CN115062656B - Tea polyphenol content prediction method and device based on electronic nose signal space domain - Google Patents

Abstract

The invention discloses a tea polyphenol content prediction method and device based on an electronic nose signal space domain, wherein the method comprises the following steps: 1. acquiring an electronic nose signal of a tea sample and measuring the content of tea polyphenol; 2. extracting response values of the tea sample electronic nose signals in a stable state as tea sample electronic nose signal characteristics; 3. based on the tea polyphenol content of the tea sample, optimizing the signal characteristics of the electronic nose of the tea sample by utilizing the maximum information coefficient to obtain a corresponding optimized electronic nose signal of the tea sample; 4. converting the tea sample preferential electronic nose signal into a spatial domain image by a code conversion method; 5. and training and testing the spatial domain image based on the convolutional neural grid CNN to obtain a prediction result of the tea polyphenol content. The invention can convert the one-dimensional time sequence of the electronic nose signal into a two-dimensional image, and can excavate the spatial domain characteristics of the electronic nose signal, and simultaneously, the CNN network can be utilized to realize timely and accurate prediction of the tea polyphenol content.

Description

A method and device for predicting tea polyphenol content based on electronic nose signal space domain

technical field

The invention relates to the fields of non-destructive testing and image processing, in particular to a method and device for predicting tea polyphenol content based on electronic nose signal space domain.

Background technique

At present, there are many detection methods for tea polyphenols content. The traditional chemical detection methods are destructive, complicated and time-consuming. As an odor sensor, the electronic nose can simulate the human sense of smell, which has strong objectivity, short response time, and easy detection. It has the advantages of fast speed and high accuracy, and does not require complex pretreatment processes and destructive traditional chemical methods. Electronic nose signal analysis has gradually become an important technical means to estimate the content of tea polyphenols. Effective feature extraction of electronic nose signals can improve the accuracy and efficiency of prediction. At present, the feature extraction of electronic nose signals is mainly performed in the time domain and frequency domain. There are two aspects of domain characteristics. The time-domain feature mainly extracts the characteristic values such as the instantaneous value, average value, and area value of the signal change, which can represent the instantaneous characteristics of the signal and the overall response trend, etc. After transforming the original time-domain signal into the frequency domain through Fourier transform or wavelet transform Frequency domain features can be extracted, which can represent the mainstream features and overall level of sensor signals. However, these methods ignore the details of the electronic nose signal changes, lose the correlation information between the original signal time sampling points, and lose a certain prediction accuracy.

Contents of the invention

In order to solve the shortcomings of the above-mentioned prior art, the present invention proposes a method and device for predicting tea polyphenol content based on the electronic nose signal space domain, in order to fully utilize the advantages of deep learning in the image field, and to mine the electronic nose signal space domain. The potential spatial domain characteristics of nasal signal can effectively improve the prediction accuracy of tea polyphenol content.

In order to achieve the above object, the technical scheme adopted in the present invention is:

The characteristics of a method for predicting the content of tea polyphenols based on the electronic nose signal space domain of the present invention include:

Step 1. Obtain the electronic nose signal of the tea sample and measure the content of tea polyphenols, wherein the electronic nose signal includes m sensor arrays, each sensor array includes a plurality of sampling points, and each sampling point corresponds to one of the electronic nose signals Response;

Step 2, according to the response trend of the tea sample electronic nose signal, extract the response value of the tea sample electronic nose signal in a steady state, thereby obtaining m electronic nose signal features;

Step 3. Based on the tea polyphenol content of the tea samples, the m electronic nose signal features are optimized by using the maximum information coefficient, and the optimal electronic nose signals corresponding to the H optimal features are obtained;

Step 4, converting the preferred electronic nose signal of the tea sample into a spatial domain image by a code conversion method;

Step 5, dividing the space domain image of the tea sample electronic nose signal constructed after the electronic nose signal encoding into a training set space domain image and a test set space domain image;

Step 6, training the spatial domain images of the training set of the electronic nose signal of the tea samples based on the CNN model to obtain a trained tea polyphenol content prediction model;

Step 7: Using the trained tea polyphenol content prediction model, process the test set spatial domain image of the tea samples to be predicted, and output the prediction result of tea polyphenol content.

A method for predicting tea polyphenols content based on electronic nose signal space domain according to the present invention is also characterized in that said step 3 includes:

Step 3.1, forming a two-dimensional data set corresponding to each electronic nose signal feature of the tea sample and the tea polyphenol content of the tea sample;

Step 3.2, divide the scatter diagram formed by any two-dimensional data set into grids with different numbers of rows and columns, so as to obtain grids with different division forms;

Step 3.3, respectively calculate the mutual information value corresponding to the grid of each division form, and perform normalization processing on each mutual information value;

Step 3.4, taking the normalized maximum mutual information value as the maximum information coefficient value of the corresponding two-dimensional data set;

Step 3.5: Sort the largest information coefficients of the m two-dimensional data sets in descending order, and select the electronic nose signal features corresponding to the first H largest information coefficients as H preferred features.

The step 4 includes:

Step 4.1, performing normalization processing on the optimized tea sample electronic nose signal to obtain the normalized tea sample optimal electronic nose signal;

Step 4.2, using the inverse trigonometric function formula to convert the normalized optimal electronic nose signal of the tea sample in the Cartesian coordinate system to the polar coordinate system, so as to obtain the radius sum corresponding to the optimal electronic nose signal of each tea sample in polar coordinates Angle; wherein, the radius represents the timestamp, and the angle represents the response value of the tea sample preferred electronic nose signal;

Step 4.3: Add the angle value of each optimal electronic nose signal of the tea sample in the polar coordinate system to the angle values of other sampling points respectively, and then take the cosine value of the addition result to form the corresponding optimal electronic nose signal Gram and corner field matrices;

Step 4.4: Add the angle value of each optimal electronic nose signal of the tea sample in the polar coordinate system to the angle values of other sampling points respectively, and then take the sine value of the addition result to form the corresponding optimal electronic nose signal Gram difference angle field matrix;

Step 4.5, performing a weighted average operation on the Gram sum angle field matrix and the Gram difference angle field matrix of each preferred electronic nose signal of the tea sample, respectively, to obtain the weighted average fusion matrix of the preferred electronic nose signal corresponding to the tea sample;

In step 4.6, the weighted average fusion matrix obtained from each optimal electronic nose signal of the tea sample is used as a channel of the image to obtain a spatial domain image.

Said step 6 comprises:

Step 6.1, using a convolution kernel to perform a convolution operation on the training set spatial domain image to obtain a training set feature map of the training set spatial domain image;

Step 6.2, using the pooling layer to perform a down-sampling operation on the feature map of the training set to obtain the spatial domain features of the training set of the electronic nose signal of the tea sample after dimensionality reduction;

Step 6.3, using the fully connected layer to integrate the spatial domain features of the training set of the electronic nose signal of the tea sample after the dimensionality reduction, to obtain the predicted value of the tea polyphenol content of the tea sample;

Step 6.4, using the gradient descent method to train the CNN model, and calculating the mean square error (MSE) loss function for updating the model parameters, and stopping the training when the loss function converges, so as to obtain the trained tea polyphenol content prediction model;

A tea polyphenol content prediction device based on the electronic nose signal space domain of the present invention is characterized in that it includes: an acquisition unit, a feature extraction unit, an optimization unit, a space domain image generation unit, a training unit, and a prediction unit, wherein,

The acquisition unit is used to acquire the electronic nose signal of the tea sample and measure the content of tea polyphenols, wherein the electronic nose signal includes a plurality of sensor arrays, each sensor array includes a plurality of sampling points, and each sampling point has a response value;

The feature extraction unit is used to extract the response value of each electronic nose signal of the tea sample in a steady state according to the response trend of the tea sample electronic nose signal and use it as the tea sample electronic nose signal feature;

The optimization unit is used to optimize the characteristics of the electronic nose signal of the tea sample by using the maximum information coefficient based on the tea polyphenol content of the tea sample, and obtain the optimal electronic nose signal of the tea sample corresponding to the optimal feature;

The spatial domain image generation unit is used to encode tea samples, preferably electronic nose signals, into spatial domain images;

The training unit trains the spatial domain image of the electronic nose signal of the tea sample based on the CNN model to obtain a trained tea polyphenol content prediction model;

The prediction unit uses the tea polyphenol content prediction model to process the spatial domain image of the tea sample to be predicted, and outputs the prediction result of the tea polyphenol content.

The tea polyphenol content prediction device of the present invention is characterized in that the spatial domain image generation unit includes the following processes:

Performing normalization processing on the optimized tea sample electronic nose signal to obtain the normalized tea sample optimal electronic nose signal;

Using the inverse trigonometric function formula, the normalized tea sample optimal electronic nose signal in the Cartesian coordinate system is transformed into the polar coordinate system, so as to obtain the radius and angle corresponding to the optimal electronic nose signal for each tea sample in polar coordinates; where , the radius represents the time stamp, and the angle represents the response value of the preferred electronic nose signal of the tea sample;

Add the angle value of each optimal electronic nose signal of the tea sample in the polar coordinate system to the angle values of other sampling points respectively, and then take the cosine value of the addition result to form the Gram sum of the corresponding optimal electronic nose signal corner field matrix;

Add the angle value of each optimal electronic nose signal of tea samples in the polar coordinate system to the angle values of other sampling points respectively, and then take the sine value of the addition result to form the Gram difference of the corresponding optimal electronic nose signal corner field matrix;

Carrying out the weighted average calculation of the Gram sum angle field matrix and the Gram difference angle field matrix of each preferred electronic nose signal of the tea sample respectively, to obtain the weighted average fusion matrix of the preferred electronic nose signal corresponding to the tea sample;

The weighted average fusion matrix obtained from each optimal electronic nose signal of the tea sample is used as a channel of the image to obtain a spatial domain image.

Compared with prior art, the beneficial effect of the present invention is:

1. The present invention encodes the one-dimensional sequence of the electronic nose signal into a two-dimensional image, constructs a deep learning prediction model of tea polyphenol content based on spatial domain features, fully utilizes the advantages of deep learning in the image field, and mines the electronic nose signal Potential spatial domain features, thereby improving the prediction accuracy of tea polyphenol content;

2. The present invention adopts electronic nose technology to collect the aroma of tea leaves, and directly predicts the content of tea polyphenols by building a model without damaging the tea samples, which greatly improves the prediction efficiency and has real-time performance;

3. The present invention performs a weighted average operation on the Gram sum angle field matrix and the Gram difference angle field matrix, and combines the complementary information of the two matrices, which can not only reduce or suppress the repeated information of the spatial domain features, but also maximize the Exploiting Spatial Domain Features of Electronic Nose Signals.

Description of drawings

Fig. 1 is a flow chart of a method for predicting tea polyphenols content based on electronic nose signal space domain of the present invention;

Fig. 2 is the transcoding image based on electronic nose signal of the present invention;

Fig. 3 is a schematic diagram of a device for predicting tea polyphenol content based on the electronic nose signal space domain of the present invention.

Detailed ways

In the present embodiment, referring to Fig. 1, a method for predicting the content of tea polyphenols based on the signal space domain of the electronic nose is carried out as follows:

Step 1. Obtain the electronic nose signal of the tea sample and measure the content of tea polyphenols, wherein the electronic nose signal includes multiple sensor arrays, each sensor array includes multiple sampling points, and each sampling point has a response value;

In this example, the PEN3 type portable electronic nose is used, and the electronic nose includes ten sensors in total. Each sensor corresponds to a different sensitive compound. Three varieties of green tea samples, Houkui, Huangshan Maofeng, and Liuan Guapian, are obtained, totaling 90 For each green tea sample, 5g of each tea sample is put into a 100mL glass beaker and sealed for 30min, the sampling interval is 1s, the cleaning time of the sensor is 100s, the return time of the sensor is 10s, and the sampling time is 75s.

Step 2, according to the response trend of the electronic nose signal of the tea sample, extract the response value of the electronic nose signal of the tea sample in a steady state, thereby obtaining m electronic nose signal features;

Step 3. Based on the tea polyphenol content of the tea samples, the m electronic nose signal features are optimized by using the maximum information coefficient, and the optimal electronic nose signals corresponding to the H optimal features are obtained;

Step 3.1, forming a two-dimensional data set corresponding to each electronic nose signal feature of the tea sample and the tea polyphenol content of the tea sample;

Step 3.2, divide the scatter diagram formed by the two-dimensional data set into grids with different numbers of rows and columns, so as to obtain grids with different division forms;

Step 3.3, respectively calculate the mutual information value corresponding to the grid of each division form, and perform normalization processing on each mutual information value;

Step 3.4, taking the normalized maximum mutual information value as the maximum information coefficient value of the corresponding two-dimensional data set;

Step 3.5: Sort the maximum information coefficients of the two-dimensional data set in descending order, and select the electronic nose signal features corresponding to the top H largest information coefficients, thereby selecting the top H electronic nose signals.

In this embodiment, 3 electronic nose signals are optimized out of 10 electronic nose signals. Assuming that the vector composed of all sample points of the time domain characteristics of each electronic nose signal is a, the vector composed of all sample points of tea polyphenol content is b, and the calculation formula of the mutual information I(a,b) between the two is as follows: (1) as shown:

In formula (1), p(a) represents the probability density of time domain feature a, p(b) represents the probability density of tea polyphenol content b, and p(a,b) represents time domain feature a and tea polyphenol content b The joint probability density function is very difficult to calculate, so MIC realizes its calculation through discrete methods. Assuming that D(a,b) is a two-dimensional data set composed of time-domain feature a and tea polyphenol content b, the two-dimensional space is divided into x segments and y segments in the X and Y directions respectively to form x×y The grid G of , let the set of grid G based on different division forms be Ω, then define:

In formula (2), D|G represents the distribution of data set D on grid G. I ^* (D,x,y) represents the mutual information of the grid divided by x rows and y columns.

The feature matrix M(D) is composed of the maximum normalization of the data set D under different division methods, as shown in formula (3):

In formula (3), min{x,y} represents the smallest x-row-y-column grid, and M(D) _x,y denotes the normalized mutual information value of the x-row-y-column grid.

Solve the maximum value of the characteristic matrix within the upper limit of the number of grid divisions to obtain the maximum information coefficient, as shown in formula (4):

In formulas (4) and (5), B(n) is the upper limit of the grid division number, and MIC(D) represents the maximum normalized mutual information value of the form grid that cannot be divided.

Step 4, converting the preferred electronic nose signal of the tea sample into a spatial domain image by a code conversion method;

Step 4.1, performing normalization processing on the optimized tea sample electronic nose signal to obtain the normalized tea sample optimal electronic nose signal;

Step 4.2, using the inverse trigonometric function formula to convert the normalized optimal electronic nose signal of the tea sample in the Cartesian coordinate system to the polar coordinate system, so as to obtain the radius sum corresponding to the optimal electronic nose signal of each tea sample in polar coordinates Angle; wherein, the radius represents the timestamp, and the angle represents the response value of the tea sample preferred electronic nose signal;

Step 4.3: Add the angle value of each optimal electronic nose signal of the tea sample in the polar coordinate system to the angle values of other sampling points respectively, and then take the cosine value of the addition result to form the corresponding optimal electronic nose signal Gram and corner field matrices;

Step 4.4: Add the angle value of each optimal electronic nose signal of the tea sample in the polar coordinate system to the angle values of other sampling points respectively, and then take the sine value of the addition result to form the corresponding optimal electronic nose signal Gram difference angle field matrix;

Step 4.5, performing a weighted average operation on the Gram sum angle field matrix and the Gram difference angle field matrix of each preferred electronic nose signal of the tea sample, respectively, to obtain the weighted average fusion matrix of the preferred electronic nose signal corresponding to the tea sample;

In step 4.6, the weighted average fusion matrix obtained from each optimal electronic nose signal of the tea sample is used as a channel of the image to obtain a spatial domain image.

In the example of the present invention, for the one-dimensional time series X={x ₁ ,x ₂ ,···, _xi ,···,x _n } of the electronic nose signal, each sampling point signal of the electronic nose signal is equal to The ratio is mapped to the [-1,1] interval, as shown in equations (5) and (6):

X′=[x′ ₁ ,x′ ₂ ,···,x′ _i ,···,x′ _n ],-1≤xi _′ ≤1 (6)

In formulas (5) and (6), x _i is the response value of the i-th sampling point, x _i ' is the response value of the i-th sampling point after normalization, n is the total number of sampling points, and X' is Normalized time series.

Map the obtained new electronic nose signal time series to the polar coordinate system, and use the inverse trigonometric function to obtain the angle corresponding to each sequence point. At this time, the sequence value range under the polar coordinates is [0, π], as shown in formula (7 ), Formula (8) shows:

x″ _i = arccos(x′ _i ),-1≤x′ _i ≤1 (7)

X″=[x″ ₁ ,x″ ₂ ,...,x″ _n ], 0≤x″ ₁ ≤π (8)

In formula (7) and formula (8), x″ _i is the angle value of the i-th response value of the electronic nose signal converted to polar coordinates, and X″ is the new sequence of electronic nose time series converted to polar coordinates.

Add the time series in the polar coordinate system and then take the cosine value and summarize them to obtain the Gram sum angle field matrix GASF, as shown in formula (9):

Add the time series in the polar coordinate system and then take the sine value for summary to obtain the Gram difference angle field matrix GADF, as shown in formula (10);

Perform weighted average summation on the two matrices to obtain the target matrix Fusion, as shown in formula (11);

The weighted average fusion matrix obtained from the time series of each electronic nose signal of the tea sample is used as the image channel to obtain the spatial domain image. The electronic nose signal of the tea sample is encoded into a two-dimensional image. This transformation can maintain the time dependence of the electronic nose signal. The texture information and color distribution information of the two-dimensional image can reflect the potential spatial domain information of the one-dimensional signal. At the same time, using CNN The model achieves the technical effect of improving the accuracy and efficiency of tea polyphenols prediction content. Figure 2 is the encoded image of the electronic nose signal.

Step 5, dividing the tea sample electronic nose signal space domain image constructed after electronic nose signal encoding into a training set space domain image and a test set space domain image;

Step 6. Based on the CNN model, the spatial domain images of the tea sample electronic nose signal training set are trained to obtain a trained tea polyphenol content prediction model;

Step 6.1, using the convolution kernel to perform convolution operation on the training set spatial domain image to obtain the training set feature map of the training set spatial domain image;

Step 6.2, using the pooling layer to perform down-sampling operation on the feature map of the training set to obtain the spatial domain features of the tea sample electronic nose signal training set after dimensionality reduction;

Step 6.3, using the fully connected layer to integrate the spatial domain features of the tea sample electronic nose signal training set after dimensionality reduction, to obtain the predicted value of the tea polyphenol content of the tea sample;

Step 6.4, use the gradient descent method to train the CNN model, and calculate the mean square error MSE loss function, which is used to update the model parameters, and stop the training when the loss function converges, so as to obtain the tea polyphenol content prediction model after training;

In the example of the present invention, the CNN network used has 128 filters in the first convolutional layer with a size of 3×3 and a stride of 1, followed by a 3×3 average pooling layer, and the second convolutional layer layer with 64 filters of size 3×3 and stride of 1, followed by a 3×3 average pooling layer, the second convolutional layer has 32 filters of size 3×3 and stride of 1, Then there is a 3×3 average pooling layer. The activation function used in each convolutional layer is relu. After each pooling layer, dropout is used to prevent overfitting. Finally, the predicted value of tea polyphenols content is output through the fully connected layer. The batch size (Batch_size) set by the model is 24, the number of training (epochs) is 200, the optimizer used is Adam, and the loss function used is mse.

Step 7: Using the trained tea polyphenol content prediction model, process the test set spatial domain images of tea samples to be predicted, and output the prediction result of tea polyphenol content.

With reference to Fig. 3, in the present embodiment, a kind of tea polyphenols content prediction device based on electronic nose signal space domain, comprises: acquisition unit, feature extraction unit, optimization unit, space domain image generation unit, training unit, prediction unit, wherein ,

The acquisition unit is used to acquire the electronic nose signal of the tea sample and measure the content of tea polyphenols, wherein the electronic nose signal includes a plurality of sensor arrays, each sensor array includes a plurality of sampling points, and each sampling point has a response value;

The feature extraction unit is used to extract the response value of each electronic nose signal of the tea sample in a steady state according to the response trend of the tea sample electronic nose signal and use it as the tea sample electronic nose signal feature;

The optimization unit is used to optimize the characteristics of the electronic nose signal of the tea sample by using the maximum information coefficient based on the tea polyphenol content of the tea sample, and obtain the optimal electronic nose signal of the tea sample corresponding to the optimal feature;

The spatial domain image generation unit is to encode the preferred electronic nose signal of the tea sample into a spatial domain image through a code conversion method; specifically, the spatial domain image generation unit first performs normalization processing on the optimized tea sample electronic nose signal, The optimal electronic nose signal of the normalized tea sample is obtained; secondly, using the inverse trigonometric function formula, the normalized optimal electronic nose signal of the tea sample in the Cartesian coordinate system is transformed into the polar coordinate system, thereby obtaining the polar coordinate The radius and angle corresponding to the preferred electronic nose signal of each tea sample; wherein, the radius represents the time stamp, and the angle represents the response value of the preferred electronic nose signal of the tea sample; then, each preferred electronic nose signal of the tea sample in the polar coordinate system is After the angle values of each sampling point are added to the angle values of other sampling points, the cosine value of the addition result is taken to form the Gram and angle field matrix corresponding to the optimal electronic nose signal; After the angle value of the electronic nose signal at each sampling point is added to the angle value of other sampling points, the sine value of the addition result is taken to form the Gram difference angle field matrix corresponding to the optimal electronic nose signal; The Gram sum angle field matrix and the Gram difference angle field matrix of the optimal electronic nose signal are respectively weighted and averaged to obtain the weighted average fusion matrix of the optimal electronic nose signal corresponding to the tea sample; finally, each optimal electronic nose signal of the tea sample is obtained The weighted average fusion matrix of is used as a channel of the image to obtain a spatial domain image.

The training unit is to train the CNN model based on the spatial domain image, and obtain the tea polyphenol content prediction model after training;

The prediction unit processes the spatial domain image of the tea sample to be predicted by using the tea polyphenol content prediction model, and outputs the prediction result of the tea polyphenol content.

The system makes full use of the advantages of deep learning in the image field. It not only maintains the time dependence of the electronic nose signal, but also mines the potential spatial domain characteristics of the electronic nose signal, thereby improving the prediction accuracy of tea polyphenols content.

Claims (4)

1. A method for predicting tea polyphenols content based on electronic nose signal space domain, its characteristics include: Step 1. Obtain the electronic nose signal of the tea sample and measure the content of tea polyphenols, wherein the electronic nose signal includes m sensor arrays, each sensor array includes a plurality of sampling points, and each sampling point corresponds to one of the electronic nose signals Response; Step 2, according to the response trend of the tea sample electronic nose signal, extract the response value of the tea sample electronic nose signal in a steady state, thereby obtaining m electronic nose signal features; Step 3. Based on the tea polyphenol content of the tea samples, the m electronic nose signal features are optimized by using the maximum information coefficient, and the optimal electronic nose signals corresponding to the H optimal features are obtained; Step 4, converting the preferred electronic nose signal of the tea sample into a spatial domain image by a code conversion method; Step 4.1, performing normalization processing on the optimized tea sample electronic nose signal to obtain the normalized tea sample optimal electronic nose signal; Step 4.2, using the inverse trigonometric function formula to convert the normalized optimal electronic nose signal of the tea sample in the Cartesian coordinate system to the polar coordinate system, so as to obtain the radius sum corresponding to the optimal electronic nose signal of each tea sample in polar coordinates Angle; wherein, the radius represents the timestamp, and the angle represents the response value of the tea sample preferred electronic nose signal; Step 4.3: Add the angle value of each optimal electronic nose signal of the tea sample in the polar coordinate system to the angle values of other sampling points respectively, and then take the cosine value of the addition result to form the corresponding optimal electronic nose signal Gram and corner field matrices; Step 4.4: Add the angle value of each optimal electronic nose signal of the tea sample in the polar coordinate system to the angle values of other sampling points respectively, and then take the sine value of the addition result to form the corresponding optimal electronic nose signal Gram difference angle field matrix; Step 4.5, performing a weighted average operation on the Gram sum angle field matrix and the Gram difference angle field matrix of each preferred electronic nose signal of the tea sample, respectively, to obtain the weighted average fusion matrix of the preferred electronic nose signal corresponding to the tea sample; Step 4.6, using the weighted average fusion matrix obtained from each optimal electronic nose signal of the tea sample as a channel of the image, thereby obtaining a spatial domain image; Step 5, dividing the space domain image of the tea sample electronic nose signal constructed after the electronic nose signal encoding into a training set space domain image and a test set space domain image; Step 6, training the spatial domain images of the training set of the electronic nose signal of the tea samples based on the CNN model to obtain a trained tea polyphenol content prediction model; Step 7: Using the trained tea polyphenol content prediction model, process the test set spatial domain image of the tea samples to be predicted, and output the prediction result of tea polyphenol content. 2. a kind of tea polyphenols content prediction method based on electronic nose signal space domain according to claim 1, is characterized in that, described step 3 comprises: Step 3.1, forming a two-dimensional data set corresponding to each electronic nose signal feature of the tea sample and the tea polyphenol content of the tea sample; Step 3.2, divide the scatter diagram formed by any two-dimensional data set into grids with different numbers of rows and columns, so as to obtain grids with different division forms; Step 3.3, respectively calculate the mutual information value corresponding to the grid of each division form, and perform normalization processing on each mutual information value; Step 3.4, taking the normalized maximum mutual information value as the maximum information coefficient value of the corresponding two-dimensional data set; Step 3.5: Sort the largest information coefficients of the m two-dimensional data sets in descending order, and select the electronic nose signal features corresponding to the first H largest information coefficients as H preferred features. 3. a kind of tea polyphenols content prediction method based on electronic nose signal space domain according to claim 1, is characterized in that, described step 6 comprises: Step 6.1, using a convolution kernel to perform a convolution operation on the training set spatial domain image to obtain a training set feature map of the training set spatial domain image; Step 6.2, using the pooling layer to perform a down-sampling operation on the feature map of the training set to obtain the spatial domain features of the training set of the electronic nose signal of the tea sample after dimensionality reduction; Step 6.3, using the fully connected layer to integrate the spatial domain features of the training set of the electronic nose signal of the tea sample after the dimensionality reduction, to obtain the predicted value of the tea polyphenol content of the tea sample; Step 6.4: Use the gradient descent method to train the CNN model, and calculate the mean square error (MSE) loss function for updating model parameters, and stop training when the loss function converges, thereby obtaining a trained tea polyphenol content prediction model. 4. A tea polyphenol content prediction device based on electronic nose signal space domain, it is characterized in that, comprising: acquisition unit, feature extraction unit, optimization unit, space domain image generation unit, training unit, prediction unit, wherein, The acquisition unit is used to acquire the electronic nose signal of the tea sample and measure the content of tea polyphenols, wherein the electronic nose signal includes a plurality of sensor arrays, each sensor array includes a plurality of sampling points, and each sampling point has a response value; The feature extraction unit is used to extract the response value of each electronic nose signal of the tea sample in a steady state according to the response trend of the tea sample electronic nose signal and use it as the tea sample electronic nose signal feature; The optimization unit is used to optimize the characteristics of the electronic nose signal of the tea sample by using the maximum information coefficient based on the tea polyphenol content of the tea sample, and obtain the optimal electronic nose signal of the tea sample corresponding to the optimal feature; The spatial domain image generation unit is used to encode tea samples, preferably electronic nose signals, into spatial domain images; The spatial domain image generation unit includes the following processes: Performing normalization processing on the optimized tea sample electronic nose signal to obtain the normalized tea sample optimal electronic nose signal; Using the inverse trigonometric function formula, the normalized tea sample optimal electronic nose signal in the Cartesian coordinate system is transformed into the polar coordinate system, so as to obtain the radius and angle corresponding to the optimal electronic nose signal for each tea sample in polar coordinates; where , the radius represents the time stamp, and the angle represents the response value of the preferred electronic nose signal of the tea sample; Add the angle value of each optimal electronic nose signal of the tea sample in the polar coordinate system to the angle values of other sampling points respectively, and then take the cosine value of the addition result to form the Gram sum of the corresponding optimal electronic nose signal corner field matrix; Add the angle value of each optimal electronic nose signal of tea samples in the polar coordinate system to the angle values of other sampling points respectively, and then take the sine value of the addition result to form the Gram difference of the corresponding optimal electronic nose signal corner field matrix; Carrying out the weighted average calculation of the Gram sum angle field matrix and the Gram difference angle field matrix of each preferred electronic nose signal of the tea sample respectively, to obtain the weighted average fusion matrix of the preferred electronic nose signal corresponding to the tea sample; The weighted average fusion matrix obtained by each optimal electronic nose signal of the tea sample is used as a channel of the image to obtain a spatial domain image; The training unit trains the spatial domain image of the electronic nose signal of the tea sample based on the CNN model to obtain a trained tea polyphenol content prediction model; The prediction unit uses the tea polyphenol content prediction model to process the spatial domain image of the tea sample to be predicted, and outputs the prediction result of the tea polyphenol content.