A tobacco leaf curing stage detection method based on electronic nose time-frequency MIC optimal features

Abstract

The application relates to a tobacco baking stage detection method based on an electronic nose time-frequency MIC preferred feature, in particular to a method for detecting the tobacco baking stage by fusing time-domain features and frequency-domain features in an electronic nose response signal, combining a maximum mutual information coefficient (MIC) feature optimization and a recognition model, which comprises the following steps: first, establishing a tobacco baking stage detection model based on the electronic nose time-frequency MIC preferred feature; and then, detecting the tobacco baking stage by using the detection model. The application realizes the recognition of the tobacco baking stage by fusing the time-domain features and the frequency-domain features in the electronic nose signal, and a higher accuracy is obtained; the MIC method is combined to realize feature optimization and data dimension reduction; the operation is simple, the detection is fast, the whole electronic nose collection system and the matched device are economical and practical, the application is beneficial to market promotion and application, and the application can provide a technical reference for improving the intelligent level of tobacco baking.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of tobacco intelligent curing detection, and particularly relates to a tobacco curing stage detection method for optimizing electronic nose time-frequency characteristics. BACKGROUND

[0002] Curing is a key link for highlighting the style and quality of tobacco leaves, and is an important step for the conversion and fixation of inherent chemical components and aroma substances in tobacco leaves. However, the curing state of tobacco leaves still relies on the subjective judgment of workers, and the curing process is not strong in pertinence, which has strong subjectivity and is difficult to realize the precise matching of the real-time state of tobacco leaves and the curing process, thereby affecting the conversion and accumulation of aroma substances in tobacco leaves. Therefore, it is of great practical significance to intelligently identify the curing state of tobacco leaves. For the identification of the curing state of tobacco leaves, machine vision, near-infrared spectroscopy and other technologies are commonly used for detection, which has certain improvement compared with eye observation, but still has the problems of limited observation area and poor representativeness. The electronic nose technology has the advantages of obtaining overall information of samples and good representativeness. At present, the related researches of electronic nose usually extract the maximum value, minimum value, average value, maximum first derivative, standard deviation and integral area and other time domain characteristics, and extract the frequency domain characteristics by using wavelet transform, and then combine the feature optimization method for discriminant analysis, which has good application prospect, but the research on the optimization of electronic nose time-frequency characteristics based on the discrimination of tobacco curing stage is rarely seen. SUMMARY

[0003] The present application aims to provide a tobacco curing stage detection method based on electronic nose time-frequency MIC optimized characteristics, taking the odor characteristics of tobacco leaves as the research object, collecting the tobacco leaf signals at different curing stages by using an electronic nose, then fusing and optimizing the time domain characteristics and frequency domain characteristics in the signals, and then comparing the discriminant models of different curing stages to obtain the optimized model for the discrimination of tobacco curing stage, which is used for the detection of tobacco curing stage, and aims to solve the problems of limited observation area and poor representativeness in the existing tobacco curing stage detection technology.

[0004] The technical scheme adopted by the present application is as follows:

[0005] A tobacco curing stage detection method based on electronic nose time-frequency MIC optimized characteristics, characterized by obtaining the time-frequency characteristics in the electronic nose signals at different curing stages, then optimizing the time-frequency characteristics by using MIC (maximum mutual information coefficient), and combining the discriminant model to discriminate the different curing stages of tobacco leaves, comprising the following steps:

[0006] Step one, obtaining the curing barn gas samples at different curing stages;

[0007] Step two, under room temperature conditions, obtaining the electronic nose response signals of the gas samples at different curing stages by using the same collection parameters;

[0008] Step three, pre-processing the signal of the electronic nose sensor array;

[0009] Step four, extracting the time domain features and frequency domain features in the electronic nose response signal respectively, fusing the extracted time-frequency features, and optimizing the time-frequency features by MIC;

[0010] Step five, dividing the optimized time-frequency features into a training set, a validation set and a test set, training and verifying the model by using the divided data set, optimizing the best model, and determining the parameters of the model;

[0011] Step six, if the effect of the established detection model meets the requirements, it represents that the model is feasible; otherwise, the sample data set is expanded and steps one to five are repeated until the requirements are met;

[0012] The specific steps of detecting the tobacco curing stage by using the model are as follows:

[0013] Step A, obtaining tobacco barn gas samples under different curing stages;

[0014] Step B, under room temperature conditions, the electronic nose response signal of the gas sample under different curing stages is obtained by using the same collection parameters;

[0015] Step C, pre-processing the signal of the electronic nose sensor array;

[0016] Step D, extracting the time domain features and frequency domain features in the electronic nose response signal respectively, fusing the extracted time-frequency features, and optimizing the time-frequency features by MIC;

[0017] Step E, inputting the optimized time-frequency features into the established detection model to discriminate the tobacco barn gas sample, and outputting the detection result.

[0018] When the MIC is used to optimize the time-frequency features of the electronic nose, the MIC coefficients between the time domain features and the frequency domain features and the tobacco curing stage are calculated, and then the threshold segmentation method is used to optimize the features, and the threshold is selected according to the basis that the model accuracy rate is less than 1% or no longer improved. The MIC coefficient greater than 0.36 is selected as the optimized feature combination, wherein 18 features are extracted in the time domain, 32 features are extracted in the frequency domain, and the total amount of features is reduced by 30.56%.

[0019] The electronic nose time-frequency characteristics, wherein the time domain characteristics include maximum value, minimum value, average value, variance, integral value and maximum first derivative of the electronic nose response signal, and the frequency domain characteristics refer to mean value, standard deviation and median value of the approximation coefficients and detail coefficients obtained by wavelet transform of the electronic nose response signal; the electronic nose time-frequency characteristics are arranged in the order of feature extraction, and then feature optimization is performed based on MIC; and the tobacco curing stage includes pre-yellowing stage, post-yellowing stage, pre-color fixing stage, post-color fixing stage and dry stem stage.

[0020] In the step one and step A, the gas samples in the curing barn at different curing stages are obtained at representative positions, such as the upper, middle and lower points of the geometric center of the curing barn.

[0021] In the step two and step B, the parameters are collected under room temperature conditions, including gas flow rate 2000 ml / min and reaction time 2 min.

[0022] In the step three and step C, when the signal of the electronic nose sensor array is preprocessed, the original signal needs to be corrected and smoothed, and the smoothing method includes but is not limited to sliding average method, exponential smoothing method and S-G filtering method.

[0023] In the step six and step E, the established discriminant model has a network structure including one input layer, two convolution layers, two batch normalization layers, two maximum pooling layers and one full connection layer, the input layer uses a standardization method for normalization processing of data, the convolution layer parameter is 2*1*16, the maximum pooling parameter is 2:1, the model learning rate is 0.001, and softmax is used as a mapping function.

[0024] The detection system applied in the method includes an electronic nose signal acquisition unit, a gas acquisition unit, a profiled gas chamber and a control analysis unit; the electronic nose signal acquisition unit is composed of a gas sensitive sensor array, a conditioning circuit and a communication module, wherein the gas sensitive sensor array is located inside the profiled gas chamber, the conditioning circuit supplies power to the gas sensitive sensor and performs A / D conversion on the signal, and the communication module adopts USB communication; the gas acquisition unit includes a gas pump, a speed regulator and a gas acquisition bag, the gas pump is used for the flow of the detected gas and the reducing gas in the detection system, and the speed regulator is used for regulating the flow rate of the gas pump; the profiled gas chamber is printed from a resin material and is coated with an anti-sticking coating and a gas flow guiding mechanism; the control analysis unit includes a PC and a software interface, wherein the software interface is used for real-time display, saving and discriminant analysis of the electronic nose signal.

[0025] Compared with the prior art, the present application has the following advantages:

[0026] Firstly, the detection method has the advantages of simple operation, economy, and easy popularization.

[0027] Second, the time-frequency characteristics of the fusion electronic nose signal in the detection method can more comprehensively obtain the characteristic information in the signal, and improve the detection precision of the tobacco curing stage.

[0028] Third, the present application provides technical support and reference for intelligent identification of tobacco curing stage and other object curing stage, and also provides new ideas and methods for fusion time-frequency characteristics in the field of agricultural product curing state detection. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 The electronic nose data acquisition system involved in embodiment 1 and embodiment 2 of the present application

[0030] Figure 2 The electronic nose response signal diagram of different curing stages involved in embodiment 2 of the present application

[0031] Figure 3 The TFF-MIC-1DCNN network model architecture diagram involved in embodiment 2 of the present application

[0032] Figure 4 The importance coefficient a) time domain b) frequency domain of the electronic nose time-frequency feature based on MIC calculation involved in embodiment 2 of the present application

[0033] Figure 5 The model training process diagram based on MIC preferred features involved in embodiment 2 of the present application

[0034] Figure 6 The TFF-MIC-1DCNN model external test set confusion matrix involved in embodiment 2 of the present application DETAILED DESCRIPTION

[0035] In order to make the purpose and advantages of the present application more clear and obvious, the content of the present application will be further described below in combination with specific embodiments.

[0036] Embodiment 1: A detection system structure and its use of a tobacco curing stage detection method based on electronic nose time-frequency MIC preferred features

[0037] Part a: A detection system structure of a tobacco curing stage detection method based on electronic nose time-frequency MIC preferred features

[0038] The tobacco leaf baking stage detection system comprises an electronic nose signal acquisition unit, a gas acquisition unit, a profiled gas chamber and a control analysis unit. The electronic nose signal acquisition unit is composed of a gas sensitive sensor array, a conditioning circuit and a communication module. The gas sensitive sensor array is located inside the profiled gas chamber. The conditioning circuit supplies power to the gas sensitive sensor and performs A / D conversion on the signal. The communication module uses USB communication. The gas acquisition unit comprises a gas pump, a speed regulator and a gas acquisition bag. The gas pump is used to detect the flow of the measured gas and the reducing gas in the system. The speed regulator is used to regulate the flow rate of the gas pump. The profiled gas chamber is printed from a resin material and has an anti-sticking coating and a gas flow guiding mechanism. The control analysis unit comprises a PC and a software interface. The software interface is used for real-time display, saving and discriminant analysis of the electronic nose signal.

[0039] Part B: A detection system using method of a tobacco leaf baking stage detection method based on electronic nose time-frequency MIC preferred features

[0040] b1. First, power on the system, preheat the sensor, and let the system run empty. When the sensor response curve is stable, the detection can be prepared.

[0041] b2. Open the computer and open the electronic nose signal online acquisition software. Check whether the electronic nose device and the computer are successfully connected through the software interface.

[0042] b3. Set the gas flow rate to 2000 ml / min, the acquisition frequency to 1 Hz, and the acquisition time to 2 min.

[0043] b4. Connect the activated carbon tube to the electronic nose cleaning port.

[0044] b5. Connect the gas sampling bag to the electronic nose inlet and open the gas sampling bag outlet valve.

[0045] b6. Click the start button. The software interface displays the collected electronic nose signal in real time and saves it.

[0046] Example 2: A tobacco leaf baking stage detection method based on electronic nose time-frequency MIC preferred features

[0047] Part C: Establishing a discriminant model for tobacco leaf baking stage detection

[0048] c1. Sample acquisition

[0049] Select mature, disease-free, and no obvious mechanical damage leaves. According to the local production process, the leaves are bunched and packed into the intensive curing barn. The curing process parameters are determined according to the local process. Sample at each temperature point 4 times. To improve the representativeness of the samples and the diversity of the samples, sample at the upper, middle and lower positions of the curing barn at the same time. A total of 540 samples.

[0050] c2. Sample electronic nose signal collection

[0051] The detection system includes an electronic nose signal acquisition unit, a gas acquisition unit, a profiling gas chamber, and a control and analysis unit. First, power on the system, preheat the sensor, and let the system run at no load. When the sensor response curve is stable, it can be prepared for detection. Set the gas flow rate to 2000ml / min, the acquisition frequency to 1Hz, and the acquisition time to 2min. Then connect the activated carbon tube to the electronic nose cleaning port, connect the gas sampling bag to the electronic nose inlet, and open the gas sampling bag outlet valve. Click the start button, the software interface can display the collected electronic nose signal in real time and save it. The electronic nose response signal at different baking stages is as follows Figure 2 shown.

[0052] c3. Electronic nose signal preprocessing

[0053] The electronic nose information obtained in this paper is a one-dimensional time series data with a duration of 2 minutes. To eliminate the influence of ambient gases on the sample response signal, the sample response signal needs to be baseline corrected. To reduce the influence of noise in the signal, the signal is smoothed using sliding average, exponential sliding average, and SG filtering.

[0054] c4. Time-frequency feature extraction and optimization

[0055] To characterize the information in the sensor response curve, six time-domain features are extracted from the time domain dimension of the original electronic nose response signal: maximum, minimum, mean, variance, integral, and maximum first-order derivative. The maximum and minimum values ​​reflect the dynamic characteristics of the signal. The mean, variance, and integral reflect the overall level of the signal. The maximum first-order derivative reflects the level of signal variation. Frequency-domain features extracted using wavelet transform can reveal the internal characteristics of the signal and are clearly distinguishable from the time-domain signal. In wavelet analysis, detail coefficients and approximation coefficients are used to describe the characteristic coefficients of the signal at different frequencies. The detail coefficient represents the detailed information of the signal in the high-frequency range, while the approximation coefficient represents the overall trend or slow changes of the signal in the low-frequency range.

[0056] The maximum value represents the maximum point of the sensor response value within 2 minutes, and the minimum value represents the minimum point of the sensor response value within 2 minutes. The maximum and minimum values ​​of the i-th sensor are defined as follows:

[0057]

[0058] in i is a sensor tag, x i1 , x i2 ,..., x i120is the response value of the i-th sensor. i

[0059] The average value represents the average value of the sensor response signal within 2 min. The variance represents the average of the sum of squares of the deviation of the sensor signal from the average value within 2 min. The integral value represents the integral area of the response curve of the sensor within 2 min with respect to the time axis. The average value, variance, and integral value of the i-th sensor are defined as follows:

[0060]

[0061] wherein i is the sensor label, x i1 , x i2 , x i120 is the response value of the i-th sensor, i is the average value of the response of the sensor within 2 min, μ is the expression of the response curve of the sensor within 2 min. f(t)

[0062] The maximum first derivative represents the degree of change of the sensor response signal within 2 min. The maximum first derivative of the i-th sensor is defined as follows:

[0063]

[0064] wherein i is the sensor label, x i1 , x i2 , x i120 is the response value of the i-th sensor, i is the average value of the response of the sensor within 2 min, Δt is the response frequency of the signal. The time domain feature and the sensor number are shown in Table 1.

[0065] Table 1 Correspondence between time domain feature parameters and sensors

[0066]

[0067] Let the received signal be x [ n ], then the detail coefficients and the approximation coefficients obtained by DWT are expressed by mathematical formula as follows:

[0068]

[0069] wherein c A [ j ]​​k ] is the approximation coefficient, c D [ j ][ k ] is the detail coefficient, j denotes the transform order, k denotes the subband index, denotes the inverse transform of the wavelet basis function, denotes the inverse transform of the scaling function. In this paper, the first wavelet of Daubechies wavelet is selected and one layer decomposition is performed. The present application uses the mean, standard deviation and median value of the detail coefficient and the approximation coefficient to represent the frequency domain features. The frequency domain features and the sensor number are shown in Table 2.

[0070] Table 2 Correspondence between frequency domain feature parameters and sensors

[0071]

[0072] The extracted time-frequency features are fused, and MIC is used for feature optimization. The maximal information coefficient (MIC) is a statistical measure of the relationship between two variables. Compared with traditional mutual information, MIC can capture a wider range of dependence relationships, including linear and nonlinear relationships. MIC as a feature selection standard can effectively measure the correlation strength between features and target variables. The core idea of MIC is to divide the data into different grid structures, then calculate the mutual information between variables under these grid structures, and finally select the grid partitioning method to maximize the mutual information. Specifically, MIC finds the maximum mutual information under a certain grid partitioning for any two variables X and Y , and normalizes it to get the MIC value.

[0073] (1) Grid partitioning

[0074] The variables X and Y are divided into grids respectively. Assuming X is divided into m intervals, Y is divided into n intervals, then a m × n grid is formed.

[0075] (2) Calculate mutual information

[0076] For each grid partitioning, the mutual information X and Y on the grid is calculated I ( X , Y ). Mutual information measures the amount of information one variable can provide about the other.X and Y the difference between the joint probability distribution and the marginal probability distributions:

[0077]

[0078] where, p ( x,y ) is the joint probability distribution of X and Y , p ( x ) and p ( y ) are the marginal probability distributions of X and Y .

[0079] (3) Finding the grid partition that maximizes

[0080] By trying different combinations of m x n grids, find the grid partition that maximizes the mutual information I ( X,Y ) for each m and n . Calculate the maximum mutual information and normalize its value:

[0081]

[0082] where, M ( X , Y ) is the normalized mutual information, and log2(min( m , n )) is the normalization factor.

[0083] (4) Calculating the MIC value

[0084] Among all possible grid partitions, the MIC is defined as the maximum value of the normalized mutual information M ( X , Y ):

[0085]

[0086] This value represents the maximum correlation between X and Y under the optimal grid partition and is normalized to the range [0, 1].

[0087] The MIC coefficients of the time-frequency features are shown in Figure 4 . In Figure 4In a, the top three important feature variables are min6, av6, and pv6, with MIC values of 0.8751, 0.8613, and 0.8583, respectively, indicating that these three features have a high correlation between the time domain and the tobacco curing stage. Figure 4 In b, the top three important feature variables are Dav6, Dme1, and Dme6, with MIC values of 0.7768, 0.6708, and 0.6702, respectively, indicating that these three features have a high correlation between the frequency domain and the tobacco curing stage. The present application selects feature variables with MIC values greater than 0.36 as the optimized feature combination. Among them, 18 features are extracted in the time domain, 32 features are extracted in the frequency domain, and the total number of features is reduced by 30.56%.

[0088] c5, sample data set division

[0089] Before building the model, all data is divided into a training set, a validation set, and an external test set. The training set is used to train the model, then the model parameters are adjusted according to the validation set to obtain a more accurate model, and finally the stability of the model is tested by the external test set. When building the model, 10% of the tobacco gas samples are randomly selected as the external test set, 20% of the samples are selected as the validation set, and the remaining 70% of the samples are selected as the training set. The number of samples in the training set, validation set, and external test set for detecting the tobacco curing stage model are 378, 108, and 54, respectively.

[0090] c6, final model determination of model optimization set

[0091] The features are optimized using MIC, and the optimized features are combined with SVM, 1DCNN, and LSTM models to build a tobacco curing stage discrimination model, respectively. The performance of the model is shown in Table 3, and the training process is shown in Figure 5The TF-MIC-1DCNN has the best effect when training the classification model using the optimized time-domain features of the electronic nose, with the accuracy of the training set, validation set, and test set being 92.59%, 90.74%, and 85.19%, respectively, followed by TF-MIC-SVM and TF-MIC-LSTM. The FF-MIC-1DCNN has the best effect when training the classification model using the optimized frequency-domain features of the electronic nose, with the accuracy of the training set, validation set, and test set being 92.86%, 92.59%, and 92.59%, respectively, followed by FF-MIC-LSTM and FF-MIC-SVM. The TFF-MIC-1DCNN has the best effect when training the classification model using the optimized time-frequency features, with the accuracy of the training set, validation set, and test set being 92.89%, 92.59%, and 92.59%, respectively, followed by TFF-MIC-LSTM and TFF-MIC-SVM. Compared with single features, the TFF-MIC-SVM, TFF-MIC-1DCNN, and TFF-MIC-LSTM models trained using the optimized fused features have improved accuracy. Among them, the improvement degree of TFF-MIC-LSTM is the highest, being 11.11%, followed by TFF-MIC-SVM of 7.41% and TFF-MIC-1DCNN of 5.55%. This indicates that the TFF-MIC-1DCNN has higher classification ability in the optimized feature matrix. The structure of the TFF-MIC-1DCNN model is shown in Figure 3

[0092] Table 3 Performance of different discriminant models based on MIC preferred features

[0093]

[0094] Part d, detection of tobacco curing stages using the recognition model

[0095] The detection of tobacco curing stages using the preferred recognition model is performed according to steps A-E. The preferred time-frequency features of the external test set are obtained, and the preferred TFF-MIC-1DCNN discriminant model is input with the preferred time-frequency features to obtain the discriminant result of the tobacco curing stage. The discriminant effect of the model is evaluated by comparing the accuracy of the real stage and the predicted stage, with the discriminant accuracy being 92.59%, and the confusion matrix is shown in Figure 6

[0096] ​​The application illustrates an operation process of a tobacco baking stage detection method based on an electronic nose time-frequency MIC optimal feature from two embodiments, respectively from the structure and use method of the detection system, the establishment and optimization of the model, and the test of the model discrimination effect, and the like. As can be seen from the detection result of the application model, the application realizes the classification and discrimination of the tobacco baking stage by fully utilizing the feature information in the time domain and frequency domain space of the electronic nose signal, establishing and optimizing the discrimination model based on the MIC feature optimization method, and provides technical support and reference for the baking stage identification of the tobacco baking stage and other objects.

[0097] For the detection of the baking stage of other objects based on the electronic nose technology, the detection method and the detection process proposed in the application can be referred to for operation.

[0098] The above embodiments are only used to illustrate the application, and are not used to limit the application, any modification, equivalent replacement, improvement made within the spirit and principle of the application, and on the basis of the technical essence of the application, should be included in the scope of the application, and the patent protection scope of the application is defined by the claims.

Claims

1. A tobacco leaf curing stage detection method based on the time-frequency MIC optimization characteristics of an electronic nose, characterized in that: The time-frequency features of the electronic nose signals at different baking stages are obtained, and then the time-frequency features are optimized using the maximum mutual information coefficient of MIC. The different baking stages of tobacco leaves are discriminated in combination with the discriminant model, including the following steps: Step 1: Obtain gas samples from the baking room at different baking stages; Step 2: Under room temperature conditions, using the same acquisition parameters to obtain the electronic nose response signals of gas samples at different baking stages; Step 3: pre-processing the signal of the electronic nose sensor array; Step 4: extract the time domain features and frequency domain features of the electronic nose response signal respectively, fuse the extracted time-frequency features, and use MIC to optimize the time-frequency features; Step 5: Divide the optimized time-frequency features into training set, validation set, and test set, and use the divided data sets to train and validate the model, select the best model, and determine the various parameters of the model; Step 6: If the established detection model meets the requirements, the model is feasible; otherwise, expand the sample data set and repeat steps 1 to 5 until the requirements are met; The specific steps of using the model to detect the tobacco leaf curing stage are as follows: Step A, obtaining gas samples from the baking room at different baking stages; Step B: Under room temperature conditions, using the same acquisition parameters to obtain electronic nose response signals of gas samples at different baking stages; Step C, preprocessing the signal of the electronic nose sensor array; Step D, respectively extracting the time domain features and frequency domain features from the electronic nose response signal, fusing the extracted time-frequency features, and optimizing the time-frequency features using MIC; Step E: Input the optimized time-frequency features into the established detection model, identify the baking room gas sample, and output the detection results; The established discriminant model has a network structure consisting of 1 input layer, 2 convolutional layers, 2 batch normalization layers, 2 maximum pooling layers, and 1 fully connected layer. The input layer uses a standardization method to normalize the data. The convolutional layer parameters are 2×1×16, the maximum pooling parameters are 2:1, the model learning rate is 0.001, and softmax is used as the mapping function.

2. The tobacco leaf curing stage detection method based on the time-frequency MIC optimization feature of the electronic nose according to claim 1, characterized in that: When optimizing the time-frequency features of the electronic nose based on MIC, the MIC coefficients between the time domain features, the frequency domain features, and the tobacco leaf curing stages should be calculated respectively. Then, the features are optimized using the threshold segmentation method. The threshold selection is based on whether the model accuracy improves by less than 1% or no longer improves. The time-frequency features with a MIC coefficient greater than 0.36 are selected as the preferred feature combination, of which 18 features are extracted from the time domain and 32 features are extracted from the frequency domain, reducing the total number of features by 30.56%.

3. The tobacco leaf curing stage detection method based on the time-frequency MIC optimization characteristics of the electronic nose according to claim 1, characterized in that: The electronic nose time-frequency characteristics, wherein the time domain characteristics include the maximum value, minimum value, average value, variance, integral value and maximum first-order derivative in the electronic nose response signal; Frequency domain features refer to the mean, standard deviation and median of the approximate coefficients and detail coefficients obtained by wavelet transform of the electronic nose response signal; the electronic nose time-frequency features are arranged in the order of feature extraction, and then feature optimization is performed based on MIC; the tobacco leaf curing stages include the early yellowing stage, the late yellowing stage, the early color fixing stage, the late color fixing stage and the dry tendon stage.

4. The method for detecting tobacco leaf curing stages based on the time-frequency MIC optimization characteristics of an electronic nose according to claim 1, characterized in that: To obtain the gas samples of the baking room at different baking stages, they should be drawn from representative positions, including the upper, middle and lower points of the geometric center of the baking room.

5. The method for detecting tobacco leaf curing stages based on the time-frequency MIC optimization characteristics of an electronic nose according to claim 1, characterized in that: The acquisition parameters included a gas flow rate of 2000 ml / min and a reaction time of 2 min at room temperature.

6. The method for detecting tobacco leaf curing stages based on the time-frequency MIC optimization characteristics of an electronic nose according to claim 1, characterized in that: When preprocessing the signal of the electronic nose sensor array, it is necessary to perform baseline correction and smoothing on the original signal. The smoothing methods include but are not limited to the sliding average method, the exponential sliding average method and the SG filtering method.

7. The method for detecting tobacco leaf curing stages based on the time-frequency MIC optimization characteristics of an electronic nose according to claim 1, characterized in that: The detection system used in the method includes an electronic nose signal acquisition unit, a gas acquisition unit, a contoured air chamber, and a control and analysis unit. The electronic nose signal acquisition unit consists of a gas sensor array, a conditioning circuit, and a communication module, wherein the gas sensor array is located inside the contoured air chamber, the conditioning circuit supplies power to the gas sensors and performs analog-to-digital conversion on the signals, and the communication module uses USB communication. The gas acquisition unit includes an air pump, a speed regulator, and a gas acquisition bag. The air pump is used to detect the flow of the test gas and the reducing gas in the system, and the speed regulator is used to control the flow rate of the air pump. The contoured air chamber is printed from a resin material and is covered with an anti-stick coating and an airflow guide mechanism. The control and analysis unit includes a PC and a software interface, wherein the software interface is used for real-time display, storage, and discrimination and analysis of electronic nose signals.

Images