Tea leaf fermentation degree intelligent discrimination method and system based on near infrared spectrum
Through an intelligent discrimination method combining near-infrared spectroscopy and chemical analysis, an intelligent discrimination model of tea fermentation degree was established, which solved the accuracy and efficiency problems in monitoring tea fermentation degree, and achieved efficient and accurate discrimination of tea fermentation degree.
Patent Information
- Application Number
- CN202411120638.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-15
- Publication Date
- 2024-11-22
AI Technical Summary
The prior art has problems of low accuracy, slow efficiency and susceptibility to external environment in monitoring and quality evaluation of tea fermentation degree, making it difficult to achieve intelligent discrimination of tea fermentation degree.
The near-infrared spectrometer was used to scan the tea samples, and the content of theaflavin, tea polyphenols and amino acids was determined in combination with chemical analysis methods. The intelligent discriminant model was established through pretreatment and feature extraction, and the support vector machine of the Gaussian radial basis function was used for the Classify to achieve intelligent discrimination of the degree of tea fermentation.
It improves the accuracy and efficiency of the determination of tea fermentation degree, reduces the influence of human factors, and enhances the scientificity and consistency of tea quality control.
Smart Images

Figure CN119000591A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of tea fermentation degree identification, and more specifically to an intelligent tea fermentation degree identification method and system based on near infrared spectroscopy. Background Art
[0002] At present, the production and export of tea plays an important role in my country's national economy, but it lags behind in quality control and quality evaluation. At present, the methods for evaluating tea quality are sensory evaluation and physical and chemical component analysis. The former has a higher degree of differentiation, but it is easily affected by human factors and the external environment, resulting in large differences in quality identification results, and the efficiency of sensory evaluation is low; the latter can quantitatively analyze a single component of tea with high accuracy, but the analysis cycle is long, the procedure is cumbersome, and it is difficult to make a comprehensive and holistic evaluation.
[0003] However, currently, the commonly used methods for controlling the degree of fermentation include sensory judgment, leaf temperature measurement, biochemical index method, electronic nose method, etc. At the same time, some fermentation indicator equipment has been developed, but these equipment are complicated to operate and expensive. In actual production, the regulation of the degree of fermentation is mainly determined by the processing personnel based on their own senses from the changes in the characteristics of the fermented leaves, such as the degree of redness of the fermented leaves, the changes in aroma, etc. to determine whether the fermentation has reached the right level. The accuracy is easily affected by experience and the external environment, resulting in light or excessive fermentation, which directly affects the product quality, the stability and uniformity of the flavor, and greatly increases the risk factors of quality and safety.
[0004] At present, infrared spectroscopy has become the first choice for detection due to its pollution-free, non-destructive sampling, real-time and easy use. At the same time, the rapid development and wide penetration of information science and technology have become an important feature of today's society. Various identification theories and technologies have emerged and have been widely used.
[0005] Therefore, how to intelligently determine the fermentation degree of tea leaves based on near-infrared spectroscopy is an urgent problem that technicians in this field need to solve. Summary of the invention
[0006] In view of this, the present invention provides a method and system for intelligently distinguishing the fermentation degree of tea leaves based on near infrared spectroscopy to solve the problems existing in the background technology.
[0007] In order to achieve the above object, the present invention adopts the following technical solution:
[0008] On the one hand, a method for intelligently distinguishing the fermentation degree of tea leaves based on near infrared spectroscopy is provided, comprising:
[0009] Use a near-infrared spectrometer to perform spectral scanning on tea samples with different fermentation degrees to obtain near-infrared spectral data; determine the contents of theaflavins, tea polyphenols, and amino acids in the tea samples by chemical analysis methods;
[0010] Preprocessing the acquired near infrared spectrum data;
[0011] The preprocessed spectral data is mapped to the feature space, the weight coefficients mapped to each feature vector in the feature space are extracted, and the measured theaflavins, tea polyphenols and amino acid content data are used as feature variables;
[0012] The eigenvector of the spectral data is concatenated with the eigenvariable to obtain a comprehensive eigenvector;
[0013] Establish an intelligent identification model for tea fermentation degree;
[0014] The comprehensive feature vector and the corresponding tea fermentation degree label data are input into the tea fermentation degree intelligent discrimination model for training and verification, so as to obtain the verified tea fermentation degree intelligent discrimination model;
[0015] After the tea samples to be tested are subjected to spectral scanning and chemical analysis, the feature vector to be discriminated after preprocessing and feature extraction is input into the verified intelligent discrimination model of tea fermentation degree to obtain the discrimination result of tea fermentation degree.
[0016] Optionally, the preprocessing includes first using a residual mean-variance distribution diagram based on Monte Carlo cross-validation to eliminate abnormal samples, then randomly selecting more than 120 samples from the data set from which abnormal samples have been eliminated as training sets and validation sets, and using the Kernard-Stone method to divide the training set and the validation set into a ratio of 3:1; and performing a second-order derivative of the spectral data of the training set.
[0017] Optionally, the formula for extracting the mapping is:
[0018] y i =w T x i ;
[0019] Among them, w represents the characteristic matrix, and the optimization equation for solving w is as follows:
[0020]
[0021] J(w) represents the cost function of the optimization equation,
[0022] In the formula, η represents the regularization parameter; x i represents the result of the second-order derivative of the near-infrared spectrum vector of the i-th sample, and x i =[xi (λ1)…x i (λ k )] T , the subscript k represents the dimension of the sample spectrum; y i For x i The weight coefficient projected onto the eigenvectors that constitute the feature matrix w determines the position of sample i in the feature space; the continuous variable u i ∈[0,1] represents x i The probability of belonging to a non-noise sample; (1-u i ) represents x i The probability of belonging to a noise sample; the fuzzy variable m∈[1,∞) represents the weight index; e(x i ) represents the original spectrum x i The error between the feature extraction and the reconstructed spectrum is
[0023] Optional, continuous variable u i The constraint formula is as follows:
[0024]
[0025] According to the constraint formula, we can get u i , whose expression is:
[0026]
[0027] According to the calculated u i The simplified formula of the optimization equation J(w) is:
[0028]
[0029] Optionally, the establishment of the intelligent discrimination model for the fermentation degree of tea leaves is specifically as follows: a support vector machine based on Gaussian radial basis function is used as a classifier, combined with a weight coefficient, to establish the intelligent discrimination model for the fermentation degree of tea leaves.
[0030] Optionally, the performance of the intelligent discrimination model for tea fermentation degree is evaluated by using classification error rate and accuracy, specifically including:
[0031] Assume the sample set is A, and the classification error rate is:
[0032]
[0033] The accuracy is given by:
[0034]
[0035] Where I(·) is the indicator function, which takes 1 when · is true and takes 0 when it is false; m is the total number of samples; o(p i ) is the predicted value, q i is the true sample value.
[0036] On the other hand, a tea fermentation degree intelligent identification system based on near infrared spectroscopy is provided, comprising:
[0037] The data acquisition module uses a near-infrared spectrometer to perform spectral scanning on tea samples of different fermentation degrees to obtain near-infrared spectral data; the content of theaflavins, tea polyphenols, and amino acids in the tea samples is determined by chemical analysis methods;
[0038] A data preprocessing module is used to preprocess the acquired near infrared spectrum data;
[0039] The data feature extraction module maps the preprocessed spectral data to the feature space, extracts the weight coefficients mapped to each feature vector in the feature space, and uses the measured theaflavins, tea polyphenols, and amino acid content data as feature variables;
[0040] The data splicing module splices the characteristic vector of the spectral data with the characteristic variable to obtain a comprehensive characteristic vector;
[0041] Model building module, to build an intelligent model for distinguishing the degree of tea fermentation;
[0042] The model training module inputs the comprehensive feature vector and the corresponding tea fermentation degree label data into the tea fermentation degree intelligent discrimination model for training and verification, thereby obtaining a verified tea fermentation degree intelligent discrimination model;
[0043] The discrimination result output module subjects the tea samples to spectral scanning and chemical analysis, and inputs the feature vector to be discriminated after preprocessing and feature extraction into the verified intelligent discrimination model of tea fermentation degree to obtain the discrimination result of tea fermentation degree.
[0044] Optionally, the data preprocessing module includes first using a residual mean-variance distribution diagram based on Monte Carlo cross-validation to eliminate abnormal samples, then randomly selecting more than 120 samples from the data set from which abnormal samples have been eliminated as training sets and validation sets, and using the Kernard-Stone method to divide the training set and the validation set into a ratio of 3:1; and performing a second-order derivative of the spectral data of the training set.
[0045] Optionally, a model performance evaluation module is also included, which uses classification error rate and accuracy to evaluate the performance of the intelligent discrimination model for tea fermentation degree, including:
[0046] Assume the sample set is A, and the classification error rate is:
[0047]
[0048] The accuracy is given by:
[0049]
[0050] Where I(·) is the indicator function, which takes 1 when · is true and takes 0 when it is false; m is the total number of samples; o(p i ) is the predicted value, q i is the true sample value.
[0051] Through the above technical scheme, it can be known that compared with the prior art, the present invention discloses a method and system for intelligently distinguishing the fermentation degree of tea based on near-infrared spectroscopy, which uses a near-infrared spectrometer to perform spectral scanning on tea samples of different fermentation degrees to obtain near-infrared spectral data; determines the content of theaflavins, tea polyphenols and amino acids in the tea samples by chemical analysis methods; pre-processes the obtained near-infrared spectral data; maps the pre-processed spectral data to a feature space, extracts weight coefficients mapped to each feature vector of the feature space, and uses the determined theaflavins, tea polyphenols and amino acid content data as feature variables; concatenates the feature vector of the spectral data with the feature variables to obtain a comprehensive feature vector; establishes an intelligent discrimination model for the fermentation degree of tea; inputs the comprehensive feature vector and the corresponding label data of the fermentation degree of tea into the intelligent discrimination model for tea fermentation for training and verification, and obtains a verified intelligent discrimination model for the fermentation degree of tea; performs spectral scanning and chemical analysis on the tea samples to be tested, and inputs the feature vector to be distinguished after pre-processing and feature extraction into the verified intelligent discrimination model for the fermentation degree of tea, and obtains a discrimination result of the fermentation degree of tea. The present invention combines spectral analysis and chemical analysis methods, and realizes intelligent discrimination of the degree of tea fermentation by comprehensively utilizing spectral data and chemical composition data, thereby improving the accuracy and efficiency of discriminating the degree of tea fermentation and providing convenience for tea production and quality control. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0053] Figure 1 A flow chart of the method provided by the present invention;
[0054] Figure 2 This is a system structure diagram provided by the present invention. DETAILED DESCRIPTION
[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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 are within the scope of protection of the present invention.
[0056] The embodiment of the present invention discloses an intelligent method for distinguishing the fermentation degree of tea leaves based on near infrared spectroscopy. Figure 1 As shown, including:
[0057] Use a near-infrared spectrometer to perform spectral scanning on tea samples with different fermentation degrees to obtain near-infrared spectral data; determine the contents of theaflavins, tea polyphenols, and amino acids in the tea samples by chemical analysis methods;
[0058] Preprocessing the acquired near infrared spectrum data;
[0059] The preprocessed spectral data is mapped to the feature space, the weight coefficients mapped to each feature vector in the feature space are extracted, and the measured theaflavins, tea polyphenols and amino acid content data are used as feature variables;
[0060] The eigenvector of the spectral data is concatenated with the eigenvariable to obtain a comprehensive eigenvector;
[0061] Establish an intelligent identification model for tea fermentation degree;
[0062] The comprehensive feature vector and the corresponding tea fermentation degree label data are input into the tea fermentation degree intelligent discrimination model for training and verification, so as to obtain the verified tea fermentation degree intelligent discrimination model;
[0063] After the tea samples to be tested are subjected to spectral scanning and chemical analysis, the feature vector to be discriminated after preprocessing and feature extraction is input into the verified intelligent discrimination model of tea fermentation degree to obtain the discrimination result of tea fermentation degree.
[0064] In a specific embodiment, the preprocessing includes first using a residual mean-variance distribution diagram based on Monte Carlo cross-validation to eliminate abnormal samples, then randomly selecting more than 120 samples from the data set from which abnormal samples have been eliminated as training sets and validation sets, and using the Kernard-Stone method to divide the training set and the validation set into a ratio of 3:1; and performing a second-order derivative on the spectral data of the training set.
[0065] Processing spectral data by second-order derivative can amplify the signal of the original spectral data to a certain extent and eliminate the background information contained in the acquired spectral data, making the chemical groups in the original spectral data more prominent in the spectral peak, effectively improving the extractability of the features in the spectral data; the second-order derivative formula is as follows:
[0066]
[0067] Wherein, r(λ) refers to the near infrared spectrum, λ∈[780, 2500] represents the wavelength, and Δλ represents the wavelength step; in a specific embodiment, the extraction formula is:
[0068] y i =w T x i ;
[0069] Among them, w represents the characteristic matrix, and the optimization equation for solving w is as follows:
[0070]
[0071] J(w) represents the cost function of the optimization equation,
[0072] In the formula, η represents the regularization parameter; x i represents the result of the second-order derivative of the near-infrared spectrum vector of the i-th sample, and x i =[x i (λ1)…x i (λ k )] T , the subscript k represents the dimension of the sample spectrum; y i For x i The weight coefficient projected onto the eigenvectors that constitute the feature matrix w determines the position of sample i in the feature space; the continuous variable u i ∈[0,1] represents x i The probability of belonging to a non-noise sample; (1-u i ) represents x i The probability of belonging to a noise sample; the fuzzy variable m∈[1,∞) represents the weight index; e(x i ) represents the original spectrum x i The error between the feature extraction and the reconstructed spectrum is
[0073] In a specific embodiment, the continuous variable u i The constraint formula is as follows:
[0074]
[0075] According to the constraint formula, we can get ui , whose expression is:
[0076]
[0077] According to the calculated u i The simplified formula of the optimization equation J(w) is:
[0078]
[0079] use Solve for w, according to the optimization equation, It is expressed as follows
[0080]
[0081] The regularization parameter η in the feature matrix w is calculated by the following steps:
[0082] Step (1), first set the number of iterations t to 1, the limit of the number of iterations to T, and the learning coefficient α0∈(0,1];
[0083] Step (2), when t is less than T, execute steps (3) to (9);
[0084] Step (3), set i = 1, σ = 0, and learning rate α t The expression is as follows:
[0085]
[0086] Step (4), when i is less than n, execute steps (5) to (8);
[0087] Step (5), calculate y = w T x i , h=yw,v=w T h
[0088] Step (6), calculate the feature weight coefficient, and its updated iterative formula is as follows:
[0089]
[0090] Step (7), the iterative formula for updating σ is as follows:
[0091] σ new =σ old +e(x i );
[0092] Step (8), update the value of i, set i=i+1;
[0093] Step (9), update the t value, set t = t + 1; calculate the regularization parameter η, the formula is as follows:
[0094] η=(σ / n).
[0095] In this embodiment, fuzzy variables and continuous variables u are used. i To judge whether the acquired near-infrared spectral data is a noise sample, and the threshold η is automatically obtained by using an iterative method, which avoids the complexity of artificially setting the threshold and effectively reduces the adverse effects of noise samples on the classification of original data.
[0096] In the feature extraction method of this embodiment, an iterative method is used to calculate and extract the feature matrix w of the near-infrared spectral data, and its spectral data is mapped to the feature space through the feature matrix, so that the separability of the sample in the feature space is significantly improved, thereby improving the classification accuracy and generalization ability of the classification model established later.
[0097] In this embodiment, the feature matrix w and the classification model are obtained through training data, and then the verification data is directly projected using w to obtain the weight coefficient in the feature space (this coefficient also determines the distribution in the feature space), and then the obtained discriminant model is used to determine which category the verification data corresponds to, and make a judgment on the degree of fermentation, such as moderate fermentation, mild fermentation, and excessive fermentation.
[0098] In a specific embodiment, the intelligent discrimination model for the fermentation degree of tea leaves is established by using a support vector machine based on Gaussian radial basis function as a classifier and combining it with a weight coefficient to establish the intelligent discrimination model for the fermentation degree of tea leaves.
[0099] In a specific embodiment, the performance of the intelligent discrimination model for tea fermentation degree is evaluated by using classification error rate and accuracy, specifically including:
[0100] Assume the sample set is A, and the classification error rate is:
[0101]
[0102] The accuracy is given by:
[0103]
[0104] Where I(·) is the indicator function, which takes 1 when · is true and takes 0 when it is false; m is the total number of samples; o(p i ) is the predicted value, q i is the true sample value.
[0105] Error rate and accuracy are the two most commonly used performance metrics in classification tasks, which are applicable to both binary and multi-classification tasks. As the name suggests, the error rate is the ratio of the number of samples that are misclassified in the classification results to the total number of samples. Conversely, the accuracy rate is the ratio of the number of samples that are correctly classified in the classification results to the total number of samples.
[0106] On the other hand, a tea fermentation degree intelligent identification system based on near infrared spectroscopy is provided. Figure 2 As shown, including:
[0107] The data acquisition module uses a near-infrared spectrometer to perform spectral scanning on tea samples of different fermentation degrees to obtain near-infrared spectral data; the content of theaflavins, tea polyphenols, and amino acids in the tea samples is determined by chemical analysis methods;
[0108] A data preprocessing module is used to preprocess the acquired near infrared spectrum data;
[0109] The data feature extraction module maps the preprocessed spectral data to the feature space, extracts the weight coefficients mapped to each feature vector in the feature space, and uses the measured theaflavins, tea polyphenols, and amino acid content data as feature variables;
[0110] The data splicing module splices the characteristic vector of the spectral data with the characteristic variable to obtain a comprehensive characteristic vector;
[0111] Model building module, to build an intelligent model for distinguishing the degree of tea fermentation;
[0112] The model training module inputs the comprehensive feature vector and the corresponding tea fermentation degree label data into the tea fermentation degree intelligent discrimination model for training and verification, thereby obtaining a verified tea fermentation degree intelligent discrimination model;
[0113] The discrimination result output module subjects the tea samples to spectral scanning and chemical analysis, and inputs the feature vector to be discriminated after preprocessing and feature extraction into the verified intelligent discrimination model of tea fermentation degree to obtain the discrimination result of tea fermentation degree.
[0114] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0115] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. An intelligent method for distinguishing the fermentation degree of tea leaves based on near infrared spectroscopy, characterized in that: include: Use a near-infrared spectrometer to perform spectral scanning on tea samples with different fermentation degrees to obtain near-infrared spectral data; The contents of theaflavins, tea polyphenols and amino acids in tea samples were determined by chemical analysis methods; Preprocessing the acquired near infrared spectrum data; The preprocessed spectral data is mapped to the feature space, the weight coefficients mapped to each feature vector in the feature space are extracted, and the measured theaflavins, tea polyphenols and amino acid content data are used as feature variables; The eigenvector of the spectral data is concatenated with the eigenvariable to obtain a comprehensive eigenvector; Establish an intelligent identification model for tea fermentation degree; The comprehensive feature vector and the corresponding tea fermentation degree label data are input into the tea fermentation degree intelligent discrimination model for training and verification, so as to obtain the verified tea fermentation degree intelligent discrimination model; After the tea samples to be tested are subjected to spectral scanning and chemical analysis, the feature vector to be discriminated after preprocessing and feature extraction is input into the verified intelligent discrimination model of tea fermentation degree to obtain the discrimination result of tea fermentation degree.
2. The method for intelligently distinguishing the degree of tea fermentation based on near infrared spectroscopy according to claim 1, characterized in that: The preprocessing includes first using a residual mean-variance distribution diagram based on Monte Carlo cross-validation to eliminate abnormal samples, then randomly selecting more than 120 samples from the data set from which abnormal samples have been eliminated as training sets and validation sets, and using the Kernard-Stone method to divide the training set and the validation set into a ratio of 3:1; and performing a second-order derivative on the spectral data of the training set.
3. The intelligent method for distinguishing the degree of tea fermentation based on near infrared spectroscopy according to claim 1, characterized in that: The formula to extract the mapping is: y i =w T x i ; Among them, w represents the characteristic matrix, and the optimization equation for solving w is as follows: J(w) represents the cost function of the optimization equation, In the formula, η represents the regularization parameter; x i represents the result of the second-order derivative of the near-infrared spectrum vector of the i-th sample, and x i =[x i (λ1)…x i (λ k )] T , the subscript k represents the dimension of the sample spectrum; y i For x i The weight coefficient projected onto the eigenvectors that constitute the feature matrix w determines the position of sample i in the feature space; the continuous variable u i ∈[0,1] represents x i The probability of belonging to a non-noise sample; (1-u i ) represents x i The probability of belonging to a noise sample; the fuzzy variable m∈[1,∞) represents the weight index; e(x i ) represents the original spectrum x i The error between the feature extraction and the reconstructed spectrum is 4. The method for intelligently distinguishing the degree of tea fermentation based on near infrared spectroscopy according to claim 3, characterized in that: Continuous variable u i The constraint formula is as follows: According to the constraint formula, we can get u i , whose expression is: According to the calculated u i The simplified formula of the optimization equation J(w) is:
5. The method for intelligently distinguishing the degree of tea fermentation based on near infrared spectroscopy according to claim 1, characterized in that: The specific method of establishing the intelligent discrimination model for the fermentation degree of tea is as follows: using a support vector machine based on a Gaussian radial basis function as a classifier and combining it with a weight coefficient to establish the intelligent discrimination model for the fermentation degree of tea.
6. The method for intelligently distinguishing the fermentation degree of tea leaves based on near infrared spectroscopy according to claim 1, characterized in that: The method also includes evaluating the performance of the intelligent discrimination model for tea fermentation degree by using classification error rate and accuracy, specifically including: Assume the sample set is A, and the classification error rate is: The accuracy is given by: Where I(·) is the indicator function, which takes 1 when · is true and takes 0 when it is false; m is the total number of samples; o(p i ) is the predicted value, q i is the true sample value.
7. An intelligent identification system for tea fermentation degree based on near infrared spectroscopy, characterized in that: The method for intelligently distinguishing the fermentation degree of tea leaves based on near infrared spectroscopy according to any one of claims 1 to 6 comprises: The data acquisition module uses a near-infrared spectrometer to perform spectral scanning on tea samples of different fermentation degrees to obtain near-infrared spectral data; the content of theaflavins, tea polyphenols, and amino acids in the tea samples is determined by chemical analysis methods; A data preprocessing module is used to preprocess the acquired near infrared spectrum data; The data feature extraction module maps the preprocessed spectral data to the feature space, extracts the weight coefficients mapped to each feature vector in the feature space, and uses the measured theaflavins, tea polyphenols, and amino acid content data as feature variables; The data splicing module splices the characteristic vector of the spectral data with the characteristic variable to obtain a comprehensive characteristic vector; Model building module, to build an intelligent model for distinguishing the degree of tea fermentation; The model training module inputs the comprehensive feature vector and the corresponding tea fermentation degree label data into the tea fermentation degree intelligent discrimination model for training and verification, thereby obtaining a verified tea fermentation degree intelligent discrimination model; The discrimination result output module subjects the tea samples to spectral scanning and chemical analysis, and inputs the feature vector to be discriminated after preprocessing and feature extraction into the verified intelligent discrimination model of tea fermentation degree to obtain the discrimination result of tea fermentation degree.
8. The intelligent identification system of tea fermentation degree based on near infrared spectroscopy according to claim 7, characterized in that: The data preprocessing module includes first using the residual mean-variance distribution diagram based on Monte Carlo cross-validation to eliminate abnormal samples, then randomly selecting more than 120 samples from the data set from which the abnormal samples are eliminated as training sets and validation sets, and using the Kernard-Stone method to divide the training set and the validation set in a ratio of 3:1; and performing second-order derivative on the spectral data of the training set.
9. The intelligent identification system of tea fermentation degree based on near infrared spectroscopy according to claim 7, characterized in that: It also includes a model performance evaluation module, which uses classification error rate and accuracy to evaluate the performance of the intelligent discrimination model for tea fermentation degree, including: Assume the sample set is A, and the classification error rate is: The accuracy is given by: Where I(·) is the indicator function, which takes 1 when · is true and takes 0 when it is false; m is the total number of samples; o(p i ) is the predicted value, q i is the true sample value.
Citation Information
Cited By
Tea leaf pile fermentation degree discrimination method and discrimination system based on near infrared spectrum
CN122113000A
A method and system for determining the degree of tea fermentation based on near-infrared spectroscopy
CN122113000B
