing tea leaves treated with different ultraviolet light by using an electronic nose as described in any one of claims 1 to 8 are implemented.

Images